jfa journal

AND option

OR option



A.T. Olagunju1,2,3, J.A. Morgan2, A. Aftab4, J.R. Gatchel5, P. Chen6, A. Dols7,8, M. Sajatovic6, W.T. Regenold9


1. Department of Psychiatry and Behavioral Neurosciences, McMaster University/St Joseph’s Healthcare Hamilton, Hamilton, ON, Canada; 2. Discipline of Psychiatry, The University of Adelaide, Adelaide, SA, Australia; 3. Department of Psychiatry, College of Medicine, University of Lagos, Lagos, Nigeria; 4. Department of Psychiatry, University of California, San Diego, CA, USA; 5. Division of Geriatric Psychiatry, McLean Hospital, Harvard Medical School, Boston, MA, USA; 6. Departments of Psychiatry & Neurology, Case Western Reserve University School of Medicine, University Hospitals Cleveland Medical Centre, Cleveland, OH, USA; 7. GGZ inGeest Specialized Mental Health Care, Department of Old Age Psychiatry, 1081 HJ, Amsterdam, The Netherlands; 8. Amsterdam UMC, Vrije Universiteit Amsterdam, Psychiatry, Amsterdam Public Health research institute and Neuroscience Amsterdam, 1081 HV, Amsterdam, The Netherlands; 9. Noninvasive Neuromodulation Unit, Experimental Therapeutics and Pathophysiology Branch, Division of Intramural Research Program, National Institute of Mental Health, National Institutes of Health, Bethesda, MD, USA.
Corresponding author: Andrew T. Olagunju, Department of Psychiatry and Behavioral Neurosciences, McMaster University/St Joseph’s Healthcare Hamilton, 100 West 5th Street, Hamilton, ON, L8N 3K7 Canada, Email:olagunja@mcmaster.ca

J Frailty Aging 2020;in press
Published online December 9, 2020, http://dx.doi.org/10.14283/jfa.2020.64


Objectives: To better understand the role of nutrition in older adults (aged 50 years or older) with bipolar disorders (OABD), we conducted a systematic review of the literature and appraise existing evidence. Methods: Following PRISMA guidelines, we searched databases including Medline/PubMed, PsychINFO, EMBASE, CINAHL, Scopus, Web of Science, Cochrane Register, FDA website, and clinical trial registries through 2019 for eligible reports. The search string combined MeSH terms for bipolar disorder, nutrition and older adults. This was supplemented by snowball searching of references in relevant studies and authors were contacted to request their work where necessary. All included studies were rated with the National Institutes of Health Study Quality Assessment Tools based on study designs. Results: Of 2280 papers screened, ten studies including eight observational and two interventional studies. The topic foci of the papers examined several nutrients, (including vitamin B12, vitamin D, coenzyme Q10, homocysteine, and folate), nutritional deficiencies and biochemical correlates. The prevalence rates of deficiencies varied with specific nutrients (3.7% to 71.6% for Vitamin B12 and 34.6% for Vitamin D), and between inpatient versus outpatient populations. While nutritional interventions appeared to be associated with improvement in both affective and cognitive outcomes, the sample sizes of OABD varied and were generally small. Conclusion: While there is evidence for the benefits of nutritional interventions on affective, cognitive and overall outcome in OABD, the quality of the evidence is limited. Our findings underscore the need for high quality studies to inform evidence-based guidelines for nutritional assessment and supplemention in OABD.

Key words: Bipolar disorder, depression, geriatrics, nutrition, older adults, mania.



Recent international collaborative research indicates that suboptimal diet contributes significantly to the global burden of diseases with 11 million deaths and loss of 255 million disability-adjusted life-years attributable to dietary risk factors in 2017 (1-3). Despite the evidence linking poor nutrition to disease, nutrition and nutritional supplements are often overlooked in the assessment and treatment of patients generally, and in individuals with bipolar disorder (BD) in particular.
Older adults with bipolar disorder (OABD) may be especially prone to nutritional deficiencies due to various age-related factors that can diminish nutrient intake and absorption. Reduced dietary intake of nutrients may occur due to decreased sensory function, decreased appetite and dysphagia with advanced age (4). Causes of decreased nutrient absorption that are more common in older adults include: age related changes in the gut or atrophic gastritis; peptic ulcer disease with Helicobacter pylori infection; gastric and intestinal resections and taking medications that can interfere with nutrient absorption. Many of these interfering medications are commonly prescribed and include inhibitors of gastric acid secretion such as proton pump inhibitors and histamine-2 receptor antagonists; anti-epileptics, metformin, methotrexate, triamterene and trimethoprim (5, 6). Importantly, nutritional status, particularly vitamin levels, can influence cognitive function and affective symptoms in older adults with mood disorders, including BD (7-10).
In addition to the lack of awareness of diet and nutritional deficiencies in the care of OABD, clinicians are often unaware of the increasing use of nutritional supplements by their patients. Herbal and nutritional compounds are used widely by older adults and have been reported to be taken by nearly one in three older adults with bipolar disorder or major depression (11). Most individuals who use these supplements do not mention this to their clinicians, putting them at risk for potential drug-supplement interactions with adverse health effects.
Given the clinical importance of nutritional deficiencies and nutritional supplement use, a working group within the International Society for Bipolar Disorders (ISBD) (12) OABD taskforce undertook a systematic review with the following objectives in mind: 1) to synthesize literature evidence on nutritional deficiencies and supplements in OABD; 2) to review the quality of research evidence on nutritional deficiencies and supplements in relation to the epidemiology, pathophysiology, clinical treatment, outcome and wellbeing of OABD, and 3) to make recommendations regarding further research that will promote evidence-based assessment and management of OABD.



Eligibility criteria

We followed the preferred reporting items for systematic reviews and meta-analyses (PRISMA) guidelines in conducting the literature search in this review (13, 14). All literature for all years through 2019 was searched. The age cut-off for OABD for eligible studies in this review is 50 years and older based on the knowledge that around 50 years adults already have decreased absorption of nutrients, which is a primary age-related cause of nutritional deficiency (15, 16). Screening for nutrient deficiencies has been recommended in people aged 50 years and older in the United States and elsewhere in the world (15-18). We included all study designs including retrospective, observational, cross-sectional, and intervention trial but excluded case reports. Other inclusion criteria were availability of study full-text in the English language,19 and a clearly described representation of OABD among the study participants.

Information sources

We searched databases including Medline/PubMed, PsychINFO, EMBASE, Web of Science, Cochrane Register, the Food and Drug Administration (FDA) website, and international clinical trial registries for all years through 2019 for eligible reports. The bibliographies of included studies and relevant reviews were snow-ball searched for additional studies. Study authors were also contacted to request their work where necessary.

Search strategy

We used a search strategy addressing the following key terms in various permutations: nutrition, older adults, and bipolar disorder combined with OR and AND. Search strings included “diet* OR Diet, Food, and Nutrition OR food OR nutrition* OR “nutritional status” OR “nutritional physiological phenomenon” OR micronutrient* vitamin* OR multivitamin* OR vitamin B OR folate OR B6 OR B12 OR niacin OR vitamin C OR vitamin D OR vitamin E OR calcium OR chromium OR iron OR magnesium OR zinc OR artemisin OR caffeine OR dehydroepiandrosterone OR DHEA OR echinacea OR fish oil OR GABA OR gamma-amino butyric acid OR garlic OR ginkgo OR glucosamine OR karaela OR melatonin OR methylsulfonylmethane OR primrose oil OR probiotics OR S-adenosylmethionine OR St. Johns wort OR tryptophan OR tyrosine OR valerian AND “old age” OR aged OR elder* OR geriatric* OR geriatric psychiatry OR older OR aging OR ageing AND “bipolar disorder” OR manic depressi* OR bipolar depression. Search strings were performed in titles, abstracts, major topic/subject headings, and MeSH headings.

Study selection

Study selection, review of included studies and data extraction were conducted by Andrew T. Olagunju (ATO), Julie A. Morgan (JAM), Awais Aftab (AA), and supervised by William T. Regenold (WTR). Titles and abstracts were screened independently by at least two authors to shortlist studies for further review. The full texts of the shortlisted studies were reviewed by at least two authors independently according to the inclusion criteria. Inconsistencies about inclusion or exclusion of studies were resolved by discussion between authors or in consultation with WTR or other members of the task force to reach consensus.

Quality/bias assessment

As studies with different designs and representation of older adults were eligible, the Study Quality Assessment Tools of the National Institutes of Health (NIH) for the assessment of Observational Cohort and Cross-Sectional Studies, Case-Control, and Controlled Intervention Studies were used to assess the quality of included studies.20 We rated individual studies on a range of 12-14 items depending on their designs to produce a comprehensive overview of the risk of bias in individual studies and highlighted relevant overall quality limitations.

Data collection process

Data items collected from eligible reports included study characteristics, nutrients, main findings and implications for OABD (Table 1). In total, we screened 2280 titles and abstracts to produce a shortlist of 92 potential reports for full-text review. Of these 92 reports, 10 studies were selected for inclusion in the final review. (See Figure 1)

Table 1
Studies included in the review examining nutrition in older adults with bipolar disorder

Legend: BD- bipolar disorders, BCAA- branched-chain amino acid, C-cross-sectional, CC-case control, MTHFR- Methylene tetrahydrofolate reductase, N-total number of sample, NR-not reported, OABD- Older adults with bipolar disorders, R-retrospective chart review, RCT-randomized controlled clinical trial, YMRS- Young Mania Rating Scale

Figure 1
Flow of studies through the systematic review



Study characteristics

Ten studies were included in the review (5, 7, 8, 10, 11, 21-24, 25). The publication dates of these reports spanned three decades from 1990 to 2015. Of the ten studies, six were retrospective chart reviews, two were cross-sectional studies, and the remaining two were clinical trials. Considering the study participants, in two studies all participants were adults diagnosed with BD with a representation of OABD, (21, 22) while in the remaining studies (n=8), participants with BD represented a subset of the total sample. In general, the topic foci in the included studies addressed different aspects of nutrition and nutritional deficiencies, including biochemical, epidemiological and clinical factors, and nutritional supplements in OABD. All were able to specify the sample sizes of OABD in the included studies aged 50 years or older, ranging from 3-50. Follow-up was limited as the majority of the studies reported cross-sectional observations. (See table 1)

Biochemical, clinical and epidemiological findings on nutrients

Table 2 presents findings on specific nutrients and nutritional supplements covered in the ten studies included in this review. In total, four classes of nutrients including vitamins, minerals, and dietary herbs and nutritional supplements are covered in the studies and described below.

Table 2
Quality assessment with NIH scale-tool based on study design

1-14 are quality assessment items/criteria; CCS- case-controlled studies; CD- cannot determine; CIS- controlled intervention studies, NIH- National Institute of Health; NA- not applicable; NR-not reported; OCCSS- observational cohort and cross-sectional studies; OR-overall rating; SD-study design.



Five retrospective chart reviews (5, 7, 8, 10, 23) focused on vitamin B12 and/or folate. Importantly, several reports looked at vitamin B12 deficiency in mixed populations of older adults including patients with BD. The prevalence rates of Vitamin B12 deficiency ranged from 3.7%8 to 71.6%.10 Lower levels of vitamin B12 correlated strongly with memory loss and poor cognitive performance in psychotic depression (8). Approximately one quarter of geriatric clinic outpatients with B12 deficiency had neuropsychiatric symptoms with behavioural disturbance, memory loss and sensorimotor disorders being most common (10). Folate deficiency was reported to be 1.3% in acute geropsychiatric inpatients (7). One study concluded that the biochemically interrelated vitamins, B12 and folate, may exert both separate and concomitant influences on affect and cognition and that poorer vitamin status may contribute to certain geropsychiatric disorders that have later life onset and lack a familial predisposition (7).
Another study investigated the relationships between Vitamin B12 and folate levels as well as homocysteine (HCY) levels on brain volume changes and brain white matter disease in geriatric patients with psychiatric disorders (23). These authors found that, low serum concentrations of folate, but not of B12, were associated with magnetic resonance imaging (MRI) evidence of brain white-matter disease and smaller hippocampus and amygdala brain-volume measures. Elevated HCY, which can result from folate deficiency, was also associated with MRI evidence of brain white matter disease (23). Finally, one study found that among older adult psychiatric inpatients, B12 serum levels and percentages of probable and possible B12 deficiency did not differ significantly between cognitive disorder patients and non-cognitive disorder patients, including OABD, suggesting that it was reasonable to routinely screen older adult psychiatric inpatients for B12 deficiency whether or not cognitive disorder symptoms were present (5). One retrospective chart review study (24) and cross-sectional study(25) addressed vitamin D in OABD. Vitamin D deficiency prevalence rates ranged from 3.7 % to 34.6%. Vitamin D insufficiency (less severe and defined as levels <30 ng/mL) was reported in approximately 75% of psychiatric inpatients in two cross sectional assessments (25). However, no associations were found between vitamin d levels and a screen of global cognitive function or psychiatric diagnoses (24, 25).

Dietary, Herbs and Nutritional Supplements

In one cross sectional study of OABD and older adults with major depression (11), herbs and nutraceutical (HNC) products were ingested by 30% of individuals. About 40% incorrectly believed that HNC products were FDA-regulated at that time. The majority (64%) had not discussed the use of HNC with treating physicians and some preferred to take HNC compared to physician-prescribed psychotropic medications (14-20%). The use of HNC was more common in those with BD (44%) than those with unipolar major depression (16%).
Two open label trials (21, 22) reported significant reduction in the severity of depressive symptoms with high-dose CoEnzyme Q10 in OABD (21). These findings could support a role of therapy targeted at mitochondrial function in affective disorders (22).

Study Quality Assessment

Quality assessment of all included studies (n = 10) was rated fair (n=8) and good (n=2) with risk of bias items included in the tool. The limited representation of OABD represented a risk of bias. Notwithstanding the limitations in the quality of the studies, the reported outcomes on nutritional deficiencies and supplements in OABD were considered the best available evidence for the recommendations made. (See table 2)



Clinical implications

Most of the studies reviewed—seven of ten studies (70%) focused on three vitamins—vitamin B12, vitamin D and folate and the consequences of their deficiencies. These studies support a relationship between vitamin B12 deficiency and both affective symptoms and cognitive impairment in OABD (7, 8). They also provide evidence for associations between folate levels and hippocampus and amygdala volumes and brain MRI white matter hyper- intensities (WMHs) in older adult psychiatric patients (23). Vitamin D deficiency was reported to be common in older adult psychiatric patients. Generally, vitamin deficiencies did not differ across psychiatric disorders, indicating that screening for these deficiencies in older adults should not be limited to particular diagnostic groups such as cognitive disorders. Evidence for a therapeutic role for vitamin supplementation was limited. There is some evidence for a benefit from folate supplementation for OABD taking lithium.
Tissue levels of homocysteine can be elevated due to vitamin B12 or folate deficiencies. Four studies examined the relationships among vitamin B12, folate and homocysteine levels and their associations with psychiatric symptoms or evidence of brain disease (23). These studies support an inverse relationship between B12 and folate blood levels on one hand and homocysteine blood levels on the other in individuals with bipolar depression. They also suggest that HCY blood levels are elevated in individuals with BD and support an association between elevated levels and impairment in executive function and in the extent of brain WMHs. These studies provide evidence for measuring HCY blood levels along with vitamin B12 and folate levels in OABD.
Studies of diet and nutritional supplements revealed that supplement use is very common among OABD (11). Furthermore, supplement use was typically not discussed with health providers (11).
Clinical trials of adjunctive nutritional supplements found preliminary evidence for a benefit from Coenzyme Q10 on depressive symptoms in OABD (21, 22). The positive results in these supplement studies require replication in a larger group of OABD prior to their consideration for routine clinical use in OABD.

Study limitations

There are several study limitations of this review. The included studies were not homogenous nor entirely focused on OABD. Of the ten articles included in the review, only two reported study results exclusively from a sample or population of OABD. By expanding our inclusion criteria to capture studies with some significant number of OABD participants, we were able to report a more comprehensive review of studies. The majority of these studies were retrospective chart reviews and cross-sectional studies, while the number of RCTs was limited. Consequently, inferences on causal relationships between the nutrients studied and the pathophysiology or treatment of OABD is limited. Furthermore, none of the included studies (n=10) looked at the potential contributions of gender and the type of bipolar disorder (I and II) to their study findings despite the clinical importance of these two factors (26).



Scientific literature on nutrition and nutritional supplements in OABD exists but is limited in both quantity and quality. However, we can draw several conclusions from this review. First, because older adults are prone to nutritional deficits and suffer from several vitamin deficiencies that influence cognition and affective symptoms, there should be consideration given to better research to develop clinical guidelines for routine screening for deficiencies of vitamin B12, folate and vitamin D in OABD. Second, screening for elevated blood levels of HCY that result from vitamin deficiencies should also be considered. Notably there is need for well-designed and powered clinical trials to examine the effects of nutritional interventions including dietary questionnaires, vitamin deficiency screening, HCY blood level screening, and adjunctive nutrient supplementation in OABD.


• Nutrition is critical to physical health and general well-being of older adults with bipolar disoder (OABD), however nutritional deficiencies are common albeit varies for specific nutrients.
• While there is evidence for the benefits of nutritional interventions on affective, cognitive and overall outcome in OABD, the quality of the evidence is limited.
• Findings in this review underscore the need for high quality studies for development of evidence-based guideline for assessing and supplementing nutrition in OABD.


Acknowledgements: The authors thank the International Society for Bipolar Disorders executives, and staff for their support for this task force. Our gratitude also goes to Ms Maureen Bell of the University of Adelaide library for her support and advice.
Conflicts of interest: The authors declare no conflicts of interest to the content of the manuscript.
Disclosure information: Financial Disclosure information for Martha Sajatovic MD are stated below: Research grants within past 3 years: Otsuka, Alkermes, Janssen, International Society for Bipolar Disorders, Reuter Foundation, Woodruff Foundation, Reinberger Foundation, National Institutes of Health (NIH), Centers for Disease Control and Prevention (CDC); Consultant: Alkermes, Bracket, Otsuka, Janssen, Neurocrine, Health Analytics; Royalties: Springer Press, Johns Hopkins University Press, Oxford Press, UpToDate; CME activities: American Physician’s Institute, MCM Education, CMEology, Potomac Center for Medical Education, Global Medical Education, Creative Educational Concepts
Financial Disclosure information: Dr. Gatchel reports other relevant financial activities from Merck, grants from Alzheimer’s Association, grants from BrightFocus Foundation, outside the submitted work.



1. GBD 2017 Diet Collaborators. Health effects of dietary risks in 195 countries, 1990-2017: a systematic analysis for the Global Burden of Disease Study 2017. Lancet (London, England). 2019;393(10184):1958-72.
2. GBD 2016 Risk Factors Collaborators. Global, regional, and national comparative risk assessment of 84 behavioural, environmental and occupational, and metabolic risks or clusters of risks, 1990-2016: a systematic analysis for the Global Burden of Disease Study 2016. Lancet (London, England). 2017;390(10100):1345-422.
3. GBD 2017 Risk Factor Collaborators. Global, regional, and national comparative risk assessment of 84 behavioural, environmental and occupational, and metabolic risks or clusters of risks for 195 countries and territories, 1990-2017: a systematic analysis for the Global Burden of Disease Study 2017. Lancet (London, England). 2018;392(10159):1923-94.
4. Schiffman SS. Intensification of sensory properties of foods for the elderly. J Nutr. 2000;130(4S Suppl):927S-30S.
5. Lachner C, Martin C, John D, Nekkalapu S, Sasan A, Steinle N, et al. Older adult psychiatric inpatients with non-cognitive disorders should be screened for vitamin B12 deficiency. Journal of Nutrition, Health and Aging. 2014;18(2):209-12.
6. Langan RC, Goodbred AJ. Vitamin B12 Deficiency: Recognition and Management. Am Fam Physician. 2017;96(6):384-9.
7. Bell IR, Edman JS, Marby DW, Satlin A, Dreier T, Liptzin B, et al. Vitamin B12 and folate status in acute geropsychiatric inpatients: affective and cognitive characteristics of a vitamin nondeficient population. Biol Psychiatry. 1990;27(2):125-37.
8. Bell IR, Edman JS, Miller J, Hebben N, Linn RT, Ray D, et al. Relationship of normal serum vitamin B12 and folate levels to cognitive test performance in subtypes of geriatric major depression. J Geriatr Psychiatry Neurol. 1990;3(2):98-105.
9. Coppen A, Chaudhry S, Swade C. Folic acid enhances lithium prophylaxis. J Affect Disord. 1986;10(1):9-13.
10. Issac TG, Soundarya S, Christopher R, Chandra SR. Vitamin B12 deficiency: an important reversible co-morbidity in neuropsychiatric manifestations. Indian journal of psychological medicine. 2015;37(1):26-9.
11. Keaton D, Lamkin N, Cassidy KA, Meyer WJ, Ignacio RV, Aulakh L, et al. Utilization of herbal and nutritional compounds among older adults with bipolar disorder and with major depression. Int J Geriatr Psychiatry. 2009;24(10):1087-93.
12. (ISBD) ISoBD. https://www.isbd.org/active-task-forces. 2019.
13. Liberati A, Altman DG, Tetzlaff J, Mulrow C, Gotzsche PC, Ioannidis JP, et al. The PRISMA statement for reporting systematic reviews and meta-analyses of studies that evaluate health care interventions: explanation and elaboration. Annals of internal medicine. 2009;151(4):W65-94.
14. Moher D, Liberati A, Tetzlaff J, Altman DG. Preferred reporting items for systematic reviews and meta-analyses: the PRISMA statement. Annals of internal medicine. 2009;151(4):264-9, w64.
15. . Hughes CF, Ward M, Hoey L, McNulty H. Vitamin B12 and ageing: current issues and interaction with folate. Ann Clin Biochem. 2013 Jul;50(Pt 4):315-29.
16. Langan RC, Goodbred AJ. Vitamin B12 Deficiency: Recognition and Management. Am Fam Physician. 2017 Sep 15;96(6):384-389.
17. Government of South Wales, Australia: A Practical Guide to Effective Engagement with Older People. Retrieved on 27/01/2020 from: https://www.sahealth.sa.gov.au/wps/wcm/connect/efc56a004efc69f1b7ccf79ea2e2f365/Better+Together+-+A+Practical+Guide+to+Effective+Engagement+with+Older+People.pdf?MOD=AJPERES&CACHEID=ROOTWORKSPACE-efc56a004efc69f1b7ccf79ea2e2f365-mMADUnH
18. Institute of Medicine Panel on folate, other B vitamins and choline. Dietary Reference Intake: Thiamin, Riboflavin, Niacin, Vitamin B6, Vitamin B12, Pantothenic Acid, Biotin and CholineWashington DC: National Academy Press, 1998, pp.306–56.
19. Bauer M, Glenn T, Conell J, Rasgon N, Marsh W, Sagduyu K, et al. Common use of dietary supplements for bipolar disorder: a naturalistic, self-reported study. Int J Bipolar Disord. 2015;3(1):29.
20. NIH National Heart LaBIS, quality assessment tools [cited 2017 M. Available from: https://wwwnhlbinihgov/health-pro/guidelines/in-develop/cardiovascular-risk reduction/tools. 2017
21. Forester BP, Harper DG, Georgakas J, Ravichandran C, Madurai N, Cohen BM. Antidepressant effects of open label treatment with coenzyme Q10 in geriatric bipolar depression. Journal of clinical psychopharmacology. 2015;35(3):338-40.
22. Forester BP, Zuo CS, Ravichandran C, Harper DG, Du F, Kim S, et al. Coenzyme Q10 effects on creatine kinase activity and mood in geriatric bipolar depression. J Geriatr Psychiatry Neurol. 2012;25(1):43-50.
23. Scott TM, Tucker KL, Bhadelia A, Benjamin B, Patz S, Bhadelia R, et al. Homocysteine and B vitamins relate to brain volume and white-matter changes in geriatric patients with psychiatric disorders. Am J Geriatr Psychiatry. 2004;12(6):631-8.
24. Lapid MI, Drake MT, Geske JR, Mundis CB, Hegard TL, Kung S, et al. Hypovitaminosis D in psychogeriatric inpatients. J Nutr Health Aging. 2013;17(3):231-4.
25. Gronli O, Kvamme JM, Jorde R, Wynn R. Vitamin D deficiency is common in psychogeriatric patients, independent of diagnosis. BMC Psychiatry. 2014;14.
26. Müller JK, Leweke FM. Bipolar disorder: clinical overview. Med Monatsschr Pharm. 2016 Sep;39(9):363- 9.





1. Department of Geriatrics, CHU Université Catholique de Louvain, Namur, Belgium; 2. Institute of Health and Society, Université Catholique de Louvain, Brussels, Belgium; 3. Departments of Public Health and Primary Care, Katholieke Universiteit Leuven, Leuven, Belgium; 4. Clinical Pharmacy Research Group, Louvain Drug Research Institute, Université catholique de Louvain, Brussels, Belgium; 5. Scientific Support Unit, CHU Université Catholique de Louvain, Namur, Belgium; 6. Unit of Research on Cellular Biology, NARILIS-Namur Research Institute for Life Sciences, University of Namur, Belgium; 7. GIGA Research Institute, University of Liège, Liège, Belgium.
Corresponding author: Florence Potier M.D., Department of Geriatrics, Centre Hospitalier Universitaire Université Catholique de Louvain Namur, 1, rue Dr G. Therasse, 5530 Mont-Godinne, Belgium. Tel 0032/81422175. Fax: 0032/81423885. Email: florence.potier@uclouvain.be

J Frailty Aging 2018;in press
Published online April 26, 2018, http://dx.doi.org/10.14283/jfa.2018.11



Background: Evidence suggests that providing care for a disabled elderly person may have implications for the caregiver’s own health (decreased immunity, hypertension, and depression). Objective: Explore if older spousal caregivers are at greater risks of frailty compared to older people without a load of care. Design: Case-control study. Setting: Participants were assessed at home in Wallonia, Belgium. Participants: Cases: community-dwelling spousal caregivers of older patients, recruited mainly by the geriatric outpatient clinic. Controls: people living at home with an independent spouse at the functional and cognitive level matched for age, gender and comorbidities. Measurements: Mini nutritional assessment-short form (MNA-SF), short physical performance battery (SPPB), frailty phenotype (Fried), geriatric depression scale (GDS-15), clock drawing test, sleep quality, and medications. The multivariable analysis used a conditional logistic regression. Results: Among 79 caregivers, 42 were women; mean age and Charlson comorbidity index were 79.4±5.3 and 4.0±1.2, respectively. Among care-receivers (mean age 81.4±5.2), 82% had cognitive impairment. Caregiving was associated with a risk of frailty (Odd Ratio (OR) 6.66; 95% confidence interval (CI) 2.20-20.16), the consumption of antidepressants (OR 4.74; 95% CI 1.32 -17.01), shorter nights of sleep (OR 3.53; 95% CI 1.37-9.13) and more difficulties maintaining a social network (OR 5.25; 95% CI 1.68-16.40). Conclusions: Spousal caregivers were at an increased risk of being frail, having shorter nights of sleep, taking antidepressants and having difficulties maintaining their social network, compared to non-caregiver controls. Older spousal caregivers deserve the full attention of professionals to prevent functional decline and anticipate a care breakdown.

Key words: Caregiving, frailty, nutrition, depression.



In 2014, 34% of the population in 20 European countries provided care to a family member (1). Among older couples, spouses are first in line to provide care (2), and they are key elements in the home support of dependent older persons (3).
As we followed our patients in geriatric consultations, it seemed to us that spouses gave care until exhaustion. In this context, we wondered how to assess the health of older spousal caregivers.
Informal care could be associated with a reduction in physical and mental health. For instance, evidence suggests that caregivers have decreased immunity observed with a lower antibody response to vaccines (4, 5), have higher rates of hypertension and coronary diseases (6, 7) and are at greater risk for depression and anxiety (8).
In later life, these effects on health may lead to increased frailty (9). The frailty phenotype is generally defined as a decline in homeostatic reserves in multiple physiological systems, resulting in greater vulnerability (10). Frailty as a phenotype has significant public health relevance because frail older adults have a greater risk of falls, disability, hospitalization, institutionalization, and death (10, 11). These risks could anticipate a care breakdown and therefore could result in hospitalizations or nursing home admission of the care-receiver.
To date, little is known about the risk of frailty among spousal caregivers. Therefore, the purpose of this study is to explore if older spousal caregivers are at greater risk for frailty compared to older people without a load of care.



This was a cross-sectional analysis of the baseline data from a cohort study of older spousal caregivers.


Caregivers were recruited through the geriatric outpatient clinic and the memory centre of the University Hospital of Louvain in Namur and because of the efforts of general practitioners and home nurses from March 2015 until May 2016. Cases were defined as spouses of older patients with a cognitive deficit (score of more than 2/7 on the Global Deterioration Scale (12)) or functional impairment (a minimum dependence of 1 activity in daily living) who were still living at home. Controls were people living at home with an independent spouse at the functional and cognitive level. Controls were recruited with the help of general practitioners and home nurses and via senior associations. Controls were matched for age, gender, and comorbidity.
All participants had to be 70 years or older. All provided written informed consent that was approved by the CHU UCL Namur Institutional Review Board (NUB: B039201422799).

Socio-demographic data

Participants were assessed at home. The following data were collected: age, gender, educational level and income level. Furthermore, medical history was taken including any medications, smoking status, alcohol consumption, sleep duration and sleep perturbation. Socioeconomic status was constructed as suggested by Cockerham (13). A total score was calculated as the sum of education (primary school=1; unfinished secondary education=2; secondary education=3; higher education=4), income (difficult=1; easy=2) and past occupation (worker-farmer-unemployed=1; craftsman-self-employed=2; employee-officer=3, manager-liberal profession=5). Questions also addressed their social network including family, friends or neighbours on a 5-point Likert-type scale (1 = very difficult to maintain contact; 5 = very easy to maintain contact).

Medical data

Comorbidity was evaluated with the Charlson Comorbidity Index (14). The occurrence of falls, hospitalizations and the numbers of medications were recorded. Body mass index (BMI) was computed as the ratio between weight in kilograms and height in square metres. The short physical performance battery (15) assessing lower extremity function included balance (ability to stand in tandem positions), gait speed (time to walk 4 metres) and strength (time to rise from a chair and return to the seated position 5 times). Nutrition was assessed with the mini Nutritional Assessment short form (16); a risk of malnutrition was considered for total scores below 12/14. The frailty phenotype was assessed according to the definition of L. Fried (10), a pre-frail status considered for a total score of 1 or 2 out of 5 and a frail status for a total score above 2/5. The grip strength of the dominant hand was measured with the Martin vigorimeter. The highest score of the three trials was retained. Finally, cognitive status was evaluated with the clock drawing test (17).

Psychosocial data

The sense of coherence of the participants was measured with the Sense of Coherence Scale (SOC-13), a 7-point Likert-type scale (18). SOC seems to be a health-promoting resource that strengthens resilience and develops a positive subjective state of health. Depressive symptoms were screened with the Geriatric Depression Scale-15 (19); a participant was considered at risk of depression with a score above 5/15.

Additional data of caregivers

Caregivers completed two additional self-report questionnaires. The Caregiver Reaction Assessment (CRA) (20) was used to estimate the various aspects of the caregiving situation by considering positive dimensions such as self-esteem. Caregiver burden was measured using the Zarit Burden Interview (ZBI) (Zarit et al. 1980) that consists of 22 self-reported items. We also collected the following information concerning the caregiving situation: the time spent giving care or supervision, the duration of being a caregiver, the activities of daily living they perform and the informal and professional support they have.

Medical data of care-receiver

Functional impairment was evaluated with the Katz Index (21) on a 6-point scale, with lower scores indicating greater dependence. In cognitive disorder cases, the severity of dementia was rated with the Global Deterioration Scale (12), and behavioural disturbances were screened with the Neuropsychiatric Index (22, 23). All data for the care-receiver were completed by their caregivers.

Statistical analyses

The sample size (79 caregivers and 79 controls) was calculated with an expected difference between caregivers and controls in Interleukin-6 (IL-6), a pro-inflammatory biomarker. Indeed, previous observational studies have found an association between frailty and elevated levels of pro-inflammatory mediators, such as IL-6, implicating a chronic, pro-inflammatory state in the pathogenesis of frailty (24). The biological results will not be discussed in this article.
Variables were compared between caregivers and controls using the Mac Nemar test for categorical variables and the Wilcoxon signed rank test for continuous measures. Variables that were significantly different between caregivers and controls were entered into a conditional logistic regression. Then, a stepwise selection based on the Akaike’s information criterion (AIC) was performed to select the final multivariable model. The results are presented as odd ratios (OR) and 95% confidence intervals (CI). We did not consider age, gender and comorbidities because controls were matched on these variables. Data were analysed using the SPSS statistical software package (version 24; SPSS Inc., Chicago, IL, USA) and R statistical software Version 3.3.1. (R Foundation for Statistical Computing, Vienna, Austria). Statistical tests were two-tailed, and a P-value < 0.05 was considered significant.



Descriptive analysis

A total of 79 community-dwelling spousal caregivers of older patients were recruited (49% by the geriatric outpatient clinic, 11% by the memory centre, 13% by general practitioners, 9% by home nurses, and 18% through different senior associations). The median age was 79.0 years [76-84], and the sample was almost equivalent in gender (53% of women). Care-receivers’ median age was 81 years [78-85]. A large majority (82%) of the care-receivers had cognitive impairment, and 68% had cognitive impairment with behavioural disorders. Their functional status was variable with a median of 3 [1-5] notes of 6 on the Katz ADL scale.

Table 1 Caregiver and control socio-demographic and psychological variables at baseline

Table 1
Caregiver and control socio-demographic and psychological variables at baseline

a. Socioeconomic status score was calculated as the sum of education, income and past occupation; b. Risk of depression: GDS>5/15


Univariate analysis

The comparison of caregivers and controls on socio-demographic and psychological variables is shown in Table 1 and on clinical variables on Table 2. We found nine variables that were significantly different between caregivers and controls. It was more difficult for caregivers to maintain their social network (family, friends, and neighbourhood). They had lower scores in physical performance and a higher risk of malnutrition and frailty (Table 2). Specifically, a higher number of caregivers had lost more than three kilograms in three months. Thirty percent of the caregivers were at risk for depression, and the consumption of antidepressants was also higher in caregivers than in controls. Caregivers reported more perturbed sleep and shorter nights (<8 hours) than controls. Finally, caregivers showed lower scores in sense of coherence.

Table 2 Caregiver and control clinical variables at baseline

Table 2
Caregiver and control clinical variables at baseline

a. SPPB: Short Physical Performance Battery (score 0-12); b. Cognition: pathologic clock drawing test
Caregiving, frailty, nutrition, depression.

Multivariable analysis

The results of the conditional logistic regression are shown in Table 3. Our model showed that for the same age, gender and comorbidities, caregiving is associated with a risk of frailty, the consumption of antidepressants, shorter nights of sleep and more difficulties maintaining a social network.

Table 3 Conditional logistic regression comparing caregivers and controls

Table 3
Conditional logistic regression comparing caregivers and controls

a. Social network: difficult or very difficult to maintain the social network



Exploratory analysis was made to identify the factors associated with caregivers’ frailty (Table 4). We could not find a relationship between frailty and age, comorbidity, burden or self-esteem of the caregiver. Caregivers’ frailty was not associated with cognitive or functional status of the care-receiver. However, frailty was associated with the involvement of a nurse at home (OR 5.80; 95%CI 1.71-19.64). Caregivers involved in care for more than three years were at a lower risk of frailty.

Table 4 Factors associated with caregiver frailty in logistic regression

Table 4
Factors associated with caregiver frailty in logistic regression

a. Katz Index on a 6-point scale; b. Global Deterioration Scale



Our study identified that older spousal caregivers were more likely to present with frailty, perturbed sleep, difficulties maintaining a social network and use of anti-depressive drugs than people without this load of care. After the multivariable analysis, caregivers showed a six times greater risk of being frail compared with non-caregiver controls. We screened frailty with the definition of L. Fried and focused on physical indicators including muscle strength, endurance or weight loss. More precisely, caregivers were more likely to be in a pre-frail stage (one or two present criteria), which is identified as a high risk of progressing to frailty (10). The difference was especially shown in the criteria of unintentional weight loss (more than 4.5 kg within the past year) and low physical activity. It is important to note that these two criteria are also considered in the MNA score. According to MNA, 35% of the caregivers were at risk for malnutrition. No differences were observed between male and female caregivers in nutritional status, although Puranen et al. (25) found that male gender caregivers were associated with a lower nutrient intake.
We found that caregiver frailty was associated with the attendance of a nurse at home. Home nurses are, thus, potentially well placed to detect frailty of the spouses of their patients. In contrast, giving care for more than three years was associated with more caregiver robustness. This is consistent with the study of Fredman et al., which notes that it is necessary to be healthy to remain a caregiver (26).
Caregivers reported more sleep problems than controls (perturbed sleep and shorter nights). Caregivers’ sleep deficits have already been confirmed with objective sleep assessments such as polysomnography (27, 28).
Finally, older spousal caregivers were more likely to use anti-depressive drugs. Several reviews indicated a positive association between frailty and depression (29, 30). Actually, depression and frailty share presenting symptoms, such as low daily activity profiles, that could result from exhaustion and loss of interest.
The theoretical framework underlying the assertion that caregiving is associated with a reduction in physical health is based on a model of the impact of stress on health (31). Caregivers may experience distress when their resources become insufficient (information, coping, finances, lack of respite, etc.). However, in our study, positive experiences, such as self-esteem, or on the contrary, negative experiences, such as the burden, were not associated with caregiver frailty.
The risk of frailty associated with caregiving can also be partly explained by the fact that caregivers are less likely to engage in preventive health behaviours. Indeed, in the Caregiver Health Effects Study, having a spouse with an ADL impairment predicted poor preventive health behaviours on the part of the caregiver, including not finding time for exercise, inadequate rest and forgetting to take medications (32).
A potential confounding factor that could explain the differences in health outcomes between caregivers and controls could be “assortative mating”. Assortative mating means that people select spouses that have similar lifestyle factors that may influence their risk of negative health outcome (33).
A recent paper based on the Health and Retirement Study showed that dementia caregivers were significantly more likely to experience increased frailty relative to non-dementia caregivers (34). To calculate this outcome, they developed a frailty index based on available data from a survey: chronic illnesses, ADL and IADL limitations, depression, obesity and self-rated health.
To our knowledge, our study is the first assessing the health of old spousal caregivers based on clinical data collected at participants’ homes. Furthermore, the majority of the studies assessing caregivers concern female caregivers. Our sample was almost equivalent in gender.


This study has limitations that should be considered when interpreting the results. First, because these are cross-sectional data, we cannot establish causality between caregiving and negative health outcomes. Second, both groups are convenience samples, and the recruitment methods were different between the caregiving group (more recruitment from the geriatric outpatient clinic) and the control group (more recruitment from general practitioners and senior associations). Third, the control group was matched for gender, age and comorbidities with Charlson’s index, which might not be the most precise tool to assess the comorbidity of geriatric patients. Furthermore, all data were completed by the caregivers. Lastly, this study concerns a specific caregiving subtype; spousal caregivers of geriatric patients that most suffer from cognitive deficit, thus limiting the generalizability of our results.


Healthcare providers have a role to play in the prevention of caregiver frailty, e.g., to develop advice relative to nutrition and depression screening. They should also propose solutions for home-care, allowing caregivers to share a minimum of social activities and find time for exercise.
In conclusion, many caregivers of geriatric patients are spouses who are old themselves. These spousal caregivers are at an increased risk of being frail, having shorter nights of sleep, taking antidepressants and having difficulties maintaining their social network. Older spousal caregivers deserve the full attention of professionals to prevent functional decline and anticipate a care breakdown


Acknowledgments: We would like to thank Dr. Eric Mormont from the memory centre of the University Hospital of Louvain in Namur, Aide et Soins à Domicile Namur and the general practitioners for referring participants to the project.
Funding: This work was supported by the Walloon region, Fond d’innovation sociale “Germaine Tillion” convention 1318184.  The sponsors had  no role in the design and conduct of the study, in the collection, analysis, and interpretation of data; in the preparation of the manuscript; or in the review or approval of the manuscript.
Disclosure of interest: The authors report no conflicts of interest.



1.    Verbakel E, Tamlagsronning S, Winstone L, Fjaer EL, Eikemo TA. Informal care in Europe: findings from the European Social Survey (2014) special module on the social determinants of health. Eur J Public Health 2017;27(suppl_1):90-95.
2.    Deboosere P, Demarest S, Lorant V, Miermans P-J, Portet M-I, Oyen HV. Santé et soins informels. 2001.
3.    Yaffe K, Fox P, Newcomer R, Sands L, Lindquist K, Dane K, et al. Patient and caregiver characteristics and nursing home placement in patients with dementia. Jama 2002;287(16):2090-7.
4.    Glaser R, Sheridan J, Malarkey WB, MacCallum RC, Kiecolt-Glaser JK. Chronic stress modulates the immune response to a pneumococcal pneumonia vaccine. Psychosom Med 2000;62(6):804-7.
5.    Li J, Cowden LG, King JD, Briles DA, Schroeder HW, Jr., Stevens AB, et al. Effects of chronic stress and interleukin-10 gene polymorphisms on antibody response to tetanus vaccine in family caregivers of patients with Alzheimer’s disease. Psychosom Med 2007;69(6):551-9.
6.    Shaw WS, Patterson TL, Ziegler MG, Dimsdale JE, Semple SJ, Grant I. Accelerated risk of hypertensive blood pressure recordings among Alzheimer caregivers. J Psychosom Res 1999;46(3):215-27.
7.    Capistrant BD, Moon JR, Berkman LF, Glymour MM. Current and long-term spousal caregiving and onset of cardiovascular disease. J Epidemiol Community Health 2012;66(10):951-6.
8.    Pinquart M, Sorensen S. Differences between caregivers and noncaregivers in psychological health and physical health: a meta-analysis. Psychol Aging 2003;18(2):250-67.
9.    Chang SS, Weiss CO, Xue QL, Fried LP. Association between inflammatory-related disease burden and frailty: results from the Women’s Health and Aging Studies (WHAS) I and II. Arch Gerontol Geriatr 2012;54(1):9-15.
10.    Fried LP, Tangen CM, Walston J, Newman AB, Hirsch C, Gottdiener J, et al. Frailty in older adults: evidence for a phenotype. J Gerontol A Biol Sci Med Sci 2001;56(3):M146-56.
11.    Bandeen-Roche K, Xue QL, Ferrucci L, Walston J, Guralnik JM, Chaves P, et al. Phenotype of frailty: characterization in the women’s health and aging studies. J Gerontol A Biol Sci Med Sci 2006;61(3):262-6.
12.    Reisberg B FS, de Leon MJ, Crook T. The Global Deterioration Scale for assessment of primary degenerative dementia. American Journal of Psychiatry 1982;139(9):1136-1139.
13.    Cockerham WC. Health lifestyles and the absence of the Russian middle class. Sociol Health Illn 2007;29(3):457-73.
14.    Charlson ME, Pompei P, Ales KL, MacKenzie CR. A new method of classifying prognostic comorbidity in longitudinal studies: development and validation. J Chronic Dis 1987;40(5):373-83.
15.    Guralnik JM, Simonsick EM, Ferrucci L, Glynn RJ, Berkman LF, Blazer DG, et al. A short physical performance battery assessing lower extremity function: association with self-reported disability and prediction of mortality and nursing home admission. J Gerontol 1994;49(2):M85-94.
16.    Kaiser MJ, Bauer JM, Ramsch C, Uter W, Guigoz Y, Cederholm T, et al. Validation of the Mini Nutritional Assessment short-form (MNA-SF): a practical tool for identification of nutritional status. J Nutr Health Aging 2009;13(9):782-8.
17.    Kokmen E, Smith GE, Petersen RC, Tangalos E, Ivnik RC. The short test of mental status. Correlations with standardized psychometric testing. Arch Neurol 1991;48(7):725-8.
18.    Eriksson M, Lindstrom B. Validity of Antonovsky’s sense of coherence scale: a systematic review. J Epidemiol Community Health 2005;59(6):460-6.
19.    Lesher EL, Berryhill JS. Validation of the Geriatric Depression Scale–Short Form among inpatients. J Clin Psychol 1994;50(2):256-60.
20.    Pascal Antoine SQ, Véronique Christophe. Valeurs psychométriques CRA. Vivre avec un proche malade :évaluation des dimensions positive et négative de l’expérience des aidants naturels. 2007.
21.    Katz S. Assessing self-maintenance: activities of daily living, mobility, and instrumental activities of daily living. J Am Geriatr Soc 1983;31(12):721-7.
22.    Cummings JL, Mega M, Gray K, Rosenberg-Thompson S, Carusi DA, Gornbein J. The Neuropsychiatric Inventory: comprehensive assessment of psychopathology in dementia. Neurology 1994;44(12):2308-14.
23.    Kaufer DI, Cummings JL, Ketchel P, Smith V, MacMillan A, Shelley T, et al. Validation of the NPI-Q, a brief clinical form of the Neuropsychiatric Inventory. J Neuropsychiatry Clin Neurosci 2000;12(2):233-9.
24.    Leng SX, Xue QL, Tian J, Walston JD, Fried LP. Inflammation and frailty in older women. J Am Geriatr Soc 2007;55(6):864-71.
25.    Puranen TM, Pietila SE, Pitkala KH, Kautiainen H, Raivio M, Eloniemi-Sulkava U, et al. Caregivers’ male gender is associated with poor nutrient intake in AD families (NuAD-trial). J Nutr Health Aging 2014;18(7):672-6.
26.    Fredman L, Lyons JG, Cauley JA, Hochberg M, Applebaum KM. The Relationship Between Caregiving and Mortality After Accounting for Time-Varying Caregiver Status and Addressing the Healthy Caregiver Hypothesis. J Gerontol A Biol Sci Med Sci 2015.
27.    Fonareva I, Amen AM, Zajdel DP, Ellingson RM, Oken BS. Assessing sleep architecture in dementia caregivers at home using an ambulatory polysomnographic system. Journal of Geriatric Psychiatry and Neurology 2011;24(1):50-59.
28.    Von Känel R, Ancoli-Israel S, Dimsdale JE, Mills PJ, Mausbach BT, Ziegler MG, et al. Sleep and biomarkers of atherosclerosis in elderly Alzheimer caregivers and controls. Gerontology 2010;56(1):41-50.
29.    Buigues C, Padilla-Sanchez C, Garrido JF, Navarro-Martinez R, Ruiz-Ros V, Cauli O. The relationship between depression and frailty syndrome: a systematic review. Aging Ment Health 2015;19(9):762-72.
30.    Mezuk B, Edwards L, Lohman M, Choi M, Lapane K. Depression and frailty in later life: a synthetic review. Int J Geriatr Psychiatry 2012;27(9):879-92.
31.    McEwen BS. Protective and damaging effects of stress mediators. N Engl J Med 1998;338(3):171-9.
32.    Burton L. C.  JTN, R Schulz, C. H. Hirsch, P.S. German. Preventive Health Behaviors among Spousal Caregivers. 1997.
33.    Buss DM. Toward a psychology of person–environment (PE) correlation: The role of spouse selection. Journal of Personality and Social Psychology 1984;47(2):361-377.
34.    Dassel KB, Carr DC. Does Dementia Caregiving Accelerate Frailty? Findings From the Health and Retirement Study. Gerontologist 2016;56(3):444-50.



Y. PARK1, N.-J. PAIK2, K.W. KIM3,4, H.-C. JANG5, J.-Y. LIM2


1. College of Medicine, Seoul National University, Jongno-gu, Seoul, Republic of Korea; 2. Department of Rehabilitation Medicine, Seoul National University College of Medicine, Seoul National University Bundang Hospital, Bundang-gu, Seongnam-si, Gyeonggi-do, Republic of Korea; 3. Department of Neuropsychiatry, Seoul National University College of Medicine, Seoul National University Bundang Hospital, Bundang-gu, Seongnam-si, Gyeonggi-do, Republic of Korea; 4. Department of Brain and Cognitive Science, College of Natural Sciences, Seoul National University, Gwanak-gu, Seoul, Republic of Korea; 5. Department of Internal Medicine, Seoul National University College of Medicine, Seoul National University Bundang Hospital, Bundang-gu, Seongnam-si, Gyeonggi-do, Republic of Korea
Corresponding author: Jae-Young Lim, MD, Department of Rehabilitation Medicine, Seoul National University College of Medicine, Seoul National University Bundang Hospital, 82, Gumi-ro 173 Beon-gil, Bundang-gu, Seongnam-si, Gyeonggi-do, Republic of Korea, Telephone: +82 31 787 7732, Fax: +82 31 787 4056, E-mail: drlim1@snu.ac.kr

J Frailty Aging 2017;6(3):144-147
Published online June 14, 2017, http://dx.doi.org/10.14283/jfa.2017.21



Abstract: Fall is a common cause of disability and death in old adults, and much research has been focused on identifying risk factors and developing preventive measures. Yet the majority of preceding research has been focused on physical performance. This study aims to evaluate the association between fall and depressive symptoms in community-dwelling elderly. Cross-sectional data of 431 men and 546 women was collected from old Korean adults living in Seongnam, Korea. Geriatric fall assessment was conducted by self-report questionnaires. Depressive symptoms were assessed by the Center for Epidemiologic Studies Depression Scale. Results indicated that depressive symptoms were associated with both fall and fear of falling in old adults. A clear gender difference was newly discovered, as depression played a stronger role in women. These results imply that clinicians should consider the negative affect of geriatric patients when assessing fall risk. Also, measures against depression might be effective in reducing falls.

Key words: Fall, fear of falling, depression.



Fall is one of the most common causes of injury-related death in the elderly (1). As the aging population increases, it has become of great interest to identify associated factors and subsequently develop prevention protocols (2). Yet most of previous studies have been focused on physical performance such as balance, gait speed, and muscle strength (3).
Geriatric depression is a major threat to health and independence in the older population (4). Physical performance such as timed walks, chair stands, and up and go tests are negatively affected by depressive mood (5). Furthermore, fall itself is influenced by depressive symptomatology. Therefore, it seems highly plausible that geriatric depression is closely associated with fall (6).
Another important aspect is fear of falling. Residents older than 65 years often report they are afraid of fall even if they have never experienced it before. This is not only a direct risk factor for fall but also an important cause of activity restriction and loss of autonomy (7). Depression and anxiety are common risk factors associated with fear of falling and might directly lead to activity restriction (8). Therefore, understanding the contribution of depressive symptoms to both fall and fear of falling is highly important.
This study attempts to analyze the association between depressive symptoms, fall, and fear of falling in community-dwelling Korean residents more than 65 years old. This study also aims to validate the gender difference in these associations.



Study Sample

This study is based on the data collected by the Korean Longitudinal Study on Health and Aging (KLoSHA). KLoSHA is a population-based cohort study conducted by Seoul National University Bundang Hospital from September 2005 to August 2006 (9). Residents more than 65 years old from Seongnam were recruited for research purpose. 439 men and 561 women agreed to enroll in the baseline cohort. All assessments of the participants were performed at the hospital under the surveillance of the institutional review board and with informed consent.



Geriatric fall assessment was conducted as part of a self-administered questionnaire on past medical history related to rehabilitation medicine. Participants had to answer whether they had ever fallen within a year or within 3 years, respectively. Detailed information regarding fall-related events was further collected. In addition to fall-related events, fear of falling was assessed. Patients reported how much they feared fall in daily life on a 0-to-5 scale.
Depression & anxiety
Participants were administered various neuropsychological surveys. Participants’ mood was evaluated by the Center for Epidemiologic Studies Depression Scale (CES-D). Anxiety was assessed by the State-Trait Anxiety Inventory (STAI), a self-administered survey with 40 items.


Age, gender, and educational level were assessed by interview. Activities of Daily Living (K-ADL) and Instrumental Activities of Daily Living (I-ADL) were assessed by standardized interviews. Body Mass Index (BMI, kg/m2) was calculated based on height and weight measured upon visit. Cognitive function was assessed by the Mini-Mental Status Examination (MMSE). The Cumulative Illness Rating Scale (CIRS) was used to assess comorbidities. The quality of life of the participants was evaluated by the Global Quality of Life Scale. The Performance-Oriented Mobility Assessment (POMA) was used to assess the participants’ physical performance.

Statistical Analyses

All statistical analyses were conducted by SPSS v. 23.0. Descriptive statistics including mean and standard deviation were calculated for all demographic, clinical, and functional variables.
The association between number of falls experienced and other factors were analyzed by the Pearson correlation coefficient. The same analysis was conducted for level of fear of falling. Multiple linear regression models for number of falls were constructed based on the results of correlation analysis.
Gender-specific analyses were performed in each gender subgroup, and the results were compared to the main analysis.



Descriptive Statistics

A total of 977 subjects were included as study participants for analysis (Table 1). Men were overall older than women. They had a higher educational level and better cognitive and physical function. However, they seemed to have more comorbidity. Men had also a lower BMI. Women possessed a lower ability to perform activities of daily living. Their level of depression, anxiety, and fear of falling were all higher compared to men. In addition, women seemed to fall more often than men: their responses on the number of falls experienced within both 1 year and 3 years were higher than the respective equivalents in men.


Table 1 Characteristics of study participants, n or mean ± SD

Table 1
Characteristics of study participants, n or mean ± SD

*. Gender difference is significant at the 0.05 level (2-tailed); †. Gender difference is significant at the 0.01 level (2-tailed).

Multiple Linear Regression Model

Correlation analysis of the studied variables proved that only STAI and CES-D were positively correlated with number of falls within both 1 year and 3 years. K-ADL, I-ADL, STAI, and CES-D showed a strong positive correlation with fear of falling. Based on the results of correlation analysis, a multiple linear regression model was developed for number of falls within 1 year (Table 2). The simple regression showed that number of falls within 1 year was associated with depressive symptoms. After controlling for anxiety level, the association was still significant. Anxiety was also independently associated with number of falls. After controlling for demographic factors, the association between depressive symptoms and number of falls within 1 year was still significant. The association between anxiety level and number of falls also remained significant. After physical function measures were added to the model, only depressive symptoms remained significantly associated with number of falls. Physical function showed no association. The association was attenuated but remained significant after cognitive function was added to the model. Lastly, the association remained significant after fear of falling was added. Fear of falling was independently associated with number of falls within 1 year.


Table 2 Multiple linear regression model of the association between number of falls within 1 year and depressive symptoms

Table 2
Multiple linear regression model of the association between number of falls within 1 year and depressive symptoms

*. Correlation is significant at the 0.05 level (2-tailed); †. Correlation is significant at the 0.01 level (2-tailed).

Gender-Based Comparison

Male and female participants showed significantly different profiles regarding depression and fall experience (Table 1). To further validate this gender difference, correlation analysis was performed for each gender subgroup. In female participants, anxiety and depression showed a positive correlation with number of falls within 1 year, and physical function showed a strong negative correlation. This pattern was consistent with the main analysis. Only anxiety level showed a significant correlation with number of falls within 3 years. In male participants, the correlation analysis revealed different results. Depressive symptoms did not show significant correlation with number of falls experienced within 1 year. Number of falls experienced within 3 years showed a statistically significant correlation only with age. This suggested that in men, the correlation between depression and fall might not be as strong as in women.



The aim of this study was to validate the association between depressive symptoms and fall. The results indicated that depressive symptoms showed a significant correlation to number of falls experienced in the past; depression remained strongly associated with fall after controlling for other variables including fear of falling. However, this result was only valid for female participants.
The association between aging and depression is clearly set (10). Geriatric depression is a rising problem, and it also leads to a chain reaction of functional deterioration in various domains (11). In addition, recent studies have observed that geriatric depression is a risk factor for falls, a major cause of disability and death in old adults (6).
This study found that old adults with fall experience obtained higher scores on depression scales. Through multiple linear regression analysis, it was shown that depressive symptoms remained significantly associated with experience of fall after covariates were controlled. In gender-specific correlation analyses, only the results of female participants matched the pattern of the original analysis. This gender difference was newly discovered.
Anxiety is a common comorbidity of depression. In geriatric patients, anxiety seems to be independently associated with fall-related psychological concerns (12). Fear of falling is associated with anxiety, depression, and physical activity. This study was consistent with previous observations in that anxiety level was significantly associated with both fall and fear of falling.
Fear of falling is a well-known fall-related psychological concern. It is also relatively common among healthy old adults who have never experienced fall before (13). Fear of falling is a common cause of physical restriction and often leads to decline in quality of life; known risk factors include age, female gender, depressive symptoms, previous falls, and subjective health status (14, 15). The results of this study were consistent with previous observations on the relationship between depression and fear of falling.
This study has certain limitations. First, the study was based on cross-sectional results; therefore, no causal relationship could be established. Longitudinal cohort studies are necessary for further analysis. Second, gender differences were identified in this study. The results of gender-specific correlation analyses suggested that depressive symptoms might not have a strong correlation to fall in males. However, as the subject number was limited, studies with a larger population would help to further validate this result.
This study suggests that depressive symptoms are related to experience of fall and fear of falling in old adults. Also, this study newly claims certain gender difference in this association. Assessment of depressive symptoms and negative affect in elderly might serve as an effective tool to screen for fall risk, and programs targeting geriatric depression might be useful in preventing fall.
Conflict of Interest: On behalf of all authors, the corresponding author states that there is no conflict of interest.
Ethical approval: All procedures performed in studies involving human participants were in accordance with the ethical standards of the institutional and/or national research committee and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards. This article does not contain any studies with animals performed by any of the authors.
Informed consent: Informed consent was obtained from all individual participants included in the study.
Funding: This study was supported by a research grant from the National Research Foundation (NRF) of the Korean Government (No. 2011-0030135) and the Korean Health Technology R&D Project, Ministry of Health & Welfare, Republic of Korea (HC15C1234). The sponsors had no role in the design and conduct of the study; in the collection, analysis, and interpretation of data; in the preparation of the manuscript; or in the review or approval of the manuscript.



1.     Majdan M, Mauritz W. Unintentional fall-related mortality in the elderly: comparing patterns in two countries with different demographic structure. BMJ Open 2015;5:e008672.
2.    Palumbo P, Palmerini L, Bandinelli S, Chiari L. Fall Risk Assessment Tools for Elderly Living in the Community: Can We Do Better? PLoS One 2015; 10(12):e0146247.
3.    Brach J, Berlin J, VanSwearingen J, Newman A, Studenski S. Too much or too little step width variability is associated with a fall history in older persons who walk at or near normal gait speed. J Neuroeng Rehabil 2005;2: 21.
4.    Steffens D. Research in Geriatric Depression: A Model for Mental Health Research in the 21st Century?. Am J Geriatr Psychiatry 2005;13:829-33.
5.    Lee Y. A study of the relationship between depression symptom and physical performance in elderly women. J Exerc Rehabil 2015;11:367-71.
6.    Kvelde T, Lord S, Close J et al. Depressive symptoms increase fall risk in older people, independent of antidepressant use, and reduced executive and physical functioning. Arch Gerontol Geriatr 2015;60:190-5.
7.    Denkinger M, Lukas A, Nikolaus T, Hauer K. Factors Associated with Fear of Falling and Associated Activity Restriction in Community-Dwelling Older Adults: A Systematic Review. Am J Geriatr Psychiatry 2015;23:72-86.
8.    Benzinger P, Zijlstra G, Lindemann U et al. Depressive symptoms and fear of falling in previously community-dwelling older persons recovering from proximal femoral fracture. Aging Clin Exp Res 2012;24:108.
9.    Huh Y, Yang E, Lee S et al. Association between executive function and physical performance in older Korean adults: Findings from the Korean Longitudinal Study on Health and Aging (KLoSHA). Arch Gerontol Geriatr 2011;52:e156-e161.
10.    Newmann J. Aging and depression. Psychol Aging 1989;4:150-65.
11.    Royall D, Schillerstrom J, Piper P, Chiodo L. Depression and Mortality in Elders Referred for Geriatric Psychiatry Consultation. J Am Med Dir Assoc 2007;8:318-21.
12.    Hull S, Kneebone I, Farquharson L. Anxiety, Depression, and Fall-Related Psychological Concerns in Community-Dwelling Older People. Am J Geriatr Psychiatry 2013;21:1287-91.
13.    Liu J. Fear of falling in robust community-dwelling older people: results of a cross-sectional study. J Clin Nurs 2014;24:393-405.
14.    Guillemin F, Martinez L, Calvert M et al. Fear of falling, fracture history, and comorbidities are associated with health-related quality of life among European and US women with osteoporosis in a large international study. Osteoporos Int 2013;24:3001-10.
15.    Oh-Park M, Xue X, Holtzer R, Verghese J. Transient Versus Persistent Fear of Falling in Community-Dwelling Older Adults: Incidence and Risk Factors. J Am Geriatr Soc 2011;59:1225-31.





1. Department of Psychology, Drexel University, USA; 2. Department of Healthcare Policy and Research, Weill Cornell Medical College, USA; 3. Department of Psychiatry, Weill Cornell Medical College, USA

Corresponding author: Alexandra Greenfield, M.S., Drexel University, 3141 Chestnut Street, Stratton Suite 119, Philadelphia, PA 19104, USA, Phone: 914-980-5720, apg54@drexel.edu

Alternate Corresponding Author: Jo Anne Sirey, Ph.D., Weill Cornell Medical College, 21 Bloomingdale Rd, White Plains, NY 10605, USA, Phone: 914-997-4333, Fax: 914-682-6979, jsirey@med.cornell.edu



Objectives: This study used the Evans model of public health determinants to identify factors associated with nutritional risk in older adults. Design: The Evans model domains (physical and mental well-being, social/environmental statuses, individual choice, and economic security) were measured in a sample of homebound older adults. Regularized logistic regression analysis with LASSO penalty function was used to determine the strongest domain of the Evans model. Using traditional logistic regression, individual variables across all domains were compared to identify the significant predictors. Setting: Older adults receiving home meal services were referred to the study by community program staff. Participants: Participants included 164 homebound older adults (age > 60) who endorsed at least one gateway symptom of depression. Measurements: Nutritional risk was determined using the Mini Nutritional Assessment. Domains of the Evans model were measured using the MAI Medical Condition Checklist, items from the IADL scale, the Structured Clinical Interview for DSM-IV Axis I Disorders, the Duke Social Support Index, living arrangements, marital status, the Alcohol Use Disorders Identification Test, items from the SCID Screening Module, and a self-report of perceived financial security. Results: Poor mental well-being, defined by a diagnosis of major depressive disorder, was identified as the strongest Evans model domain in the prediction of nutritional risk. When each variable was independently evaluated across domains, instrumental support (Wald’s Z=-2.24, p=0.03) and a history of drug use (Wald’s Z=-2.40, p=0.02) were significant predictors. Conclusions: The Evans model is a useful conceptual framework for understanding nutritional health, with the mental domain found to be the strongest domain predictor of nutritional risk. Among individual variables across domains, having someone to help with shopping and food preparation and a history of drug use were associated with lower nutritional risk. These analyses highlight potential targets of intervention for nutritional risk among older adults.  


Key words: Depression, nutrition, model selection, Grouped LASSO, redundancy analysis. 



Malnutrition is a substantial problem among older adults. More than one out of five older adults suffer from malnutrition and more than two out of five are at risk for malnutrition, which is commonly referred to as “nutritional risk” (1). Poor nutrition may lead to a range of negative health outcomes by causing a decline in functional status, worsening of existing medical problems, and significantly shorter survival rates (2). It is important to identify individuals at nutritional risk because interventions that target factors associated with nutritional risk may reduce the adverse effects of malnutrition and reverse its course. Nutritional interventions have also been linked to shorter length of hospitalizations (3) and lower overall costs of medical services (4).

A number of factors have been shown to contribute to the development of malnutrition in older adults. While malnutrition is not an inevitable consequence of aging, many of the physical changes associated with aging are also related to decreased food intake and appetite, declining energy, loss of taste and smell, poor oral health, and dysphagia (5). Chronic illnesses such as diabetes, arthritis, oral manifestations, and gastrointestinal disorders are strongly associated with poor nutrition due to decreases in functional ability, changes in dietary intake, and weight loss (5). In addition to these physical risk factors, being female or a racial/ethnic minority (6), having greater levels of depressive symptoms (7), living alone (6), and loneliness (8) have all been found to be independently related to poor nutritional status in older adults.

Given the complexity of nutrition-related challenges facing the older adult, it is important to apply a theoretical model that takes into account the physical, mental, social, and environmental domains that affect nutritional risk. Few studies have examined nutritional health using a multidimensional model that can explore the relative importance of the predictors. 

One study (9) used hierarchical regression analysis to test the dynamics of several factors shown to be individually important to nutritional health in a sample of community-dwelling older adults. The theoretical framework for this study was the Roy Adaptation Model (RAM), a biopsychosocial approach which posits that health outcomes are a function of an individual’s ability to adapt to changing environmental demands. In the context of nutritional health, this may refer to one’s functional status in response to multiple medical comorbidities. After adjusting for the confounding effects of the various factors, depressive symptoms, functional status, oral health, and income were identified as significant predictors of nutritional health.

The current study seeks to determine factors associated with nutritional risk using the Evans model of the determinants of population health (10) as a conceptual framework. This model’s approach is based on the World Health Organization’s definition of health as not merely the absence of disease or injury, but a state of complete physical, mental, and social well-being. The Evans model considers the major domains that contribute to illness, including (1) physical well-being, (2) social/environmental status, (3) individual choice, and (4) economic security [10]. To capture the potential impact of mental health, we expanded the “physical well-being” domain into two distinct domains: physical well-being and mental well-being. 

The Evans model has been used to guide the analyses of factors that predict hypertension (11), high cholesterol (11), and coronary heart disease (10). In a previous study that used the Evans model to assess nutritional risk in community-dwelling older adults, the overall model was not found to be a statistically significant predictor of nutritional status (12). However, nutritional status was operationalized as body mass index and weight change, which may not be valid markers of malnutrition in the older adult population due to expected changes in body composition due to aging. 

The current study measures the Evans model domains (physical well-being, mental well-being, social/environmental status, individual choice, and economic security) in a sample of homebound older adults receiving home-delivered meals. Specifically, we investigate two questions: 1) which domain of the Evan model has the strongest association with nutritional risk, and 2) which variables among all the domains of the Evans model are the strongest predictors of nutritional risk. Regularized logistic regression analysis with grouped LASSO penalty function was used to determine the strongest predictive domain (group of variables) affecting nutritional risk. Using traditional logistic regression, individual variables across all domains were compared to identify the best predictor(s). By applying these two statistical techniques to a theoretical framework for the prediction of nutritional risk, we hope to improve the ability to identify older adults at risk of developing malnutrition and its associated negative outcomes. 



Sample: This study uses cross-sectional data collected from a sample of homebound older adults recruited for Open Door, a community-based mental health intervention study (NIMH R01 MH087557) (13). Study participants were adults age 60 or older who are homebound, eligible for home delivered meals services, and endorse either depressed mood or lack of interest or pleasure, the gateway symptoms necessary for a diagnosis of depression. This sample with depressive symptoms, physical limitations and medical burden was chosen for the analysis because of their complex clinical presentation and potential vulnerability to high nutritional risk and poor outcomes. The sample also provided sufficient variability across all of the Evans model domains to allow for robust analyses. The home delivered meals program has served approximately 2.6 million frail older adults (14) who are ‘confined’ due to a condition, illness or injury that restricts the ability of the individual to leave their home without assistance. Compared to the overall U.S. population above age 60, recipients of home delivered meals are more likely to be older, Black, living alone in poor health, have greater difficulty performing everyday tasks, and be at a high nutritional risk (15, 16).  

When an eligible study participant was identified by home delivered meals staff, verbal consent was obtained for the study personnel to contact the older adult, describe the study, and assess interest. If the subject chose to participate, a counselor made an in-home visit to obtain informed consent and conduct a baseline assessment. Exclusion criteria included the presence of significant substance abuse history or psychotic disorder, active suicidal ideation requiring immediate attention, cognitive impairment (Mini-Mental State Examination (17); MMSE < 24), aphasia, inability to speak English or currently in mental health treatment (either antidepressant medication or psychotherapy). This study was approved by the Weill Cornell Medical College Institutional Review Board (Protocol 0808009247).

Of 362 older adults referred for evaluation by community staff, 106 were ineligible after a phone screening as they were already in mental health treatment or too cognitively impaired to understand study recruitment procedures. Of those who were contacted for an in-person screen, 77 were excluded from the sample. Of the 77 excluded, 44 were too cognitively impaired or medically ill, 18 were already receiving treatment for depression, 6 had an alternative psychiatric diagnosis or substance abuse history, 3 could not speak English, 1 changed his/her mind about participation in research, 1 was at high risk for suicide, 1 was not homebound, and 3 were identified in the “other” category. 164 subjects were included in this analysis. 

Outcome Variable

Nutritional risk status was assessed using the Mini Nutritional Assessment (18). This assessment is a validated tool for nutritional screening and assessment in older adults (19), and it is routinely administered throughout the state by providers receiving funding for home meal programs. The measure assesses nutritional intake, food preparation and yields a continuous score from 0 to 21 with high nutritional risk defined as a score of 6 or greater.   

Factors Associated with Nutritional Risk 

To apply the Evans model to determine the factors associated with nutritional risk, measures collected by the Open Door study were matched with the associated Evans model domains (mental well-being, physical well-being, social/environmental status, individual choice, and economic security) as described in Table 1. We consider each domain of the Evan’s model as potential predictors of nutritional risk after controlling for demographic factors and cognitive status. In the mental domain of the Evans model, poor mental well-being was defined as suffering from major depression. A research diagnosis of Major Depressive Disorder (MDD) was determined using the Structured Clinical Interview for DSM-IV Axis I Disorders (SCID-1) (20) administered by trained social workers under the supervision of a clinical psychologist. Depression status was dichotomized into those individuals who met DSM-IV (SCID) criteria for MDD and those who did not. The physical domain of the model included measures of medical burden and physical capacity for food preparation. Medical burden was defined as the number of current medical conditions endorsed in the Multi-level Assessment Inventory Medical Condition Checklist (MAI) (21). Capacity for food preparation was defined as the ability to shop and prepare foods. It includes the degree to which an individual is reliant on others for food shopping and was measured using self-report questions from Instrumental Activities of Daily Living (IADL) (22). For this older adult sample, the social/environmental domain of the model combined measures of social support and living arrangements. Social support was defined using the three subscales of the Duke Social Support Index (DSSI) (23), measuring subjective support, social integration and instrumental support. The individual’s living arrangements indicated whether there was someone in the home who could support the individual’s nutritional health. Those individuals who lived alone and were unmarried or unpartnered were defined as having an environment that is less supportive of nutritional intake. The economic domain of the model was measured using the individual’s perceived current financial security. This measure captured the degree to which an individual felt with or without resources that could affect nutritional status. Subjects described their financial security as: 1) “can’t make ends meet”, 2) “have just enough to get along”, or 3) “are currently comfortable.”  The negative individual choice domain in the Evans model was defined by alcohol and drug use based on the growing number of older adults using and abusing alcohol and drugs (24). The SCID Screening Module was used to evaluate current and past street drug or prescription drug abuse. Current alcohol consumption was measured by the Alcohol Use Disorders Identification Test (AUDIT-C) (25). The AUDIT-C has been shown to be useful in the detection of hazardous drinking among older adults, who may be particularly sensitive to the toxic effects of alcohol due to impaired metabolism and potential interactions with medications (26).


Table 1 Evans Factors, Domains Assessed, and Associated Measures


Statistical Covariates

Demographic variables (summarized in Table 2) were included as covariates in the analyses to control for the effects of characteristics such as gender, age, and race. Although participants did not have cognitive impairment, cognitive status as measured by the MMSE was included as a covariate to account for the potential association between cognitive functioning and nutritional health in older adults (27).


Table 2 Demographic Characteristics and Evans Factors by Nutritional Risk Status

Notes. MMSE: Mini-Mental Status Examination; MAI Medical Condition Checklist: Multi-level Assessment Medical Condition Checklist; IADL: Instrumental Activities of Daily Living; AUDIT-C: Alcohol Use Disorder Identification Test; SCID: Structured Clinical Interview for the DSM-IV


Data Analysis Plan

Prior to testing hypotheses, analyses were conducted to determine interdependence between variables to ensure that each variable is independent and represents unique aspects of the sample. Hierarchical clustering and redundancy analysis, as described below, were performed to evaluate redundancy. Nutritional risk was defined as a binary outcome, using the cut-off score of 6 or greater on the Mini Nutritional Assessment. The goals of the statistical analyses were: 1) to identify the strongest domain among the five Evans model domain associated with nutritional risk; and 2) to investigate association of the individual variables (making up the domains) with nutritional health. 

Redundancy Analysis to Evaluate Interdependence

The goal of this analysis was to evaluate the correlations and interrelationships among the predictor variables (both demographic and the variables within the Evans model domains) and determine collinearity between these variables. To this end, we performed hierarchical clustering and redundancy analysis (28, 29). Hierarchical clustering is an algorithm that assesses similarity between independent variables. It starts out by assigning each variable to its own cluster, finds the closest pair of clusters using a linkage metric and merges them into a single cluster. The algorithm then computes the similarity between the new cluster and each of the old clusters. This process is repeated until one single cluster is generated. Spearman’s rank correlation is used as the similarity metric and complete linkage function is used as the linkage metric between clusters. A dendrogram depicting the relationships between the variables is presented where the arrangement of ‘branches’ depicts which variables or clusters of variables are similar to each other and the height of ‘branches’ indicates strength of the similarity (See Figure 1). 


Figure 1 Dendogram representing hierarchical clustering of 17 predictors with Spearman’s rank correlation as the distance metric


To determine if a variable is redundant, we perform a redundancy analysis to reveal how well each independent variable can be predicted from all other variables. Redundancy analysis considers the regression of each variable on the other variables, dropping the most predictive variables in a stepwise manner until no variable can be predicted with an R2 of at least 0.8. As a result of this analysis, a variable is considered redundant if it can be predicted from a linear combination of all other variables. Non-redundant variables are retained for the statistical model.

Domains Analysis of Nutritional Risk with Grouped LASSO

To identify the strongest domain associated with nutritional risk among the five domains of the Evans model, a regularized logistic regression was performed with a Grouped LASSO (Least Absolute Shrinkage and Selection Operator) penalty function (Yuan, 2006). A traditional logistic regression analysis allows us to find associations of individual variables with nutritional risk, but grouped LASSO analysis provides the relative importance of groups of variables combined into a single domains. For example, a domain has five variables A, B, C, D and E, one variable (say C) may be highly associated with nutritional risk, but the domain as a whole may still be weakly associated with nutritional risk. The grouped LASSO analysis is a regularized regression technique, which operates by optimizing estimates of regression coefficients with a LASSO penalty function. LASSO penalty, when used in conjunction with a regression model, is a method of variable selection to select a subset of predictors which fit the model as well as the entire set of predictors. Grouped LASSO performs selection on groups of variables instead of individual variables. The selection is performed by penalizing groups of regression coefficients (corresponding to the variables in a particular group) with the LASSO penalty function. 

The LASSO penalty function shrinks the coefficients of unimportant groups of variables to zero (effectively removing them from the regression model); all regression coefficients below a certain threshold are treated as zero and removed from the model. The degree of penalty or shrinkage is controlled by a tuning parameter (λ), which determines the threshold below which regression coefficients will be turned into zero.  We chose a series of values for λ such that the regression model corresponding to the largest value of λ only includes the intercept, demographic variables and cognitive status. The five domains of the Evans model represent five groups of variables upon which Grouped LASSO selection was performed. The demographic variables and cognitive status represented an additional group which was always retained in the model. The first group to be included in the model along this decreasing series of λ is deemed as the strongest domain among the Evans model domains. 

Variable Analysis of Nutritional Risk with Logistic Regression

To evaluate the association of the twelve specific variables (that make up the Evans model domains) with nutritional risk, a logistic regression model was performed controlling for demographic factors and cognitive status. The strongest predictor of risk was determined by the ranks of the standardized coefficients or equivalently z-statistic of logistic regression (30). Strength of prediction of a future observation is represented by the c-statistic which measures the area under the Receiver Operating Characteristics or ROC curve (plot of sensitivity vs. 1-specificity).  



The sample consisted of 164 participants who screened positive for depression as a part of routine assessment for home delivered meals eligibility. Table 2 presents the demographic and clinical characteristics of the sample. 

The redundancy analysis performed on the set of seventeen variables (Evans model variables combined with the demographic variables) shows that at R2=0.8 level there are no redundant predictors, i.e. no single variable can be predicted from the other variables with an R2 of 0.8. Instrumental social support had the highest R2=0.47 (in a model with instrumental support as a dependent variable and all variables except instrumental support as independent variables) among all variables. The dendrogram (Fig 1) shows the hierarchical clustering of independent variables using Spearman’s rank correlation as the distance metric. This shows that current living situation and marital status have highest similarity (ρ2=0.30) followed by subjective social support (DSSI) and instrumental social support (DSSI) (ρ2= 0.23). As a result of the redundancy and clustering analysis we can conclude that the interdependency and correlations among the seventeen variables of nutritional risk is negligible and the proposed association analysis does not need to take these into account.


Table 3 Logistic Regression Analysis of Nutritional Risk

Note. IADLs: Instrumental Activities of Daily Living


In the Grouped LASSO analysis, to identify the domain most associated with nutritional risk, the intercept, demographic and cognitive variables were fixed and not subjected to selection. Among the Evans model domains most associated with nutritional risk was the mental domain defined as a diagnosis of MDD after controlling for demographic variables such as race, ethnicity, age, gender and education and cognitive status measured by MMSE. The second domains selected were the physical domain (defined as medical burden and impairments in instrumental activities of daily living), individual choice (alcohol use and report of drug use) and economic (self-reported financial status) domains. The social domain was the final domain to enter the model. The relative importance of each domain is determined by when it is included in the model along the path of the shrinkage parameter (see Methods).

Using a traditional logistic regression model, all twelve variables from the Evans model and all demographic factors were included in the model with nutritional risk as the dependent variable. The strongest associated variables were a history of drug use (Wald’s Z=-2.40, p=0.02) and instrumental support (Wald’s Z=-2.24, p=0.03). They were followed by major depression (Wald’s Z=1.95, p=0.05) and medical burden (Wald’s Z=1.86, p=0.06), although not statistically significant. The c-statistic for this regression was c=0.79 indicating that the model predicted nutritional risk significantly better than chance (c=0.5). The model with all variables (Evans model variables and demographic and cognitive status variables) and a model with only demographic and cognitive status variables were compared using a likelihood ratio test and showed significant improvement (χ2(11)=1.93, p = 0.0013). This indicates the usefulness of the Evans model variables as predictors of nutritional risk. History of drug use (OR=0.17, 95% CI 0.02-0.6) and having higher instrumental social support (OR = 0.84, 95% CI 0.71- 0.97) were significantly associated with lower odds of nutritional risk.  



Using the Evans population health model as a conceptual framework, we evaluated factors associated with nutritional risk in a sample of homebound older adults with depressive symptoms, medical burden, limited mobility and isolation. We employed innovative statistical strategies to investigate which domain of the Evans model had the strongest association with nutritional risk and traditional statistical analyses to determine which variables among all the domains of the Evans model were associated with nutritional risk. Our findings indicated that the domain of mental well-being, which we defined as a diagnosis of major depressive disorder, had the strongest association with nutritional risk. Across all variables, poor instrumental support with respect to food shopping and preparation and a history of drug use were significantly associated with lower nutritional risk. These two findings capture the outcomes evaluated by two different statistical approaches. Among the Evans model domains, grouped LASSO analysis identified the mental domain as having the strongest association with nutritional risk, whereas traditional logistic regression analysis picked instrumental support (belonging to the social/environmental domain) and history of drug use (individual choice domain) as the strongest associated variables. As domains, social and individual choice domains are less important relative to the mental domain, but individual variables within these domains are highly associated with nutritional risk.

The results of this study identify mental well-being, or the presence of major depressive disorder, as the strongest among all Evans model domains in predicting nutritional risk. This supports the data from community samples and extends the finding that depression may increase homebound older adults’ vulnerability to malnutrition (7, 31). Major depression can have a pervasive impact on an individual’s life, influencing many areas of functioning. Depression has been associated with involuntary weight loss in older adults (32), as well as increased risk of disability (33), disease (34) health risk behaviors (35), and social isolation (36), all of which have been shown to be independently related to nutritional status (8, 37-39). In this sample where depression is untreated, the interrelation of nutrition and depression may be particularly evident. 

Our finding that greater instrumental support is related to lower nutritional risk is consistent with other literature suggesting the importance of social resources in supporting the nutrition of older adults (40). These older adults have others who can provide assistance for food shopping and meal preparation. Access to this specific type of social support seems to play a critical role in nutritional health outcomes for homebound older adults.

The significant relationship between prior drug use and low nutritional risk was unexpected. However, the measure used in this study captured history of drug use rather than contemporary use. It is possible that older adults who have engaged in previous risky behaviors when younger have become more attentive to self-care as they age. It is also possible that the limited variability in our sample yielded spurious findings, as only 6 study participants reported a history of drug use. Future work should assess the association between current and past substance use and nutritional behaviors to further explore this relationship.

A limitation of this study is the selection of a homebound sample with depressive symptoms who have already been identified as potentially at risk for poor nutritional health due to restricted mobility. While there was a wide range of depressive symptoms within this sample, we recognize that we have increased the possibility of finding a relationship between major depression and nutritional risk due to increased prevalence of depressive symptoms. However, we believe that this is important to explore given the findings that more than 1 out of 10 home meal recipients has clinically significant depression and an additional 1 out of 8 have mild depression (Sirey et al., 2008). We also recognize that the cross-sectional nature of the study limits our ability to determine the impact of the identified factors on the course of nutritional risk. Our exploration of the Evans model was limited to the available scales administered by the Open Door study. Some of the variables included in our analyses may have had insufficient variability to demonstrate significant relationships with nutritional risk. To improve the integrity of the analyses, we included demographic variables and cognitive status as covariates, and we revised the original Evans model by separating physical and mental well-being into distinct domains. However, the model lacked the inclusion of additional factors that may contribute to nutritional risk in older adults. For instance, a higher frequency of service use has been shown to be associated with lower nutritional risk (41). 

Despite its shortcomings, this study helps to illuminate potential future targets of intervention for nutritional risk. In particular, depression, instrumental support, drug use, and nutrition may be interdependent health concerns for older adults. The use of innovative statistical procedures allowed us to evaluate the associations between nutritional risk and both the individual variables and conceptual domains of the Evans model. It is important to better understand the relative importance of individual predictors and larger conceptual domains in the context of nutritional risk, as this will help health care providers target the most vulnerable patients for preventative measures and the promotion of positive health outcomes.  


Conflict of Interest: All authors have disclosed that there are no financial, personal or potential conflicts. 

Author Contributions: Alexandra P. Greenfield, M.S., Samprit Banerjee, Ph.D., Alyssa DePasquale, B.A., and Nathalie Weiss, B.A., Jo Anne Sirey, Ph.D.: preparation of the paper. Greenfield, Sirey. Banerjee, and DePasquale: study concept, design, and methods. Sirey and Banerjee: analysis and interpretation of data. 

Sponsor’s Role: This research is supported by a grant from the National Institute of Mental Health (R01 MH079265, PI: J Sirey; P30 MH085943, PI GS Alexopoulos). The study design and conclusions are the sole responsibility of the authors and not the sponsor. 



1. Kaiser MJ, Bauer JM, Rämsch C, et al. Frequency of malnutrition in older adults: a multinational perspective using the mini nutritional assessment. J Am Geriatr Soc 2010;58(9):1734-1738.

2. Shakersain B, Santoni G, Faxén-Irving G, Rizzuto D, Fratiglioni L, Xu W. Nutritional status and survival among old adults: an 11-year population-based longitudinal study. Eur J Clin Nutr 2015;1-6.

3. Smith PE, Smith A.E. High-quality nutritional interventions reduce costs. Healthc Financ Manage 1997;51(8):66-69.

4. Arnaud-Battandier F, Malvy D, Jeandel C, et al. Use of oral supplements in malnourished elderly patients living in the community: a pharmaco-economic study. Clin Nutr 2004;23(5):1096-1103.

5. Nickols-Richardson SM, Johnson MA, Poon LW, Martin P. Mental health and number of illnesses are predictors of nutritional risk in elderly persons. Exp Aging Res1996;22(2):141-154.

6. Wham CA, McClean C, Teh R, Moyes S, Peri K, Kerse N. The Bright Trial: What are the factors associated with nutrition risk? J Nutr Health Aging 2013;18(7):692-697.

7. Yoshimura K, Yamada M, Kajiwara Y, Nishiguchi S, Aoyama T. Relationship between depression and risk of malnutrition among community-dwelling young-old and old-old elderly people. Aging Ment Health 2013;17(4):456-460.

8. Ramic E, Pranjic N, Batic-Mujanovic O, Karic E, Alibasic E, Alic A. The effect of loneliness on malnutrition in elderly population. Med Ar 2011;65(2):92-95.

9. Chen CCH, Chang CK, Chyun, DA, McCorkle R. Dynamics of nutritional health in a community sample of american elders: A multidimensional approach using Roy Adaptation Model. Adv Nurs Sc 2005;28(4):376-389.

10. Evans RG, Barer ML, Marmor TR. Why are some People Healthy and others not? The Determinants of Health of Populations. 1994. Aldine de Gruyter, New York.

11. Evans RG, Stoddart GL. Producing health, consuming health care. Soc Sci Med 1990;31(12):1347-63.

12. Callen BL, Wells TJ. Screening for Nutritional Risk in Community-Dwelling Old-Old. Public Health Nurs 2005;22(2):138-146.

13. Sirey JA, Greenfield A, DePasqaule A, et al. Improving engagement in mental health treatment for home meal recipients with depression. Clin Interv Aging 2013;8:1305-1312.

14. Colello KJ. Older Americans Act: Title III Nutrition Services Program. 2011. Congressional Research Service, Washington DC.

15. Greenlee K. Senior Hunger and the Older Americans Act (testimony of Kathy Greenlee), in U.S. Senate Subcommittee on Primary Health and Aging of the Committee on Health, Education, Labor, and Pensions. 2011. National Academies Press, Washington, DC.

16. Ponza M, Ohls JC, Millen BE, et al. Serving Elders At Risk. The Older Americans Act Nutrition Programs – National Evaluation of the Elderly Nutrition Program, 1993–1995, 1996. Mathematica Policy Research, Princeton, NJ.

17. Folstein MF, Folstein SE, McHugh PR. “Mini-mental state”: a practical method for grading the cognitive state of patients for the clinician. J Psychiat Res 1975;12(3):189-198.

18. Guigoz Y, Vellas B, Garry PJ, Vellas BJ, Abarede JL. Mini Nutritional Assessment: a practical assessment tool for grading the nutritional state of elderly patients. J Nutr Elder, 1997;15-60.

19. Kaiser MJ, Bauer JM, Rämsch C, et al. Validation of the Mini Nutritional Assessment Short-Form (MNA®-SF): A practical tool for identification of nutritional status. J Nutr Health Aging 2009;13(9):782-788.

20. First MB, Spitzer RL, Gibbon M, Williams JB. Structured Clinical Interview for DSM-IV Axis I Disorders: Patient Edition (February 1996 Final), SCID-I/P. 1996. New York State Psychiatric Institute, New York, NY.

21. Lawton MP, Moss M, Fulcomere MC, Kleban, MH. A Research and Science Oriented Multi-Level Assessment Instrument. J Gerontol 1982;37:91-99.

22. Katz S. Assessing self-maintenance: activities of daily living, mobility, and instrumental activities of daily living. J Am Geriatr Soc 1983;31(12):721-727.

23. Landerman R, George LK, Campbell RT, Blazer DG. Social support, stress, and depression: alternative models of the stress buffering hypothesis. Am J Community Psychol 1989;17(5):625-642.

24. Blow FC, Barry KL. Alcohol and substance misuse in older adults. Curr Psychiatry Rep 2012;14(4):310-319.

25. Bush K, Kivlahan DR, McDonell MB, Fihn SD, Bradley KA, The AUDIT alcohol consumption questions (AUDIT-C): an effective brief screening test for problem drinking. Arch Intern Med 1990;158(16):1789-1795.

26. Gomez A, Conde A, Santana JM, Jorrin A, Serrano IM, Medina R. The diagnostic usefulness of AUDIT and AUDIT-C for detecting hazardous drinkers in the elderly. Aging Ment Health 2006;10(5):558-561.

27. La Rue A, Koehler KM, Wayne SJ, Chiulli SJ, Haaland KY, Garry PJ. Nutritional status and cognitive functioning in a normally aging sample: a 6-y reassessment. Am J Clin Nutr 1997;65(1):20-29.

28. Harrell FE. Regression modeling strategies: with applications to linear models, logistic regression, and survival analysis. 2nd ed. 2015. Springer, New York.

29. Harrell FE. Harrell Miscellaneous (Version R package, Version 3.12-2), in http://CRAN.R-project.org/package=Hmisc2013.

30. Bring J. How to standardize regression coefficients. Am Stat 1994;48(3):209-213.

31. Cabrera MAS, Mesas AE, Garcia ARL, de Andrade SM. Malnutrition and depression among community-dwelling elderly people. J Am Med Dir Assoc 2007;8(9):582-584.

32. Huffman GB. Evaluating and treating unintentional weight loss in the elderly. Am Fam Physician 2002;65(4):640-650.

33. Bruce ML. Depression and disability in late life: directions for future research. Am J Geriat Psychiatr 2001;9(2):102-112.

34. Penninx BW, Guralnik JM, Ferrucci L, Simonsick EM, Deeg DJ, Wallace RB. Depressive symptoms and physical decline in community-dwelling older persons. JAMA 1998;279(21):1720-1726.

35. Cranford JA, Nolen-Hoeksema S, Zucker RA. Alcohol involvement as a function of co-occurring alcohol use disorders and major depressive episode: Evidence from the National Epidemiologic Survey on Alcohol and Related Conditions. Drug Alcohol Depen 2011;117(2):145-151.

36. Vanderhorst RK, McLaren S. Social relationships as predictors of depression and suicidal ideation in older adults. Aging Ment Health 2005;9(6):517-525.

37. Kiesswetter E, Pohlhausen S, Uhlig K, et al. Malnutrition is related to functional impairment in older adults receiving home care. J Nutr Health Aging 2013;17(4):345-350.

38. Charlton K, Nichols C, Bowden S, et al. Poor nutritional status of older subacute patients predicts clinical outcomes and mortality at 18 months of follow-up. Eur J  Clin Nutr 2012;66(11):1224-1228.

39. Ory MG, Jordan PJ, Bazzarre T. The Behavior Change Consortium: setting the stage for a new century of health behavior-change research. Health Educ Res 2002;17(5):500-511.

40. Locher JL, Ritchie CS, Roth DL, Baker PS, Bodner EV, Allman RM. Social isolation, support, and capital and nutritional risk in an older sample: ethnic and gender differences. Soc Sci Med 2005;60(4):747-61.

41. Weddle D, Wilson FL, Berkshire SD, Heuberger R. Evaluating nutrition risk factors and other determinants of use of an urban congregate meal program by older African Americans. J Nutr Gerontol Geriatr 2012;31(1):38-58.





Sociomedical Division of the Preventive Medicine and Community Health department at the University of Texas Medical Branch, Galveston, TX 77555

Corresponding author: C. Siordia, phone number: 1-409-772-5899; e-mail: csiordia@utmb.edu

J Frailty Aging 2013;2(3):153-164
Published online February 12, 2016, http://dx.doi.org/10.14283/jfa.2013.23


The almost irrefutable hypothesis that place matters continues to grow in popularity.  Epidemiological and public health researchers are studying social and physical environment’s effect on individual health outcomes. Advances in the field are hindered by the lack of consistency in measuring and labeling social contexts. Greater definitional precision is required. In order to give an example of this, “neighborhood” studies between 2000 and 2012 dealing with depression symptomatology in older adults were identified with an exhaustive search. Only those where the terms neighborhood, and mental health, or mental well-being, or CES-D appear were included for the initial review. After additionally selecting for age and the presence of the Center for Epidemiologic Studies Depression (CES-D) measure, from an initial 98 articles, we end up with 11 articles. We focus on how neighborhoods are defined and briefly highlight findings on CES-D. For the most part, the definition of neighborhood is limited, frequently justified, and typically hidden in the methodological details and closing arguments of an article. In general, articles evade the discussion of polygon appropriateness with relation to the term neighborhood. Our review suggests that a good starting point for advancing this field of inquiry would be to increase the definitional precision of the term neighborhood and to offer an upfront disclosure with more appropriate terminology. Doing so may lead place-effect investigations in population aging and frailty down a more productive road.


Key words: Neighborhood, health, aging, CES-D, spatial, depression.



Depressive mood may impair an aged adult’s functional ability and thus their frailty—their multisystem reduction in physiological capacity. Place effect studies using multilevel and/or spatial modeling aim at informing researchers how an aged adult’s area of residence influences their health outcomes. Given the complex nature of trying to determine how the individual and his/her environment interact, attempts to inform research on how context matters use many statistical and theoretical assumptions. When attempting to investigate place effects, researchers are usually bounded by the availability of data—measures of health outcomes, sample sizes, and predefined nesting units. Some investigators make use of predefined nesting units (e.g., counties, tracts) to measure individuals’ social and/or physical environment. Some of them label their non-theory driven geographical polygons as “neighborhoods.” Using the term neighborhood during the interpretations of results in quantitative research may affect the level of meaning assigned to the study’s findings—where using the term lends more face validity to the research. Our article substantively contributes to place effect literature that focuses on the ageing population by offering a detailed discussion on the treatment of the word “neighborhood” in studies who investigate depression symptomatology using the complete or abbreviated version of the Center for Epidemiologic Studies Depression (CES-D) instrument with samples from both urban and non-urban areas.

Defining “Neighborhood”

Understanding frailty requires that an aged adult’s psycho-social interactions with his/her environment be understood. Investigating context’s effects on mental health requires the production of place measures. The creation of place measures demands that researchers identify a set of geographical boundaries to identify a place. For example, a hypothetical study on how an area’s criminal activity affects an individual’s depression symptomatology would require that places be given a score on some “criminal activity” scale. In order to give a place a criminal activity score, the researcher must determine what geographical boundaries constitute a place. That is, criteria for determining a place’s geographical boundaries would have to be (either explicitly or implicitly) chosen. Consequently, the need for geoboundarization (i.e., the act of determining geographic boundaries) mandates that the investigator make implicit or explicit theoretical decisions of what a person perceives as their area of residence. Investigations on mental health with United States (US) samples typically make use of US Census Bureau geographical units (e.g., counties, tracts). These methods have existed for many decades (1) and remain utilized as we move towards developing a clearer understanding of the role that geographical dependence plays in mental health outcomes (2).

The spatial units created by the geoboundarization process necessary for the production of place level measures create another condition; researchers must stamp the boundarized geospatial unit with a label. The labeling of geospatial units affords investigators a greater proficiency in communicating their findings on how place affects the individual. Some investigators unsuitably make use of the word neighborhood to label their geospatial units. We say unsuitable because the term currently possess a high but as of yet a scientifically difficult to define degree of significance—a condition which may change in the future after more investigations are done to determine what a neighborhood is and how it can be geographically boundarized. For now and because the meaning of neighborhood is given by its use and context, we argue selected geographic units should be referred to by their origin-assigned label. By origin-assigned we mean that a geographic unit should be referred to using the label that the polygon was assigned by the creator. For example, if the US Census creates a polygon and labels it “tract,” then all subsequent references to that polygon should label the geographic entity a tract.

Why should aging and frailty researchers care about the geographical boundarization and its subsequent labeling? The reduction in life-space mobility in the aged may make their frailty status more vulnerable to environmental effects. Advancing our understanding of how place affects their frailty requires that we advance our discourse on the proper use and meaning of terms.

Discussions on the production of social space (3) and geoboundarization of place have been going on for many decades (4). Early and current theorists have posited that human behavior is influenced by the context that the individual perceives as well as the social structures beyond his/her perceptions. This is why attempts to define the term neighborhood abound in the literature. Recent theoretical perspectives explain neighborhood as a geographical construct of place—where the term evokes images of an environment where local members share a common sense of place. It may be the term neighborhood has a deep psychosocial meaning for many, but the meaning of what a neighborhood is has yet to be scientifically defined—in particular, how neighborhoods should be geoboundarized. We hope that if research consistently adapts the origin-assigned labeling approach, we will continue to have a clearer discussion of how readily available spatial units capture (or fail to encompass) areas where individuals share a common sense of place.

Because its’ theoretical conceptualization are still bearing out, the operationalization, and thus measurement of neighborhood remains empirically unattainable. Despite these limitations, the recent proliferation of articles investigating the effects of sociophysical contexts on health have benefited from using the term neighborhood to make their arguments more compelling and easier to understand. Some literature makes use of the term with little attempt at a critical review of its meaning. Clapp and Wang (5) explain that a lack in definitional precision may occur because “the ordinary language definition is considered so compelling as to require little elaboration.” By way of example, when talking about “context” or “tract” effects on depression symptomatology, arguments could be rendered uninformative, weak, and/or confusing. In contrast, using the word neighborhood strengthens the flow of the prose and gives the feeling that greater insights—than those actually afforded by the research—can be derived from the findings. This is all possible because neighborhood is loaded with meaning and serves as a powerful rhetorical construct with intuitive (albeit not as of yet scientific) meaning.

Physically bounded spatial units, referred to as neighborhoods in some research, vary in their geographical scope and conceptual meaning. The dearth of attention given to this crucial matter focuses on how selected place-measures affect individual-level outcomes. Since the gold standard for defining a neighborhood is yet to exist, their geoboundarization ranges on various dimensions. For example, as described in detail elsewhere (6), neighborhoods may be defined by: personal views (e.g., place attachment); functional views (e.g., public transportation); social life views (e.g., social interaction); physical views (e.g., sidewalks); economic views (e.g., homeownership); or political views (e.g., organizations); and we add, policy views (e.g., school districts).  Because a discourse on how neighborhoods should be geoboundarized has yet to become a matter of common academic debate, advances in developing a precise scientific meaning of place effects remain hindered.

As is clear from the previous paragraph, the rapid growth and continuing interest in neighborhood effects on health has not eradicated all methodological concerns. Because the widely used abstract neighborhood remains undefined, its geographical delineation remains muddled in non-theory driven polygons. Advances in the field of mental health, have generally shown that areas with poor-quality housing, few resources, and unsafe conditions can impose stress that can lead to depression (7). In general, sociophysical environments have been found to impact individual-level outcomes—an influence that exists above and beyond the effects of the individual’s own personal attributes. How we control the definition of context will ultimately bring into being how we define place-on-individual problems, design studies, and diagnose interventions to moderate the onset of disabling frailty. The main goal of this project is to echo previous requests (8) to increase definitional precision when investigating and labeling place and its effects on mental health outcomes. Our manuscript advances the efforts of others (73) by focusing on the use and meaning of the word neighborhood. To validate these arguments, we conduct a narrow literature review of place effects on depression symptomatology in the aged.

Place Effects on Mental Health

A complete review of the literature on place and mental health can be found in Appendix A. In general, place effect studies range over a wide variety of study designs, geographical units, and area measurements. While there is some evidence that context affects health (25), there are some reasons why findings on place effects on mental well-being remain somewhat inconclusive. The inconclusiveness of existing research is in part due to the fact that cross study comparisons are difficult (if not impossible) to do. Cross study comparability on place effect studies is difficult to accomplish for two main reasons: (1) the different geographical polygons employed to measure place; and (2) the various forms of statistical methods employed to account for how macro-level factors influence micro-level outcomes. Within our literature review, the majority of studies do use the same geographical polygon: this leads to a series of more complicated questions (discussed in closing)—like, even if cross-study comparability is possible, how will it benefit us if we fail to capture a “true” measure of what a person perceives as his/her neighborhood? This study focuses on the first challenge and proposes a solution that can be accomplished through greater definitional precision—a step we hope will mitigate the plethora of issues faced by place effect investigations.

Current Study

The main purpose of this study is to discuss implications from labeling arbitrarily defined geographical polygons as neighborhoods and to provide evidence from a narrow review of literature on the aged and depression symptomatology. We focus our study on the aged because their population merits special attention given the importance that frailty in the ageing population continues to exert in shaping our society. Please note that the “age 65” threshold is used because it is the most widely used demarcation for labeling somebody as an older adult in the US. Some have argued that it may have little social meaning for many non-U.S. nations and that the 65-age-threshold will steadily become less relevant as retirement ages fluctuate and life expectancies change (26).

In our literature review, we focused on studies which investigate depression symptomatology because some have argued that depression is “the most frequent cause of emotional suffering in later life” and “significantly decreases quality of life in older adults” (27). Depression influences health outcomes beyond mental well-being. For example, some have linked it to higher rates of mortality (28). Others have recently shown that including environmental factors is important in such studies because areas may significantly affect depression in people whose activity space is limited to their area of residence (29). Depressive mood is important for mental capacity and physical activity—both of which play a key role in an aged adult’s frailty status.

While other depression screening instruments exist, we are only interested in articles where the CES-D scale is used. We focused on the CES-D complete or partial measure because it has been administered to aged adults around the world (30), has been validated with various populations (31), has been show to be reliable (31), and has been found to provide valuable insights into aged adults mental wellbeing.

We have limited our study to articles using samples from the US. This means we have excluded some high quality and relevant research (33). We decided to limit our investigation to studies using U.S. samples to facilitate the discussion of geographical polygons. That is, by only reviewing studies using US data, we can more readily critique the label used in referencing the selected geospatial unit in contrast to available geographical polygons in the area of study. The inclusion of other nations would necessitate an extensive discussion of how the various geographical entities vary by location.

Our call for greater definitional precision is aimed not at creating or eliminating controversy, but rather to transfer the discourse in place effects, aging, frailty, and mental health to a more appropriate stage: Where the substance of an investigation is not muddled by the misappropriation of terms. We are not aware of any articles where the definition of neighborhood is discussed and where a review focusing on how the term neighborhood is defined in mental health literature of older adults where the CES-D scale is used is given simultaneously. Our investigation fills this gap by first discussing implications from labeling geographical polygons as neighborhoods and by providing an example to validate our arguments by systematically identifying scientific articles published between 2000 and 2012 and empirically researching if and how the term “neighborhood” is defined.



The initial inclusion of articles for review followed three basic requirements. The first criterion for inclusion was that the item be published between January 1, 2000 and September 30, 2012. Articles meeting the first criterion were then selected if the terms “neighborhood” or its alternate spelling “neighbourhood”, and “mental health,” and/or “mental well-being,” and/or “CES-D” appeared in the title and/or the abstract. We only included articles published using the English language, which were peer-reviewed, and only used US samples in the analysis. The last selection criterion was used in order to reduce the amount of polygon comparisons made in our literature review—since compatibility of geographical units across nations would require an extensive discussion. From those being screened for relevance, we exclude all the articles where a neighborhood measure was not used as a key variable in the equation. This means all non-quantitative studies were excluded. Since all the aforementioned procedures produced a host of results, we limited our review to articles that only used a sample from our main population of interest: those age 65 and beyond. All others were disqualified from the abstract review stage.

The article search started with PubMed, after selecting for all the criteria outlined above, our initial search yielded over 102 articles. Their abstracts were reviewed by one reviewer. Any abstracts that showed potential for inclusion (i.e., met all the criteria) were retained for a full review. In order to insure full coverage, the reference sections of the full review articles were searched to identify potential articles for inclusion. The abstracts, derived from the reference section search in the initial full review sample, were evaluated for inclusion in the review. While using search engines to seek out desired articles, the abstracts in list of related articles, provided by search engines, were also searched to seek out potential candidates for inclusion.

These exhaustive procedures produced a total of more than 200 articles.  An abstract review condensed the sample to undergo full review to 98. From these, thirteen of the articles used children or youth, another sixty-three were made up of studies with samples of adults (aged 21 and below 65), and only twenty contained analysis with aged study participants.  From the aged articles 10 were excluded because they either excluded the CES-D scale or omitted the inclusion of a context-level measure (a list of all the 85 excluded articles is available from the author upon request). After completing a full review of the 98 articles, we ended up with a final sample of 11 articles to include in our literature review. Articles from the final sample of 11 are summarized in Table 1. The table provides information on how neighborhood is defined, which spatial units are used, what context measures are employed in the analysis, and a brief description of the key findings on CES-D.

Table 1 Studies using the Term “Neighborhood” and Investigating Depression Symptomatology among Older Adults (N=12)

a. Number of people per nesting unit not given by author; b.Three waves of data used in study. The numbers given in table represent the baseline configuration; c. No version of software given by author.

Since our specific interest is on how neighborhood is being defined and secondarily on how it is related to depression symptomatology, the assessment of the 11 selected articles was conducted using the following criteria: primarily (1) a description on the treatment of the term neighborhood; and secondarily on the (2) spatial unit used in the geographical boundarization of neighborhood; (3) neighborhood measure; (4) type of data; (5) statistical modeling and software used; (6) age range of sample; and (7) findings related to CES-D.



Finding related to CES-D are discussed in Appendix B; while a detail discussion on the treatment of the term neighborhood is given in Appendix C and a discourse on the use of tracts as neighborhoods is delineated in Appendix D.

Homogeneity of Tracts

Most of our reviewed articles geographically located individuals into census tract neighborhoods. The use of tracts in the investigation of health outcomes is decades old (54-55) as such, our reviewed articles using tracts could be argued to be within standard protocols in research—where the labeling of tracts as neighborhood may also be within existing customs. Since their inception in 1906, tracts have been utilized to understand the population (56). Other studies (12), as did ours, find that “neighborhood studies” are in large part forced to rely on geographic boundaries defined by governmental agencies such as the US Census Bureau. The appealing nature of tracts comes from their consistency and availability. Their benefits are however diminished given that they are primarily created for administrative purposes and not as per social scientific theory. Please note that although homogeneity need not be a necessary condition for what an individual defines as a neighborhood, we echo others and encourage researchers to examine the level of homogeneity within their spatial units (57).

Because most of our articles use tracts to measure individuals’ area characteristics, an important question then becomes: How homogeneous are tracts?  Warnings on using tracts as neighborhoods with care have been issued with greater eloquence before (12, 49).  Recent work explored within-tract homogeneity along indicators of active living, and concluded that there was “substantial within-tract heterogeneity” (57).  It is beyond the scope of this paper to either give a full review of this topic or to conduct a full quantitative investigation on it.  However, we do provide a simple example that validates the need to explore and better understand how “tract homogeneity” varies by location and variable of interest.  Our goal here is to simply give the reader a visual representation of potential within-tract heterogeneity (the opposite of the much acclaimed within-tract homogeneity)—neither of which may be required for what people define as a neighborhood area.

Explorations on tracts have shown since long ago (58) and continue to date (57) that intra-tract heterogeneity must first be explored since its presence is so frequently found. Here we only give a descriptive figure in order for the reader to get a picture of what is meant by intra-tract heterogeneity. In order to facilitate our example, we have arbitrarily selected a section in the city of Galveston, Texas, USA.  The following could easily be replicated with any other part of the country. By using Summary File 1 from the 2000 Decennial Census, we produced Figure 1 using ArcMap 9.3 (59). The figure displays the “White Only” concentration by blocks with tract boundaries outlined. As can be visually appreciated, some tracts are more homogenous on percent-White-only than others.  The readers can judge by themselves if this tiny example substantiates the need for future studies to first explore the assumed within-tract homogeneity of their macro-level variable of interest.

Figure 1 Percent “White Only” in 2000 for Galveston Texas: By Blocks with Tract Boundaries for the Central part of the Island



It seems that the geographical boundaries and measures of “neighborhoods” are largely influenced by the need to identify them in ways that optimize the availability of data.  On the treatment of “neighborhood,” we find that most use the term liberally and give limited regard to how their geographical unit may play a role in their measure of environment. With regards to CES-D, we find that in general, macro-level measures on context instability (here broadly defined) are statistically related with an increase in depression symptomatology (the micro-level outcome of interest). In short, an aged adult’s environment may play a key role in how frailty operates.

The specific aim of our project was to offer a literature review on articles focusing on the aged, who use the CES-D scale, and account for context. In particular, our aim was to show how greater definitional precision in the labeling of context measures can improve how neighborhood is defined in scientific research. There are some limitations with our project.  First, our final article sample for review may be considered by some as too narrow (although our criterion only provided such a number of cases). Second, selecting articles that investigate an aged population means we are neglecting to explore how researchers of non-aged populations treat/use/define the term neighborhood. Our use of a geographical section in the city of Galveston, Texas, U.S. no doubt leaves room for some criticisms (e.g., it is a island with a unique population and city layout). Our focus on CES-D means that we excluded the treatment of the term by all those who may be interested in investigations of other mental health measures or other ailments. Despite these limitations, we feel our project achieves its mission in delivering the clear message that context labeling needs to occur with greater care.

Our article’s first substantive contribution to the literature is our detailed discussion on the treatment of the word “neighborhood” in studies who investigate depression symptomatology using the CES-D instrument with an aged population. Our discussion amounts to more than just a long grievance, we provide a comprehensive discussion on how the use of the term limits place effect research. In particular, we offer critiques which can serve future advances in theory. Our second substantive contribution is on briefly showing that in general, more place-level instability is associated with greater levels of depression symptomatology. We now extend our contributions to outline a future research agenda.

Future research should investigate the same topic but with different age populations and different mental health outcomes (60), including studies that account for physical environment macro-level measures (61).  Researchers could also continue what others began long ago (58) and explore if/how initial tract homogeneity morphs into heterogeneity as populations within tract increase (62). Lastly, investigations on how various statistical approaches (e.g., aspatial clustering algorithms) can inform “neighborhood” boundarization should continue while being informed by theory (63).

Over a decade ago, researchers with the Social Science Research Council concluded that there was “much to be learned about neighborhood effects from studies that use census-based sources of data.” In their report, researchers are advised to continue searching for “alternative procedures for measuring neighborhoods” (64).  Most of our reviewed articles used census-based data or spatial units. Even with the large number of studies, our knowledge on the consequences of environmental conditions for older adult mental well-being remains limited. Place effect studies on the mental health of aged adults lack congruency in their definition and measure of social and physical environments.

Using the meaning-rich term “neighborhood” allows authors to make powerful arguments on how a selected environmental measure influences an individual’s mental health. We believe the use of term can at times be misleading. As was extensively discussed before, there is no agreed upon definition of what a neighborhood is—much less of what its geographical boundaries should be. Convincing arguments for the sensitivity of boundary definitions have been given elsewhere (65). Some have clearly shown that people vary in how they geographically boundarize their neighborhood (66). This is all further complicated by the fact that contexts may not affect all people in the same way—because people have different personality characteristics that allow them to adjust in different ways to challenging environments (7).

Various studies are exploring how people define their neighborhood. Early work on this topic found that when residents’ neighborhood boundary definitions are explored, the resulting geospatial units covered different spaces and created different social measures than did census-defined units (67). Articles advocating the use of a “bottom-up GIS” approach in defining a person’s neighborhood exist (68) and cutting edge “participatory photo mapping” techniques allowing for a transdisciplinary approach that integrates various digital tools with qualitative data to depict the complex nature of person’s neighborhood are advancing (69). Investigations from the phenomenological viewpoint have found that individual’s “cognitive maps” are not entirely fixed, but that they do share meaning with other geographically proximal residents (70). This line of research may be able to better validate appropriate geographical polygons, albeit proxy, for capturing a persons’ neighborhood. More research using alternative nesting units may help develop these methods into units that could be used in larger datasets and at a more practical level for research.

A call for greater definitional precisions is more than just an argument on semantics. Moving place effect studies on aging and frailty away from inconclusiveness requires that researchers be clear on how their geographical boundarizations affect their macro-level measures. The point is that until a definition of neighborhood is settled, aggregate measures of contexts will have to be arbitrarily produced by a subjectively chosen geographical polygon. Our call for greater definitional precisions is a call for making sure that place effect publications explicitly state the logic (or necessity) for their chosen geographic unit—and that said geospatial unit be initially labeled with its origin assigned label. When we begin to outline a framework for what the differing geographic units mean in place effect research, we will begin to make cross study comparability a more beneficial enterprise for understanding frailty in the aged.

Hypotheses should be born out of theory, the credibility of which should not depend on how a key term is used (71). Our call for greater definitional precision is aimed at transferring the discourse atarax is an antihistamine. this medicine is used for the treatment of allergy symptoms. it is also used to treat anxiety and tension. this medicine can be used in place effect and mental health to a more advance stage: Where the substance of an investigation is not muddled by the misappropriation of terms. We find that publications investigating place effects on the mental health (using CES-D measures) of the ageing population need greater definitional precision when labeling their selected geographical units.

Appendix A
Place Effects on Mental Health

A recent publication (74) clearly explains that growing evidence seems to support the idea that an aged person’s environment is related to their depressive mood. Recent counts of the population show that the proportion of people age 65 and older is growing, according to the 2010 US Census [see http://2010.census.gov/news/press-kits/summary-file-1.html], there are over 40 million older adult (65 and over) residents in the US. In 2010, there were 22 people over the age of 65 for every 100 people between the ages of 20 and 64—a dependency ratio which is expected to rapidly climb from 22 to 35 by 2030 [see http://www.census.gov/newsroom/releases/ archives/aging_ population/cb10-72.html].  Since older ages are, in general, characterized by the loss of psychological resources (9), depression is one of the most prominent challenges faced by aged adults (10). Some estimate that about 15 million people aged 65 and above will have a diagnosable mental illness by 2030—up from 4 million in 1970 (11). Consequently, the mental health of older adults has many policy, social, and economic implications.
Political debates on the health care of older adults have saturated media outlets for several years. There are many reasons fuelling the growing discussion. While mental health care is an issue that applies to and affects individuals of all ages (72), the ageing of the population and its accompanying health care costs are amongst the most poignant reasons for why so many are paying attention to this topic. For example, Medicare (a government funded health insurance in the U.S.) total expenditures in 2010 were 522.8 million and are projected to rise to nearly double (or 932.1 million) by 2020 as the older adult population increases (11).  In 2010, out of the 47.5 million people covered by Medicare, 83% (or 39.6 million) of them were people age 65 and older [see http://www.cms.gov/reportstrustfunds/down loads/tr2011.pdf].
The geography of health remains vibrant despite its continued search for a place in the academic setting. Research on the effects of social and physical environment on mental health continues to grow. Investigations in the field have found links between individual mental health and various network units such as family, community, and local service providers (12). Others have found that individual mental wellbeing affects the economic wellbeing of households, communities, and local governments (13). The built environment has also been shown to play a role in individual health (14). Health research’s growing capability to model statistical relationships while accounting for their geographical location has produced evidence that health is spatially patterned (15)—a theoretical assumption which has created the need for “multilevel thinking” in place effect studies on health.
Widely cited work has convincingly argued that social and physical environments are important units of analysis that merit measurement strategies and theoretical frameworks (12). Such studies have made it necessary for researchers to acknowledge the importance of contexts, and if possible, account for it in the study of mental well-being. Place effect theorist (16) have at the very least, convincingly argued that geographically proximal sociophysical structures (i.e., social and physical contexts) can either promote or undermine the mental health of mobility-limited older adults (17). For example, existing work has used the exposure-disease paradigm to explain that ambient hazards exacerbate and influence psychosocial stress—a process that affects the functioning of body systems (e.g., immune system) to the point of leading directly to illness(18).
The environment is uniquely important with aged adults because it has been found to play a more significant role in older ages as the degree of mobility decreases (19).  For example, researchers have found that older adults on average spend about 80% of their day in their home (20). If mobility in fact decreases with age, then the immediate environment may play a stronger role among aged adults as their multisystem reduction in physiological capacity increases. These findings have lead researchers to theorize that physical environment (e.g., availability of sidewalks) and social contexts (e.g., local economic deprivation) may be more important when individual deficits (e.g., decreased mobility) occur in older ages (21). Place effect research on mental health has been around for many years and its continued growth necessitates that both theory and methods be advanced in order for us to successfully answer how context affects individual-level outcomes.
In general, we can say that context affects an aged adult’s functional capacity. When this theoretical premise is combined with the aging of the US population, it becomes crucial for researchers to continue expanding our understanding of how environments affect mental well being. To date, the most general theoretical propositions to be made are that context influences health in two ways: (1) through relatively short-term influences on behaviors; and (2) through longer-term processes that accumulate to affect health outcomes (22). As summarized elsewhere (23), place effects on health investigations have focused on four general groups: (1) mortality; (2) disease; (3) mental health; and (4) health behaviors. This review focuses on mental health and in particular on the definition of neighborhood in studies of older adults where depression symptomatology is measured with the CES-D scale (24).

Appendix B
Place Effects on CES-D

Each study (listed in Table 1) reports the relationships between neighborhood level constructs and depressive symptoms (CES-D scores), cognitive function (MMSE scores with CES-D scores as a covariate) or psychological distress [operationalized as anxiety (measured by the Spielberger State Anxiety Inventory) and depressive symptoms (measured by the CES-D)]. Eight of the studies use only social context measures and two include physical attributes. Two studies found Latino/a ethnic homogeneity within census tracts to be associated with fewer depressive symptoms among Latinos/as (34, 35). Having more depressive symptoms was associated with faster rates of cognitive decline (36)—no context on CES-D cross-level interaction is reported.  Social support was shown to be a mediating variable for the association between both built environment and neighborhood climate (e.g., perceived danger) on decreased psychological distress (37, 38). Greater walkability within buffers (i.e., spatial zones) was associated with fewer depressive symptoms (39). Higher tract socioeconomic status and a higher concentration of residents 65+ were associated with fewer depressive symptoms (40). Three studies found no tract level variables associated with depressive symptoms beyond individual characteristics (41-43). Higher block socioeconomic ranking was found to be associated with more depressive symptoms (73).

Appendix C
Treatment of Neighborhood Term

The eleven place-effect studies will now be described in terms of how the neighborhood construct is treated. A total of three articles attempted to define and justified the use of the term neighborhood (35, 39, 40, 73). Only two articles questioned the use of the term neighborhood before discussing results (34, 36), while another two only do so in closing (38, 42). Three articles neither defined neighborhoods nor questioned the use of the term neighborhood (37, 41, 43). This means that the majority, of the reviewed articles do not question the use of the term neighborhood. Seven of the articles explicitly justify the use of term while the remaining three articles do so implicitly by omitting any discussion on the meaning of the term (37, 41, 43). As shown in Table 1, the majority of the articles nest (i.e., geographically identify) individuals into census tracts (34-36, 40-43). Three of the articles nest individuals into census blocks (37-38, 73) and one makes use of a spatial buffer—derived from a spatial analysis (39). Since most of the articles in our sample nest individuals into tracts, we decided to explore how the term neighborhood and tracts were linked in the articles.

Appendix D
Tracts as Neighborhoods

Most of the justification for using tracts as neighborhoods comes from a shortlist of articles and one book chapter. Some of the reviewed articles point to Subramanian et al. (44), where the author argues that when using area-based socioeconomic measures, “census tracts have been shown to be reasonable approximations for delineating neighborhoods” (Pg. 829). Subramanian et al. (44) in turn makes such a claim by citing Sampson et al. (12). However, Sampson et al. (12) point out census tracts “offer imperfect operational definitions of neighborhoods for research and policy.”  Sampson et al. (12) even goes on to explain in the following sentence that this is the reason why researchers are interested in designing strategies for defining neighborhoods. From our reading, Sampson et al. (12) does not offer a blanket validation for calling tracts neighborhoods.
There are two widely used citations, within our list of reviewed articles, which are said to justify the use of tracts as neighborhoods. Even though the authors from both Krieger et al. (45) and Krieger et al. (46) never actually advocate the use the term neighborhood, they are widely cited as providing justification for labeling tracts as neighborhoods. In the first piece, the word neighborhood is only mentioned one time in the text—when discussing others work.  Krieger and colleagues only use the term “area-based” measures when referring to their selected context measures (45). On the second publication, the word neighborhood is not mentioned in the text at all (46). In this latter piece, the authors’ again only use terms such as “census tract and block group measures” to identify their macro-level factors (46). Here again, cited work fails to lend justification in labeling tracts as neighborhoods.
In our reviewed articles, Diez-Roux et al. (47) is also credited with providing justification for calling tracts neighborhoods. The authors in this article explain that in their study, they found little difference between using census tracts or block-group measures (47). Again, the only time the term neighborhood is used in the text is in a footnote given for one of their figures (47). The authors interpret their findings using the “block-group” or “census tract” labels (47)—they never justify using the neighborhood label to refer to tracts. Another popular citation seen in our reviewed articles was Ricketts and Sawhill (48) who use tracts and make use of the word neighborhood extensively within their text. The only justification for why tracts are referred to as neighborhood is given in a footnote where they state that “as the best available statistical equivalent of a neighborhood, a tract is a small statistical subdivision of a county with generally stable boundaries and homogeneous population characteristics” (48). They do not provide citations in their footnote—nor do they reference any theoretical premises for why tract geographical boundaries should be referred to as neighborhoods.
Tienda (49) is also cited as a source that legitimizes the labeling of tracts as neighborhoods. Our reading of her piece however leads us to believe she was attempting to outline the complications with deciphering neighborhood effects. In particular, Tienda (49) makes it clear that using tracts as proxies to neighborhoods only means that researchers are measuring global characteristics of arbitrarily defined places. Tienda (49) points out that “interaction patterns” are what create neighborhoods, and that consequently, tracts as proxies to neighborhoods only capture a bounded-space that is unlikely to be defined in social terms—since the boundaries of the arbitrary geographical polygons are not determined by the social interactions within them. From our reading of Tienda (49), she does not offer a justification for labeling tracts as neighborhoods.
South and Crowder (50) are cited by some in the literature as providing evidence for why tracts can be labeled as neighborhoods. In their paper, South and Crowder (50) do use tracts and refer to them as neighborhoods extensively. Ross and Mirowsky (51) are also cited within our reviewed articles as validating the use of the term neighborhood to describe tracts. Ross and Mirowsky (51) do use the term extensively in the text and justifies it by citing Tienda (49) and South and Crowder (50). Ross and Mirowsky (51) write that census tracts are the best approximation to neighborhood.  They write: “following most prior research, we use census tracts as geographical representation of neighborhoods» (51). They cite Tienda (49) as being a source that explains how tracts offer an imperfect operationalization of neighborhoods. Ross and Mirowsky (51) validate the use of the term neighborhood to refer to tracts because they are the most “commonly available spatial entity in approximating the usual conception of neighborhood”—a statement that is followed by a citation to our previously discussed Rickets and Sawhill (48) article.
Many of our reviewed articles cited a statement issued by the US Census Bureau regarding tracts to justify their use of the neighborhood label. Since the US Census Bureau kept being cited, it is worth mentioning that in 1990 and 2000, the Census Bureau reported [see http://www.census.gov/geo/www/cob/tr_metadata.html] that when first delineated, census tracts were “designed to be homogeneous with respect to population characteristics, economic status, and living conditions.”  In 2010, the Census Bureau modified their description and explained [see http://www.census.gov/geo/www/2010census /gtc_10.html] that tracts are “small, relatively permanent statistical subdivisions of a county or equivalent entity.” Statements by the US Census Bureau were probably never meant as a scientific justification for labeling tracts as neighborhoods. The Census recognizes that their geographical polygons must only serve the needs of the legal mandate that create the institution. That is, the Census creates geographies that serve governmental administrative needs and not the theories of academicians.
Curiously, most of the articles within our full review paid little attention to alternative nesting units. Existing research has advocated the use of census block groups as more adequate neighborhood units (52-53). Others have even advocated testing various geospatial units to explore which ones are capable of producing the most robust estimates (53). Many alternative nesting units are used in the literature that have been argued to adequately serve as proxies for neighborhoods. Clearly the debate on what geographical boundaries best capture neighborhoods continues—a discussion that is unintentionally hidden when the term neighborhood is employed without regard to its scientifically ambiguous nature.



1.    Coulter, E. J. & Guralnick, L. (1959). Analysis of Vital Statistics by Census Tracts. Journal of the American Statistical Association, 54(228), 730-740.
2.    Chakraborty, J. (2011). Revisiting Tobler’s First Law of Geography: Spatial Regression Models for Assessing Environmental Justice and Health Risk Disparities. In J. A. Maantay & S. McLafferty (Eds.), Geospatial Analysis of Environmental Health. New York: Springer.
3.    Lefebvre, H. (1991). The Production of Space (Translation by Donald Nicholson-Smith). UK: Blackwell.
4.    Duncan, O. D. (1957). The Measurement of Population Distribution. Population Studies: A Journal of Demography, 11(1), 27-45.
5.    Clapp, J. M. & Wang, Y. (2006). Defining Neighborhood Boundaries: Are Census Tracts Obselete? Journal of Urban Economics, 59(2), 259-284.
6.    Talen, E. & Shah, S. (2007). Neighborhood Evaluation Using GIS: An Exploratory Study. Environment and Behavior, 39(5), 583-615.
7.    Cutrona, C. E., Wallace, G. & Wesner, K. A. (2006). Neighborhood Characteristics and Depression: An Examination of Stress Processes. Current Directions in Psychology Science, 15(4), 188-192.
8.    Chaskin, R. J. (1997). Perspectives on Neighborhood and Community: A Review of the Literature. Social Service Review, 71(4), 521–547.
9.    Koenig, H., Smiley, M. & Gonzalez, J. A. (1988). Religion, Health, and Aging: A Review and Theoretical Integration. Westport, CT: Greenwood.
10.    Solhaug, H. I., Romuld, E. B., Romild, U., et. al. (2012). Increased Prevalence of Depression in Cohorts of the Elderly: An 11-Year Follow-Up in the General Population—The HUNT Study. International Psychogeriatics, 24, 151-158.
11.    Jeste, D. V., et. al. (1999). Consensus Statement on the Upcoming Crisis in Geriatric Mental Health. Archives of General Psychiatry, 56(9), 848-853.
12.    Sampson, R. J., Morenoff, J. D., & Gannon-Rowley, T. (2002). Assessing “Neighborhood Effects”: Social Processes and New Directions in Research. Annual Review of Sociology, 28, 443-478.
13.    Mair, C., Diez-Roux, A. V. & Galea, S. (2008). Are Neighbourhood Characteristics Associated with Depressive Symptoms? A Review of Evidence. Journal of Epidemiological Community Health, 62(11), 940-946.
14.    Renalds, A., Smith, T. H. & Hale, P. J. (2010). A Systematic Review of Built Environment and Health. Family Community Health, 33(1), 68-78.
15.    Diez-Roux, A. V. & Mair, C. (2010) Neighborhoods and Health. Annals of the New York Academy of Sciences, 1186, 125-145.
16.    Macintyre, S., Ellaway, A. & Cummins, S. (2002). Place Effects on Health: How Can we Conceptualize, Operationalize and Measure Them? Social Science & Medicine, 55(1), 125-139.
17.    Burton, L., Kemp, S. P., Leung, M., Matthews, S. and Takeuchi, D. T. (2011). Communities, Neighborhoods, and Health: Expanding the Boundaries of Place. New York, NY: Springer.
18.    Gee, G. C. & Payne-Sturges, D. C. (2004). Environmental Health Disparities: A Framework Integrating Psychosocial and Environmental Concepts. Environmental Health Perspectives, 112(17), 1645-1653.
19.    Kaiser, H. J. (2009). Mobility in Old Age: Beyond the Transportation Perspective. Journal of Applied Gerontology, 28(4), 411-418.
20.    Horgas, A. L., Wilms, H. U., & Baltes, M. M. (1998). Daily Life in Very Old Age: Everyday Activities as Expression of Successful Living. The Gerontologist, 38(5), 556-568.
21.    Lawton, M. P. (1985). Housing and Living Environments of Older People. In R. H. Binstock & E. Shanas (Eds.), Handbook of Aging and the Social Sciences, 2nd Ed. (pp. 450-478). New York: Van Nostrand Reinhold.
22.    Ellen, I. G., Mijanovich, T. & Dillman, K. N. (2001). Neighborhood Effects on Health: Exploring the Links and Assessing the Evidence. Journal of Urban Affairs, 23(3-4), 391–408.
23.    Yen, I. H., Michael, Y. L. & Perdue, L. (2009). Neighborhood Environment in Studies of Health of Older Adults: A Systematic Review. American Journal of Preventive Medicine, 37(5), 455-463.
24.    Radloff, L. S. (1977). The CES-D Scale: A Self-Report Depression Scale for Research in the General Population. Applied Psychological Measurement, 1(3), 385-401.
25.    Pickett, K. E., & Pearl, M. (2001). Multilevel Analyses of Neighborhood Socioeconomic Context and Health Outcomes: A Critical Review. Journal of Epidemiology and Community Health, 55, 111-122.
26.    Stewart, R. & Lindesay, J. (2011). The Epidemiology of Depression and Anxiety. In M. T. Abou-Saleh, C. Katona & A. Kumar (Eds.), Principles and Practice of Geriatric Psychiatry (pp. 616-623). West Sussex, UK: John Wiley and Sons.
27.    Blazar, D. G. (2003). Depression in Late Life: Review and Commentary. Journal of Gerontology, 58(3), M249-M256.
28.    Ryan, J., Carriere, I., Ritchie, K. et. al. (2008). Late-Life Depression and Mortality: The Influence of Gender and Antidepressant Use. British Journal of Psychiatry, 200(3), 12-18.
29.    Vallee, J., Cadot, E., Roustit, C. et. al. (2011). The Role of Daily Mobility in Mental Health Inequalities: The Interactive Influence of Activity Space and Neighbourhood of Residence on Depression. Social Science and Medicine, 73(8), 1133-1144.
30.    Chen, R. & Copeland, J. (2011). Epidemiology of Depression: Prevalence and Incidence. In M. T. Abou-Saleh, C. Katona & A. Kumar (Eds.), Principles and Practice of Geriatric Psychiatry (pp. 479-488). West Sussex, UK: John Wiley and Sons.
31.    Beekman, A. T. F., Deeg, D. J. H., Limbeek, J. V. et. al. (1997). Criterion Validity of the Center for Epidemiologic Studies Depression Scale (CES-D): Results from a community-Based Sample of Older Subjects in the Netherlands. Psychological Medicine, 27, 231-235.
32.    Hertzong, C., Alstine, J.V., Usala, P.D. (1990). Measurement Properties of the Center for Epidemiological Studies Depression Scale (CES-D) in Older Populations. Journal of Consulting and Clinical Psychology, 2(1), 64-72
34.    Gerst, K., Miranda, P. Y., Eschbach, K. et. al. (2011). Protective Neighborhoods: Neighborhood Proportion of Mexican Americans and Depressive Symptoms in Very Old Mexican Americans. Journal of the American Geriatrics Society, 59(2), 353-358.
33.    Stafford, M., McMunn, A., & DeVogli, R. (2011). Neighborhood Social Environment and Depressive Symptoms in mid-life and beyond. Ageing and Society, 31(1),  893-910.
35.    Ostir, G. V., Eschbach, K., Markides, K. S. et. al. (2003). Neighborhood Composition and Depressive Symptoms among Older Mexican Americans. Journal of Epidemiology and Community Health, 57, 987-992.
36.    Sheffield, K. M. & Peek, M. K. (2009). Neighborhood Context and Cognitive Decline in Older Mexican Americans: Results from the Hispanic Established Populations for Epidemiologic Studies of the Elderly. American Journal of Epidemiology, 169(9), 1092-1101.
37.    Brown, C. S., Mason, C. A., Lombard, J. L. et. al. (2009a). The Relationship of Built Environment to Perceived Social Support and Psychological Distress in Hispanic Elders: The Role of “Eyes on the Street”. Journal of Gerontology, 64B(2), 234-246.
38.    Brown, S. C., Mason, C. A., Spokane, A. R. et. al. (2009b). The Relationship of Neighborhood Climate to Perceived Social Support and Mental Health in Older Hispanic Immigrants in Miami, Florida. Journal of Aging and Health, 21(3), 431-459.
39.    Berke, E. M., Gottlieb, L. M., Moudon, A. V. et. al. (2007). Protective Association Between Neighborhood Walkability and Depression in Older Men. Journal of the American Geriatrics Society, 55(4), 526-533.
40.    Kubzansky, L. D., Subramanian, S. V., Kawachi, I. et. al. (2005). Neighborhood Contextual Influences on Depressive Symptoms in the Elderly. American Journal of Epidemiology, 162(3), 253-260.
41.    Wight, R. G., Cummings, J. R., Karlamangla, A. S. et. al. (2009). Urban Neighborhood Context and Change in Depressive Symptoms in Late Life. Journal of Gerontology, 64B(2), 247-251.
42.    Aneshensel, C. S., Wight, R. G., Miller-Martinez, D. et. al. (2007). Urban Neighborhoods and Depressive Symptoms Among Older Adults. Journal of Gerontology, 62B(1), S52-S59.
43.    Hybels, C. F., Blazer, D. G., Pieper, C. F. et. al. (2006). Sociodemographic Characteristics of the Neighborhood and Depressive Symptoms in Older Adults: Using Multilevel Modeling. American Journal of Geriatrics and Psychiatry, 14(6), 498-506.
44.    Subramanian, S. V., Chen, J. T., Rehkopf, D. H. et. al. (2006). Comparing Individual- and Area-based Socioeconomic Measures for the Surveillance of Health Disparities: A Multilevel Analysis of Massachusetts Births,1989–1991. American Journal of Epidemiology, 164(9), 823-834.
45.    Krieger, N., Chen, J. T., Waterman, P. D. et. al. (2002). Geocoding and Monitoring of US Socioeconomic Inequalities in Mortality and Cancer Incidence: Does the Choice of Area-based Measure and Geographic Level Matter? The Public Health Disparities Geocoding Project. American Journal of Epidemiology, 156(5), 471-482.
46.    Krieger, N., Chen, J. T., Waterman, P. D. et. al. (2003). Choosing area based socioeconomic measures to monitor social inequalities in low birth weight and childhood lead poisoning: The Public Health Disparities Geocoding Project (US). Journal of Epidemiology and Community Health, 57(3), 186-199.
48.    Ricketts, E. R. & Sawhill, I. V.  (1988). Defining and Measuring the Underclass. Journal of Policy Analysis and Management, 7(2), 316-325.
49.    Tienda, Marta. (1991). Poor People and Poor Places: Deciphering Neighborhood Effects on Poverty Outcomes. In J. Huber (ed.), Macro-Micro Linkages in Sociology (pp. 244-262). Newbury Park, California: Sage.
50.    South, S. J. & Crowder, K. D. (1997). Escaping Distressed Neighborhoods: Individual, Community, and Metropolitan Influences. American Journal of Sociology, 102(4), 1040-1084.
51.    Ross, C. E. & Mirowsky, J. (2001). Neighborhood Disadvantage, Disorder, and Health. Journal of Health and Social Behavior, 42(3), 258-276.
52.    Goodman, A. C. (1977). Comparison of Block Group and Census Tract Data in a Hedonic Housing Price Model, Land Economics, 53(4), 483-487.
53.    Shultz, S. D. & King D. A. (2001). The Use of Census Data for Hedonic Price Estimates of Open-Space Amenities and Land Use. Journal of Real Estate Finance and Economics, 22(2-3), 239-252.
54.    Green, H. W. (1933). The Use of Census Tracts in Analyzing the Population of a Metropolitan Community. Journal of the American Statistical Association, 28(181), 147-153.
55.    Terris, M. (1948). Relation of Economic Status to Tuberculosis Mortality by Age and Sex. AmericanJournal ofPublic Health, 38(8), 1061-1071.
56.    Krieger, N. (2006). A Century of Census Tracts: Health and the Body Politic (1906-2006). Journal of Urban Health: Bulletin of the New York Academy of Medicine, 83(3), 355-361.
57.    Riva, M., Arraricio, P., Gauvin, L. et. al. (2008). Establishing the Soundness of Administrative Spatial Units for Operationalising the Active Living Potential of Residential Environments: An Exemplar for Designing Optimal Zones. International Journal of Health Geographics, 7(1), 43-56.
58.    Myers, J. K. (1954). Note on the homogeneity of census tracts: A methodological Problem in urban ecological research. Social Forces, 4(2):364-366.
59.    ESRI 2011. ArcGIS Desktop: Release 10. Redlands, CA: Environmental Systems Research Institute.
60.    Weich, S.,Twigg, L., Holt, G. et. al. (2003). Contextual Risk Factors for the Common Mental Disorders in Britain: A Multilevel Investigation of the Effects of Place. Journal of Epidemiology and Community Health, 57, 616-621.
61.    Payne-Sturges, D., & Gee G. C. (2006). National Environmental Health Measures for Minority and Low-income Populations: Tracking Social Disparities in Environmental Health. Environmental Research, 102(3);154-171.
62.    Chen, W., Myles, J., & Picot, G. (2011). Why Have Poorer Neighbourhoods Stagnated Economically while the Richer Have Flourished?: Neighbourhood Income Inequality in Canadian Cities. Urban Studies 1(1), 1-32.
63.    Rey, S.J., Anselin, L., Folch, D.C., et. al. (2010). Measuring Spatial Dynamics in Metropolitan Areas. Working paper number 12 from the GeoDa Center for Geospatial analysis and computation at the Arizona State University.
64.    Brooks-Gunn, J. Duncan, G. J., Leventhal, T. & Aber, J. L. (1997). Lessons learned and future directions for research on the neighborhoods in which children live. In J. Brooks-Gunn, G. J. Duncan & J. L. Aber (Eds.), Neighborhood Poverty: Contexts and Consequences for Children Volume 1 (pp. 268-267). New York: Russell Sage Foundation.
65.    Hipp, J. R. (2007). Block, Tract, and Levels of Aggregation: Neighborhood Structure and Crime and Disorder as a Case in Point. American Sociological Review, 72(5), 659-680.
66.    Coulton, C. J., Korbin, J., Chan, T., Su, M. (2001). Mapping Residents’ Perceptions of Neighborhood Boundaries: A Methodological Note. American Journal of Community Psychology, 29(2), 371-383.
67.    Coulton, C. J., Chan, T., & Mikelbank, K. (2011) Finding Place in Making Connections Communities: Applying GIS to Residents’ Perceptions of Their Neighborhoods. The Urban Institute, Making Connections Research Series; 2011b:1-45.
68.    Talen, E. (2000). Bottom-Up GIS: A New Tool for Individual and Group Expression in Participatory Planning. Journal of the American Planning Association, 66(3), 279-294.
69.    Dennis, S. F., Gaulocher, S., Carpiano, R. M. et. al. (2009). Participatory Photo Mapping (PPM): Exploring an Integrated Method for Health and Place Research with Young People. Health & Place, 15(2), 466-473.
70.    Campbell, E., Henly, J. R., Elliott, D. S., Irwin, K. (2009). Subjective Constructions of Neighborhood Boundaries: Lessons from a Qualitative Study of Four Neighborhoods. Journal of Urban Affairs, 31(4), 461-490.
71.    Selznick, P. (1996). In Search of Community. In W. Vitek & W. Jackson (Eds.), Rooted in the Land: Essays on Community and Place (pp. 195-203). New Haven, Connecticut: Yale University Press.
72.    Knight, B. G., & Sayegh, P. (2011). Mental health and Aging in the 21st Century. Journal of Aging and Social Policy, 23(3), 228-243.
73.    Everson-Rose, S., Skarupski, K. A., Barnes, L. L., et al. (2011). Neighborhood socioeconomic conditions are associated with psychosocial functioning in older black and white adults. Health and Place, 17, 793-800.
74.    Julien, D., Richard, L., Gauvin, L., et al. 2012. Neighborhood characteristics and depressive mood among older adults: An integrative review. International Psychogeriatrics, doi: 10.1017/S1041610211002894.





1. Department of Clinical & Health Psychology, University of Florida, Gainesville, FL, USA; 2. Department of Neuroscience, University of Florida, Gainesville, FL, USA; 3. Department of Biostatistical Sciences, Wake Forest School of Medicine, Winston-Salem, NC, USA; 4. Department of Pharmacotherapy and Translational Research, University of Florida, Gainesville, FL, USA; 5. Center for Pharmacogenomics, University of Florida, Gainesville, FL, USA; 6. Stanford Prevention Research Center, Department of Medicine, Stanford University, Stanford, CA, USA; 7. Department of Aging and Geriatric Research, University of Florida, Gainesville, FL, USA; 8. Department of Epidemiology and Medicine, University of Pittsburgh, Pittsburgh, PA, USA; 9. Sticht Center on Aging, Wake Forest School of Medicine, Winston-Salem, NC, USA; 10. Pennington Biomedical Research Center, Louisiana State University, Baton Rouge, LA, USA; 11. Klein Buendel Inc., Golden, CO, USA

Corresponding author: Vonetta Dotson, Ph.D., Department of Clinical & Health, Psychology, University of Florida, P.O. Box 100165, Gainesville, FL 32610-0165, USA. Phone: +1 (352) 273-6041. Fax: +1 (352) 273-6156. Email: vonetta@phhp.ufl.edu.



Background: Converging evidence suggests that physical activity is an effective intervention for both clinical depression and sub-threshold depressive symptoms; however, findings are not always consistent. These mixed results might reflect heterogeneity in response to physical activity, with some subgroups of individuals responding positively, but not others. Objectives: 1) To examine the impact of genetic variation and sex on changes in depressive symptoms in older adults after a physical activity (PA) intervention, and 2) to determine if PA differentially improves particular symptom dimensions of depression. Design: Randomized controlled trial. Setting: Four field centers (Cooper Institute, Stanford University, University of Pittsburgh, and Wake Forest University). Participants: 396 community-dwelling adults aged 70–89 years who participated in the Lifestyle Interventions and Independence for Elders Pilot Study (LIFE-P). Intervention: 12-month PA intervention compared to an education control. Measurements: Polymorphisms in the serotonin transporter (5-HTT), brain-derived neurotrophic factor (BDNF), and apolipoprotein E (APOE) genes; 12-month change in the Center for Epidemiologic Studies Depression Scale total score, as well as scores on the depressed affect, somatic symptoms, and lack of positive affect subscales. Results: Men randomized to the PA arm showed the greatest decreases in somatic symptoms, with a preferential benefit in male carriers of the BDNF Met allele. Symptoms of lack of positive affect decreased more in men compared to women, particularly in those possessing the 5-HTT L allele, but the effect did not differ by intervention arm. APOE status did not affect change in depressive symptoms. Conclusions: Results of this study suggest that the impact of PA on depressive symptoms varies by genotype and sex, and that PA may mitigate somatic symptoms of depression more than other symptoms. The results suggest that a targeted approach to recommending PA therapy for treatment of depression is viable.


Key words: Exercise intervention, depression, aging, BDNF, APOE, serotonin transporter.



Traditional treatment for late-life depression with pharmacotherapy is efficacious in many patients (1) but often is accompanied by deleterious side effects, including increased falls and hyponatraemia (2). Consequently, recent efforts have focused on alternative non-pharmacological interventions, particularly for older adults experiencing clinically significant sub-threshold depressive symptoms. Converging evidence suggests that physical activity (PA) is an effective intervention for both clinical depression and sub-threshold depressive symptoms (3). Clinical trials have found similar efficacy for PA and antidepressant medication in the treatment of depression (4), with PA providing greater protection against relapse compared to medication (5).

Other studies, including previous work in the Lifestyle Interventions and Independence for Elders Pilot Study (LIFE-P) (6), showed no overall impact of physical activity increases on depressive symptoms. These inconsistencies may be due in part to methodological issues, but may also reflect heterogeneity in response to PA, with some subgroups of individuals responding to PA, but not others. Attempts to target traditional antidepressant treatments have examined the impact of genetic variation on treatment efficacy. These studies focus on genes that are putatively related to antidepressant mechanisms, such as the serotonin transporter gene (5-HTT) for selective serotonin reuptake inhibitors, or genes that are related to the risk of depression, including the brain-derived neurotrophic factor (BDNF) and apolipoprotein E (APOE) genes. The role of these particular genes in the antidepressant response has been documented by recent meta-analyses (7-9), which showed that the APOE ε4 allele, the BDNF Met allele, and the 5-HTT long (L) allele are associated with higher likelihood of positive response and remission after antidepressant treatment. It is unclear whether genetic differences also impact the efficacy of PA in treating depression or depressive symptoms, as the evidence is limited and results are mixed. One study reported that young adults with at least one 5-HTT L allele showed greater reductions in depressive symptoms after a 5-week exercise intervention (10). In contrast, a recent cross-sectional study in middle-aged adults found that the BDNF Val66Met polymorphism did not moderate the relationship between self-reported physical activity and depressive symptoms (11). This question has not been investigated in older adults.

Also unclear is whether PA impacts particular symptom dimensions of depression more than others. Depression is a clinically heterogeneous disorder that comprises a variety of different symptoms (e.g., negative affect, reduced positive affect, and somatic symptoms). Emerging evidence suggests that specific dimensions of depressive symptoms are related to specific brain changes and domains of cognitive dysfunction (12, 13). Corroborating the distinction between symptom dimensions of depression, there is evidence of distinct vascular, degenerative and inflammatory contributors to different depressive symptom clusters (14), and genetic work has shown significant positive familial correlations for different symptom dimensions (15). As such, it is possible that PA would improve certain types of depressive symptoms, but not others. 

Moreover, the impact of PA on depressive symptoms may vary by sex. Numerous studies have shown that men and women not only differ in their risk for depression and vulnerability to depression-related negative sequelae, but also in the associations of genotype with depression risk and response to depression treatment (16, 17). Some studies have reported a sex difference in the relationship between PA and depressive symptoms, with the effect being found exclusively or to a greater extent in either men (e.g., 18) or women (e.g., 11). A recent meta-analysis of randomized trials showed a stronger effect of exercise in men (19).

The goal of the present investigation was to expand upon previous work in the LIFE-P cohort (6) by examining the role of variants in the BDNF, 5-HTT, and APOE genes in the antidepressant response to a PA intervention, and by separately examining different symptom dimensions of depression. Based on previous studies documenting a better treatment response in depressed carriers of the APOE ε4 allele, the BDNF Met allele, and the 5-HTT L allele, we expected LIFE-P participants possessing these genetic markers to show the greatest reduction in depressive symptoms after a 12-month PA intervention compared to an educational control intervention.



Data for the present investigation came from the Lifestyle Intervention and Independence for Elders Pilot (LIFE-P) Study, a randomized controlled trial evaluating the effect of PA on physical performance measures linked to mobility disability. Details of the study design for LIFE-P have been described elsewhere (20). Briefly, community-dwelling adults aged 70–89 years were recruited from four field centers (Cooper Institute, Stanford University, University of Pittsburgh, and Wake Forest University). Participants were required to have a sedentary lifestyle (<20 minutes of structured PA per week during the previous month) and a score of <10 on the Short Physical Performance Battery (21), but be able to complete a 400-m walk unaided in less than 15 minutes. Major exclusion criteria included presence of severe heart failure, uncontrolled angina, and other severe illnesses that might interfere with PA. The National Institutes of Health and the Institutional Review Boards of all participating institutions approved the study procedures. All participants provided written informed consent. A total of 424 participants were enrolled between May 2004 and February 2005. Participants were randomly assigned to either a PA or health education control arm, described briefly below and in more detail elsewhere (20). At baseline and at 6- and 12-month follow-up, comprehensive standardized assessments were conducted by trained research staff masked to intervention assignment. Of the 424 participants in LIFE-P, 396 consented to DNA testing, of which 365 had available baseline and follow-up data for the variables used in the present investigation. Sample characteristics for the present investigation are provided in Table 1.


Table 1 Demographic characteristics, genotype and baseline CES-D scores

* Total sample varied across gene: BDNF = 362, 5-HTT = 347, APOE = 359


Physical Activity (PA) Intervention

Participants randomly assigned to the PA arm performed aerobic, strength, flexibility, and balance training in both center- and home-based settings. Walking was the primary mode of aerobic actvity, given its widespread popularity and ease of administration across a broad segment of the older adult population. Other forms of endurance activity (e.g., stationary cycling) were utilized when regular walking was contraindicated. Each center-based session began with a brief warm-up, followed by 40 minutes of moderate-intensity walking, and concluded with 15 minutes of flexibility and balance training exercises. 

PA intensity was gradually increased over the first 2-3 weeks, with a target of reaching moderate intensity as assessed by the Borg scale (22), a numerical scale indicating a rating of perceived exertion from minimal to maximal. Participants were asked to walk at a target intensity of 13 (somewhat hard) and perform strength training at an intensity of 15-16 (hard). The proportion of center-based to home-based sessions changed during the course of treatment: The number of center-based sessions was reduced to two times per week and home-based activities were increased during weeks 9-24. In the maintenance phase (week 25 to the end of the study), participants were encouraged to perform home-based PA a minimum of 5 days per week, and one weekly center-based session was offered.

Control Intervention

A successful aging health education intervention served as an attention control arm. Half of participants were randomly assigned to the control arm and attended education workshops on health topics of relevance to older adults, such as nutrition, medication use, foot care, and preventive medicine. Each class was concluded with gentle seated upper extremity stretching. These classes lasted approximately 60 minutes and were given weekly for the first 26 weeks, and then monthly until the end of the study. This amount of contact time is similar to the contact time for control arms employed in other successful randomized trials of PA in older adults (6).

Depressive Symptom Measurement

The Center for Epidemiologic Studies Depression Scale (CES-D; 23) was administered at baseline and at 12 months. The CES-D has shown a consistent factor structure across numerous patient populations and ethnic groups, which has been confirmed by meta-analysis (24). The four-factor structure of the CES-D includes depressed affect (e.g., sadness and fearfulness), somatic symptoms (e.g., loss of appetite, concentration difficulties), lack of positive affect (e.g., diminished capacity to experience pleasure), and interpersonal difficulties (e.g., perceived problems in social relationships) subscales. Baseline and 12-month scores were used in analyses for the current investigation.  


DNA was isolated from whole blood using a commercially available DNA isolation kit, and the concentration and purity (260/280) were determined by a UV spectrophotometer. The 5-HTT gene short (484 bp) and long (528 bp) variant alleles (insertion/deletion polymorphism) were genotyped by PCR amplified fragment length polymorphism. The following forward 5’-GCGTTGCCGCTCTGAATGC-3’ and reverse 5’-GAGGGACTGAGCTGGACAACCAC-3’ PCR primers were used to amplify the short (484 bp) and long (528 bp) variant alleles. The amplified PCR products from DNA samples along with positive and negative controls were run on 1% agarose gel and the results were interpreted by two trained personnel. The BDNF Val66Met rs6265 and APOE rs4412 polymorphisms were genotyped by Taqman® genotyping method on ABI 7900 HT platform using the genotyping probes C__11592758_10 and C___2230322_20, respectively. The assays were performed and analyzed according to the manufacture’s recommendations (Life Technologies, CA, USA). The APOE rs420358 polymorphism was genotyped by Pyrosequencing genotyping method (25), using the following PCR and sequencing primers: forward biotinylated PCR primer 5’-GCGGACATGGAGGACGTG-3’, reverse PCR primer 5’-TACACTGCCAGGCGCTTCT-3’, and reverse sequencing primer 5’-ACTGCACCAGGCGGC-3’. The Pyrosequencing reactions were carried out using the Pyrosequencing HS 96 platform according to the manufacturer’s recommendations and the genotypes were automatically called by PSQ HS 96 SNP software (Qiagen, Valencia, CA, USA). 

Statistical Analyses

Analyses of variance compared baseline continuous demographic characteristics and CES-D scores for the full sample, as well as CES-D scores by sex and genotype. Chi-square tests compared genotype frequency and categorical demographic characteristics across sex.

Our primary hypotheses were tested with analyses of covariance. Dominant models rather than additive models were used for analysis of 5-HTT (presence vs. absence of the L allele) and APOE (presence vs. absence of the ε4 allele) due to small sample sizes for the interaction terms of genotype by sex in the additive models. The 12-month changes in total CES-D score and its subscales were treated as dependent variables in separate models. The interpersonal problems subscale was not included in the analyses due to the restricted range of scores on the subscale, which comprises only 2 items. Intervention arm, genotype, and sex were included as predictors. The 3-way interaction among intervention arm, genotype, and sex and their pairwise two-way interactions were also tested. Stratified analyses were explored for interactions with p <.10. Non-significant interaction terms were removed in stages until a final, parsimonious model was reached. All models included age, race/ethnicity, testing site, and baseline outcome as covariates. A significance threshold of p <.05 was used.


Demographic characteristics, genotype distribution and baseline CES-D scores are summarized for the total sample and for women and men separately in Table 1. No significant sex differences were observed in demographic characteristics or baseline CES-D scores (ps ≥ .085). The percentage of participants who possessed the BDNF Met allele, 5-HTT L allele, and APOE ε4 allele was 29.6%, 77.5%, and 25.9%, respectively. All frequencies were in Hardy-Weinberg equilibrium. Genotype distribution was similar for men and women (ps ≥ .41). Baseline scores on the CES-D, including the depressed affect, somatic symptoms, and lack of positive affect subscales are presented by sex and genotype in Table 2. The mean CES-D total score at baseline was 7.29±6.75; scores were similar across genotype and sex × genotype (ps ≥ .15). 


Table 2 Baseline CES-D scores by sex and genotype

CES-D = Center for Epidemiologic Studies Depression Scale; BDNF = brain-derived neurotrophic factor; APOE = apolipoprotein E


Figure 1 Least squares means for 12-month change in somatic symptoms by sex and intervention group


Table 3 Least squares means for change in CES-D scores after adjusting for intervention by sex, intervention group, and genotype


PA = physical activity intervention group; SA = successful aging education control group; CES-D = Center for Epidemiologic Studies Depression Scale; BDNF = brain-derived neurotrophic factor; APOE = apolipoprotein E


The main effect of intervention arm was not significant in any of the models (ps ≥ .11). ANCOVAs examining the impact of PA, BDNF status and sex on 12-month change in depressive symptoms revealed an intervention arm × sex effect, F(1, 347) = 4.13, p = .043, for somatic symptoms (Figure 1; Table 3). Men in the PA arm showed the greatest decrease in somatic symptoms over 12 months, with minimal change in the control arm. Symptoms decreased to a lesser extent in women, and were similar for the PA and control arms. Based on a marginal 3-way interaction between intervention, BDNF status, and sex, F(1, 347) = 3.09, p = .079, we performed analyses stratified by intervention arm. In the PA arm, men who possessed the Met allele showed greater decreases in somatic symptoms compared to Met negative men and to women, F(1, 172) = 4.54, p = .043 (Figure 2). No significant effects were observed in the control arm.


Figure 2 Least squares means for 12-month change in somatic symptoms by BDNF genotype, sex and intervention group


In the analysis of 5-HTT status, significant sex, F(1, 335) = 4.99, p = .026, and genotype × sex, F(1, 335) = 4.27, p = .035, effects were observed for the lack of positive affect subscale, such that those depressive symptoms decreased most after 12 months in men compared to women, and more so in 5-HTT L-negative compared L-positive men (Figure 3). Results did not differ by intervention arm.

No significant effects were found in the APOE analyses (ps ≥ .20).


Figure 3 Least squares means for 12-month change in symptoms of lack of positive affect by 5-HTT genotype and sex



Converging evidence suggests that PA improves mood, leads to depression remission, and protects against recurrence of depressive symptoms, and previous work has identified genetic variations that predict treatment response to antidepressant medication (3-5, 7-9). It is less clear whether or not factors such as genetic variation and sex moderate the antidepressant effect of PA, or if PA differentially affects distinct symptom dimensions of depression in a nonclinical older population. The current study addressed these questions. We found a three-way relationship between intervention, BDNF genotype and sex for somatic symptoms of depression, suggesting that the impact of PA on depressive symptoms indeed varies by genotype, sex, and symptom dimension.

Our primary finding was a preferential benefit of PA on somatic depressive symptoms in male BDNF Met allele carriers. The impact of BDNF genotype on change in depressive symptoms after PA is consistent with our hypothesis and parallels the antidepressant literature. Although there are reports of no effect of BDNF on antidepressant response (26), recent meta-analyses have concluded that Met carriers have a more positive response to pharmacotherapy for depression (7). Additionally, a recent study reported that Met-positive individuals had a more positive acute mood response to a bout of moderate intensity exercise relative to those without a Met allele (27). Our finding is in contrast to a recent study that reported no moderating effect of BDNF on the relationship between PA and depressive symptoms (11); however, the former study was based on self-report of PA in middle aged participants, unlike our randomized PA trial in older adults.

Taken together, it appears that for some individuals, PA, similar to antidepressant medication, may impact depressive symptoms through its effect on BDNF. This is not surprising given the functions of the BDNF protein in the nervous system. BDNF is a member of the neurotrophin family that has a critical role in central nervous system functions including cell survival and differentiation, axonal growth, and the function and plasticity of synapses (28). BDNF is a neurotransmitter modulator that is highly expressed in the central nervous system, particularly in the hippocampus and other brain regions related to mood, such as the frontal lobes and striatum (29). In depression, BDNF expression is reduced in areas such as the hippocampus and prefrontal cortex and increased in the nucleus acumbens and amygdala, but antidepressant treatment normalizes BDNF levels (30). PA has also been shown to increase levels of BDNF (31), which may explain the antidepressant effect of PA. In light of the association between the BDNF Met allele and reduced BDNF secretion (32), it is plausible that Met carriers preferentially benefit from PA due to their inherent physiological disadvantage.

In the present study, the benefits of PA on depressive symptoms were specific to somatic symptoms; there was no impact of PA on total symptoms or on symptoms of depressed affect or lack of positive affect. A growing body of evidence supports the importance of examining symptom dimensions of depression separately rather than treating symptoms as homogeneous. Different symptom dimensions of depression appear to have distinct cognitive and neural correlates, etiological contributors, and genetic associations (12-15), and at least one study found that the impact of genotype on antidepressant treatment response is specific to particular symptom dimensions (33). Given the physical nature of PA interventions, the specificity of our finding to somatic symptoms could be seen as an indication that PA simply improves physical functioning. However, the somatic subscale of the CES-D includes items such as “I had trouble keeping my mind on what I was doing” and “…everything I did was an effort,” which do not directly reflect physical functioning. Thus, the nature of the items comprising the subscale suggests that the improvements are not solely due to changes in physical functioning. Future work relating changes in depressive symptoms after PA to depression biomarkers may clarify the mechanism underlying the effect. For example, depressive symptoms, particularly in older adults, are associated with white matter changes in the brain. We recently showed that this may be driven by somatic symptoms and depressed affect, as only these factors were associated with greater white matter lesion volume in older men, and with increases in white matter lesion volume over time in both men and women (12). PA improves white matter integrity (34), thus, it is plausible that PA-related changes in white matter results in reduced somatic symptoms of depression after PA. 

We found that men, but not women, showed improvements in somatic symptoms with increases in PA. This is in line with previous findings of sex differences in depression risk and correlates, as well as response to depression treatment (12, 16, 17). For example, there is evidence that BDNF genotype is associated with higher depression risk only in men or to a greater extent in men (11). Previous work has also shown sex effects on the association of BDNF with antidepressant treatment response, but these studies found an effect only in women (35). The reasons for our sex effect are unclear. The result is not explained by greater regression to the mean in men, since men and women did not differ in somatic symptom scores at baseline. A recent meta-analysis (36) concluded that the BDNF Val66Met polymorphism may play a larger role in the neurobiological underpinnings of depression in men than in women. This conclusion was based on evidence of sexual dimorphisms in brain structures involved in the neurobiology of depression, particularly the hippocampus (37). The possibility that the biological underpinnings of depressive symptoms differ in men and women is supported by studies showing stronger and more consistent neural correlates of depression in men compared to women (e.g., 38). Thus, men may benefit more from PA interventions if PA leads to change in underlying neurobiological mechanisms that are affected more in depressed men than in depressed women. Future work with larger samples will be important for replicating our finding, particularly given the fairly small number of men in our pilot study.

We did not find the expected effect of 5-HTT or APOE on the antidepressant effect of PA. Our hypothesis was based on evidence linking the 5-HTT L allele and APOE ε4 genotype with better response to treatment with antidepressant medication (7-9). Additionally, 5-HTT has recently been shown to impact reductions in depressive symptoms after a 5-week exercise intervention in young adults (10). However, the relationship between these genotypes and response to PA is not fully established, as the literature is limited and some studies failed to find these associations (39, 40). Our ability to detect possible relations may have been affected by insufficient power or the low severity of depressive symptoms. There is also evidence that gene-gene interactions should be considered, particularly the interaction between 5-HTT and BDNF polymorphisms (41, 42). We were unable to test possible gene-gene interactions in this pilot study due to the sample size. Future extensions of the present work in the full LIFE trial (n = 1635) will allow us to address this issue in a larger sample. 

Our results should be considered in the context of the characteristics of the study sample. Participants in LIFE-P were sedentary older adults at risk for disability who ranged in age from 70-89 years and who were not recruited based on a depression diagnosis. The impact of genetic variation on changes in depressive symptoms after PA may differ in clinical samples of individuals diagnosed with major depression or in a different age cohort. Given the variety of genes that have been linked to depression risk and antidepressant treatment response, future studies should also examine the association of additional genotypes on changes in depressive symptoms after PA in subthreshold depression as well as major depression. Nonetheless, this pilot study provides initial evidence that the type of comprehensive PA intervention used in the LIFE-P trial may potentially impact the somatic domain of depressive symptoms in older men, even without a diagnosis of clinical depression. In addition, the results suggest that PA may be particularly impactful in mitigating somatic depressive symptomatology in men with specific genetic markers, paving the way for future research aimed at targeting such subgroups of older adults. This line of work will be important for establishing the clinical significance of changes in depressive symptoms after PA. Based on evidence that the presence of even one depressive symptom increases the risk for negative functional outcomes (43, 44), the 1- to 2-point changes in depressive symptoms in the current study are likely to be clinically meaningful. Thus, this work has the potential to inform targeted non-pharmacological interventions for subthreshold depression in older adults.


Funding: The LIFE-P study was supported by the National Institutes of Health/National Institute of Aging Cooperative Agreement (U01AG22376) and sponsored in part by the Intramural Research Program, National Institute for Aging, and National Institutes of Health. Dr. Dotson is partially supported by the UF Claude D. Pepper Center (NIA P30 AG028740-01). The Pittsburgh Field Center was partially supported by the Pittsburgh Claude D. Pepper Center P30 AG024827. The Wake Forest University Field Center is, in part, supported by the Claude D. Older American Independence Pepper Center #1 P30 AG21332. Yale University Thomas M. Gill, M.D. Dr. Gill is the recipient of a Midcareer Investigator Award in Patient-Oriented Research (K24AG021507) from the National Institute on Aging. The sponsors had no role in the design and conduct of the study; in the collection, analysis, and interpretation of data; in the preparation of the manuscript; or in the review or approval of the manuscript.

Acknowledgments: The LIFE Study: Lifestyle Interventions and Independence for Elders, Pilot. http://clinicaltrials.gov/show/NCT00116194. 

Research Investigators for Pilot Phase of LIFE: Cooper Institute, Dallas, TX: Steven N. Blair, P.E.D. – Field Center Principal Investigator, Timothy Church, M.D., Ph.D., M.P.H. – Field Center Co-Principal Investigator, Jamile A. Ashmore, Ph.D. Judy Dubreuil, M.S. Georita Frierson, Ph.D. Alexander N. Jordan, M.S., Gina Morss, M.A. Ruben Q. Rodarte, M.S. Jason M. Wallace, M.P.H. National Institute on Aging: Jack M. Guralnik, M.D., Ph.D. – Co-Principal Investigator of the Study: Evan C. Hadley, M.D. Sergei Romashkan, M.D., Ph.D. Stanford University, Palo Alto, CA Abby C. King, Ph.D. – Field Center Principal Investigator: William L. Haskell, Ph.D. – Field Center Co-Principal Investigator: Leslie A. Pruitt, Ph.D. Kari Abbott-Pilolla, M.S. Karen Bolen, M.S. Stephen Fortmann, M.D. Ami Laws, M.D. Carolyn Prosak, R.D. Kristin Wallace, M.P.H. Tufts University: Roger Fielding, Ph.D. Miriam Nelson, Ph.D. Dr. Fielding’s contribution is partially supported by the U.S. Department of Agriculture, under agreement No. 58-1950-4-401. Any opinions, findings, conclusion, or recommendations expressed in this publication are those of the author(s) and do not necessarily reflect the view of the U.S. Dept of Agriculture. University of California, Los Angeles, Los Angeles, CA Robert M. Kaplan, Ph.D., M.A. VA San Diego Healthcare System and University of California, San Diego, San Diego, CA Erik J. Groessl, Ph.D. University of Florida, Gainesville, FL Marco Pahor, M.D. – Principal Investigator of the Study Michael Perri, Ph.D. Connie Caudle Lauren Crump, M.P.H Sarah Hayden Latonia Holmes Cinzia Maraldi, M.D. Crystal Quirin University of Pittsburgh, Pittsburgh, PA Anne B. Newman, M.D., M.P.H. – Field Center Principal Investigator Stephanie Studenski, M.D., M.P.H. – Field Center Co-Principal Investigator Bret H. Goodpaster, Ph.D., M.S. Nancy W. Glynn, Ph.D. Erin K. Aiken, B.S. Steve Anthony, M.S. Sarah Beck (for recruitment papers only) Judith Kadosh, B.S.N., R.N. Piera Kost, B.A. Mark Newman, M.S. Jennifer Rush, M.P.H. (for recruitment papers only) Roberta Spanos (for recruitment papers only) Christopher A. Taylor, B.S. Pam Vincent, C.M.A. Diane Ives, M.P.H Wake Forest University, Winston-Salem, NC Stephen B. Kritchevsky, Ph.D. – Field Center Principal Investigator Peter Brubaker, Ph.D. Jamehl Demons, M.D. Curt Furberg, M.D., Ph.D. Jeffrey A. Katula, Ph.D., M.A. Anthony Marsh, Ph.D. Barbara J. Nicklas, Ph.D. Jeff D. Williamson, M.D., M.P.H. Rose Fries, L.P.M. Kimberly Kennedy Karin M. Murphy, B.S., M.T. (ASCP) Shruti Nagaria, M.S. Katie Wickley-Krupel, M.S. Data Management, Analysis and Quality Control Center (DMAQC) Michael E. Miller, Ph.D. – DMAQC Principal Investigator Mark Espeland, Ph.D. – DMAQC Co-Principal Investigator Fang-Chi Hsu, Ph.D. Walter J. Rejeski, Ph.D. Don P. Babcock, Jr., P.E. Lorraine Costanza Lea N. Harvin Lisa Kaltenbach, M.S. Wei Lang, Ph.D. Wesley A. Roberson Julia Rushing, M.S. Scott Rushing Michael P. Walkup, M.S. 

Conflict of Interest: The authors have no conflicts of interest to report.



1. Tedeschini E, Levkovitz Y, Iovieno N, et al. Efficacy of antidepressants for late-life depression: a meta-analysis and meta-regression of placebo-controlled randomized trials. J Clin Psychiatry 2011;72(12):1660-8.

2. Coupland C, Dhiman P, Morriss R, et al. Antidepressant use and risk of adverse outcomes in older people: population based cohort study. BMJ 2011;343:d4551.

3. Stathopoulou G, Powers MB, Berry AC, Smits JAJ, Otto MW. Exercise Interventions for Mental Health: A Quantitative and Qualitative Review. Clin Psychol Sci Pract 2006;13(2):179-193.

4. Blumenthal JA, Babyak MA, Doraiswamy PM, et al. Exercise and pharmacotherapy in the treatment of major depressive disorder. Psychosom Med 2007;69(7):587-96.

5. Babyak M, Blumenthal JA, Herman S, et al. Exercise treatment for major depression: maintenance of therapeutic benefit at 10 months. Psychosom Med 2000;62(5):633-8.

6. Matthews MM, Hsu FC, Walkup MP, et al. Depressive symptoms and physical performance in the lifestyle interventions and independence for elders pilot study. J Am Geriatr Soc 2011;59(3):495-500.

7. Kato M, Serretti A. Review and meta-analysis of antidepressant pharmacogenetic findings in major depressive disorder. Mol Psychiatry 2010;15(5):473-500.

8. Murphy GM, Kremer C, Rodrigues H, Schatzberg AF. The apolipoprotein E epsilon4 allele and antidepressant efficacy in cognitively intact elderly depressed patients. Biol Psychiatry 2003;54(7):665-73.

9. Porcelli S, Fabbri C, Serretti A. Meta-analysis of serotonin transporter gene promoter polymorphism (5-HTTLPR) association with antidepressant efficacy. Eur Neuropsychopharmacol 2012;22(4):239-58.

10. Rethorst CD, Landers DM, Nagoshi CT, Ross JT. Efficacy of exercise in reducing depressive symptoms across 5-HTTLPR genotypes. Med Sci Sports Exerc 2010;42(11):2141-7.

11. Gujral S, Manuck SB, Ferrell RE, Flory JD, Erickson KI. The BDNF Val66Met polymorphism does not moderate the effect of self-reported physical activity on depressive symptoms in midlife. Psychiatry Res 2014;218(1-2):93-7.

12. Kirton JW, Resnick SM, Davatzikos C, Kraut MA, Dotson VM. Depressive Symptoms, Symptom Dimensions, and White Matter Lesion Volume in Older Adults: A Longitudinal Study. Am J Geriatr Psychiatry 2014;22(12):1469-77.

13. Korten NC, Penninx BW, Kok RM, et al. Heterogeneity of late-life depression: relationship with cognitive functioning. Int Psychogeriatr 2014; 26(6): 953-63.

14. Naarding, P, RA Schoevers, JG Janzing, et al. A study on symptom profiles of late-life depression: the influence of vascular, degenerative and inflammatory risk-indicators. J Affect Disord 2005;88(2):155-62.

15. Korszun A, Moskvina V, Brewster S, et al. Familiality of symptom dimensions in depression. Arch Gen Psychiatry 2004;61(5):468-74.

16. Brummett BH, Boyle SH, Siegler IC, et al. Effects of environmental stress and gender on associations among symptoms of depression and the serotonin transporter gene linked polymorphic region (5-HTTLPR). Behav Genet 2008;38(1):34-43.

17. Regan CO, Kearney PM, Savva GM, Cronin H, Kenny RA. Age and sex differences in prevalence and clinical correlates of depression: first results from the Irish Longitudinal Study on Ageing. Int J Geriatr Psychiatry 2013;28(12):1280-7.

18. Bhui K, Fletcher A. Common mood and anxiety states: gender differences in the protective effect of physical activity. Soc Psychiatry Psychiatr Epidemiol 2000;35(1):28-35.

19. Rethorst CD, Wipfli BM, Landers DM. The antidepressive effects of exercise: a meta-analysis of randomized trials. Sports Med 2009;39(6):491-511.

20. Rejeski WJ, Fielding RA, Blair SN, et al. The lifestyle interventions and independence for elders (LIFE) pilot study: design and methods. Contemp Clin Trials 2005;26(2):141-54.

21. Guralnik JM, Simonsick EM, Ferrucci L, et al. A short physical performance battery assessing lower extremity function: association with self-reported disability and prediction of mortality and nursing home admission. J Gerontol 1994;49(2):M85-94.

22. Borg G. Borg’s Perceived Exertion and Pain Scales. 1998, Champaign, IL: Human Kinetics. viii, 104 p.

23. Radloff LS. The CES-D Scale: A self-report depression scale for research in the general population. Applied Psychological Measurement 1977;1(3):385-401.

24. Shafer AB. Meta-analysis of the factor structures of four depression questionnaires: Beck, CES-D, Hamilton, and Zung. J Clin Psychol 2006;62(1):123-46.

25. Langaee T, Ronaghi M. Genetic variation analyses by Pyrosequencing. Mutat Res 2005;573(1-2):96-102.

26. Illi A, Viikki M, Poutanen O, et al. No support for a role for BDNF gene polymorphisms rs11030101 and rs61888800 in major depressive disorder or antidepressant response in patients of Finnish origin. Psychiatr Genet 2013;23(1):33-5.

27. Bryan A, Hutchison KE, Seals DR, Allen DL. A transdisciplinary model integrating genetic, physiological, and psychological correlates of voluntary exercise. Health Psychol 2007;26(1):30-9.

28. Huang EJ, Reichardt LF. Neurotrophins: roles in neuronal development and function. Annu Rev Neurosci 2001;24:677-736.

29. Martinowich K, Manji H, Lu B. New insights into BDNF function in depression and anxiety. Nat Neurosci 2007;10(9):1089-93.

30. Krishnan V, Nestler EJ. The molecular neurobiology of depression. Nature 2008;455(7215):894-902.

31. Adlard PA, Perreau VM, Cotman CW. The exercise-induced expression of BDNF within the hippocampus varies across life-span. Neurobiol Aging 2005;26(4):511-20.

32. Rybakowski JK. BDNF gene: functional Val66Met polymorphism in mood disorders and schizophrenia. Pharmacogenomics 2008;9(11):1589-93.

33. Huezo-Diaz P, Uher R, Smith R, et al. Moderation of antidepressant response by the serotonin transporter gene. Br J Psychiatry 2009;195(1):30-8.

34. Voss MW, Heo S, Prakash RS, et al. The influence of aerobic fitness on cerebral white matter integrity and cognitive function in older adults: results of a one-year exercise intervention. Hum Brain Mapp 2013;34(11):2972-85.

35. Domschke K, Lawford B, Laje G, et al. Brain-derived neurotrophic factor ( BDNF) gene: no major impact on antidepressant treatment response. Int J Neuropsychopharmacol 2010;13(1):93-101.

36. Verhagen M, van der Meij A, van Deurzen PA, et al. Meta-analysis of the BDNF Val66Met polymorphism in major depressive disorder: effects of gender and ethnicity. Mol Psychiatry 2010;15(3):260-71.

37. Madeira MD, Lieberman AR. Sexual dimorphism in the mammalian limbic system. Prog Neurobiol 1995;45(4):275-333.

38. Lavretsky H, Kurbanyan K, Ballmaier M, et al. Sex differences in brain structure in geriatric depression. Am J Geriatr Psychiatry 2004;12(6):653-7.

39. Poland RE, Lesser IM, Wan YJ, et al. Response to citalopram is not associated with SLC6A4 genotype in African-Americans and Caucasians with major depression. Life Sci 2013;92(20-21):967-70.

40. Taylor MJ, Sen S, Bhagwagar Z. Antidepressant response and the serotonin transporter gene-linked polymorphic region. Biol Psychiatry 2010;68(6):536-43.

41. Martinowich K, Lu B. Interaction between BDNF and serotonin: role in mood disorders. Neuropsychopharmacology 2008;33(1):73-83.

42. Lenze E, Rawsom K, Dixon D, et al. An interaction between polymorphisms in brain-derived neurotrophic factor and serotonin transporter genes predicts depressive symptoms in older adults after hip fracture. Am J Geriatr Psychiatry 2014;22(3):S108-S109 [Abstract].

43. Judd LL, Akiskal HS, Paulus MP. The role and clinical significance of subsyndromal depressive symptoms (SSD) in unipolar major depressive disorder. J Affect Disord 1997;45(1-2):5-17; discussion 17-8.

44. Zuidersma M, Ormel J, Conradi HJ, de Jonge P. An increase in depressive symptoms after myocardial infarction predicts new cardiac events irrespective of depressive symptoms before myocardial infarction. Psychol Med 2012;42(4):683-93.