Depression is frequent in older adults, affecting

Transcription

1 Longitudinal Associations Between Frequency and Depressive Symptoms in Older Adults: Results from the VoisiNuAge Study Dominic Julien, PhD,* Lise Gauvin, PhD, Lucie Richard, PhD,* ** Yan Kestens, PhD, and Hélene Payette, PhD BACKGROUND: Cross-sectional studies show that walking is associated with depression among older adults, but longitudinal associations have rarely been examined. The aim of this study was to investigate longitudinal associations between walking frequency and depressive in older adults to determine which variable is the stronger prospective predictor of the other. DESIGN: Longitudinal; four repeated measures over 5 years. SETTING: Population-based sample of urban-dwelling older adults living in the Montreal metropolitan area. PARTICIPANTS: Participants from the VoisiNuAge study aged 68 to 84 (N = 498). MEASUREMENTS: Main exposures: depressive (Geriatric Scale) and number of walking days in previous week (Physical Activity Scale for the Elderly). Covariates: age, education, and number of chronic illnesses. Cross-lagged panel analyses were performed in the entire sample and in sex-stratified subsamples. RESULTS: Depressive predicted walking frequency at subsequent time points (and more precisely, higher depressive were related to fewer walking days), but walking frequency did not predict depressive at subsequent time points. Stratified analyses From the *Institut de Recherche en Santé Publique de l Université de Montréal, Centre de Recherche du Centre Hospitalier de l Université de Montréal, Département de Médecine Sociale et Préventive, Centre de Recherche Léa-Roback sur les Inégalités Sociales de Santé de Montréal, Faculté des Sciences Infirmieres,**Centre de Recherche de l Institut Universitaire de Gériatrie de Montréal, Université de Montréal, Montréal, Québec, Centre de Recherche sur le Vieillissement, Centre de Santé et des Services Sociaux, Institut Universitaire de Gériatrie de Sherbrooke, and Département des Sciences de la Santé Communautaire, Faculté de Médecine et des Sciences de la Santé, Université de Sherbrooke, Sherbrooke, Québec, Canada. Address correspondence to Dominic Julien, Direction de Santé Publique de l Agence de la Santé et des Services Sociaux de Montréal, 1301, Sherbrooke East, Montreal, Quebec, Canada, H2L 1M3. dominic. julien@umontreal.ca DOI: /jgs revealed that prospective associations were statistically significant in women but not men. CONCLUSION: The longitudinal association between walking frequency and depressive is one in which depressive predict reduced walking frequency later. Higher depressive are more likely a cause of reduced walking because of time precedence than vice versa. Future research on longitudinal relationships between meeting physical activity recommendations and depression are warranted. J Am Geriatr Soc 61: , Key words: depression; walking; motor activity; aged; longitudinal studies is frequent in older adults, affecting between 1% and 5% of people aged 65 and older. 1 is associated with mental suffering; risk of suicide; and poor physical, cognitive, and social functioning. 2 Older adults with depression use health and medical services two to three times as often as those without depression. 3 With the aging of the global population, it is important to identify factors that may alleviate depressive in elderly people. 4 Cross-sectional studies have reported significant associations between walking and depression in older adults 5 7 but do not provide any information on the direction of the causality. To the knowledge of the authors of the current study, three longitudinal studies have examined prospective relationships between walking and depression in older adults, and results were mixed. In one study, walking was not associated with future depression, 8 whereas in another, walking distance predicted future depression. 9 In a third study, walking was not associated with future depression, but depression was related to future walking habits. 10 One of these studies focused on older Japanese-American men 9 and another on older Hispanic adults, 10 so results may not generalize to other populations of older adults. Two of JAGS 61: , , Copyright the Authors Journal compilation 2013, The American Geriatrics Society /13/$15.00

2 JAGS DECEMBER 2013 VOL. 61, NO. 12 LONGITUDINAL ASSOCIATIONS WALKING AND DEPRESSION 2073 these studies 8,9 used regression analyses, which may not be well suited to the study of phenomena that change across time. Cross-lagged panel analysis is a better approach to investigating longitudinal associations between two variables than regression analysis because it allows for examining prospective associations between variables at multiple time points in a single analysis. Moreover, although cross-lagged panel analyses are not intended to determine causality per se, they examine prospective relationships in a way that establishes which variable is the stronger prospective predictor of the other, thereby suggesting which variable is a more likely cause of the other because of time precedence 11 (the cause occurring before the effect). Prospective associations between walking and depression in older adults have not been extensively examined, so additional research is needed to clarify the plausible direction of associations. Uncovering these associations is important for health promotion professionals, clinical psychologists, physicians, and nurses. For example, if walking habits predict future depression, with more walking being associated with lower depression, then promoting walking may be an effective strategy for preventing depression in elderly adults. In contrast, if walking is a consequence of depression, poor mental health may interfere with walking and therefore negatively affect physical health. Thus, helping to alleviate depressive may also affect physical health. The aim of this study was to investigate longitudinal associations between walking and depressive in a population-based sample of urban-dwelling older adults in an effort to determine which variable is the stronger prospective predictor of the other. METHODS Participants and Procedure Participants were taken from the VoisiNuAge Study, which investigates relationships between neighborhood environments and health-related behaviors such as walking and social participation in older adults. The VoisiNuAge database was created from the merging of two existing datasets: the Québec Longitudinal Study on Nutrition and Successful Aging (NuAge), a 5-year observational study on nutrition and successful aging, 12,13 and Montreal Epidemiological and Geographic Analysis of Population Health Outcomes and Neighbourhood Effect (MEGAPHONE), a database for health research derived from geographic information systems, allowing for geocoding of VoisiNuAge participants at the address level. 14 The NuAge cohort (N = 1,793) is an age- and sexstratified random sample for the regions of Montreal, Laval, and Sherbrooke in the province of Quebec, Canada. Inclusion criteria for the NuAge cohort were aged 68 to 84, free of activity of daily living disabilities (bathing and showering, cooking, dressing), without cognitive impairment (Modified Mini-Mental State Examination score >79), able to walk one block or climb a flight of stairs without rest, willing to commit to a 5-year study, and French or English speaking. Those reporting heart failure ( New York Heart Association Class II); chronic obstructive pulmonary disease requiring oxygen therapy or oral corticosteroids; inflammatory digestive diseases; or cancer treated using radiation therapy, chemotherapy, or surgery in the past 5 years were excluded. Global participation rate (sample studied/total eligible subjects) was 58.6%. Participants were followed four times over 5 years ( ; (baseline),,, and ) and underwent a series of nutritional, functional, medical, biological, and social computer-assisted interviews (William, Multispectra, Montreal, Canada) conducted by trained research dieticians and nurses following standardized procedures. 13 The VoisiNuAge study focused on participants who resided in the Montreal metropolitan area (n = 848). The sample was limited to those who were still in the cohort at (n = 725), meaning drop-outs (n = 102) and participants who died (n = 21) were excluded. Participants with incomplete data on the variables described below were also excluded from analyses (n = 227), leaving a sample of 498 participants. A flowchart of participant inclusion appears in Figure 1. Measures Depressive Symptoms The Geriatric Scale (GDS) 15 was used to assess depressive. The GDS is a 30-item questionnaire with a yes no response format; scores from 11 to 20 suggest mild depression and scores of 21 and higher suggest NuAge, a study on nutrition and successful aging N = 1,793 Participants living in Montreal, Laval, and Sherbrooke (Canada) VoisiNuAge, a study on neighborhood environments and healthy aging n = 848 Participants living in Montreal metropolitan area (Montreal and Laval) VoisiNuAge participants still in the cohort at n = 725 Participants in the current study n = 498 MEGAPHONE, a database for health research Excluded participants n = 945 Participants not living in Montreal metropolitan area Excluded participants n = 102 dropouts n = 21 deceased Excluded participants Figure 1. Flowchart of participant inclusion. n = 227 for incomplete data

3 2074 JULIEN ET AL. DECEMBER 2013 VOL. 61, NO. 12 JAGS moderate to severe depression, but continuous GDS scores were used in the cross-lagged panel analyses. Higher scores on the GDS indicate higher levels of depressive. The mean Cronbach alpha across different measurement times in the VoisiNuAge sample (n = 848) was 0.83 (range ), indicating high internal consistency of the measure. One question taken from the Physical Activity Scale for the Elderly (PASE) 16 was used to assess the number of days that walking episodes occurred over the past week: Over the past 7 days, how often did you walk outside your home or yard for any reason? For example, for fun or exercise, walking to work, walking the dog, etc.? (never (0 days), seldom (1 2 days), sometimes (3 4 days), and often (5 7 days); recoded as 0, 1.5, 3.5, and 6 days). Although there are no validity data on the PASE item assessing frequency of walking episodes, the item shares considerable resemblance with a validated question from the International Physical Activity Questionnaire 17 and thus shows face validity. This single PASE item has been used in other studies 5,18 to assess walking. The PASE showed good test retest reliability (correlation coefficient = 0.75) and satisfactory convergent validity with health, strength, and balance. 16 Sociodemographic and Health Characteristics Covariates were sex, age, years of education, and number of chronic illnesses based on a list of 23 reported medical conditions (Table 1). Study Design The current study used a longitudinal design. Age and years of education were collected at ; GDS scores, walking, and number of chronic illnesses were collected at each time point. The ethics committees of the University Geriatrics Institutes in Montreal and Sherbrooke approved the research, and respondents signed an informed consent form. Statistical Analysis Group differences between included and excluded VoisiNuAge participants were tested on variables of interest. Descriptive analyses were conducted to characterize respondents included in the final sample (n = 498). Using a procedure described elsewhere, 19 scores for change from to, to, and to were computed for depressive and walking frequency, respectively, to exclude the possibility that a small proportion of participants who experienced change drove cross-lagged panel analyses results. Average proportions of participants presenting increase, decrease, or unchanged depressive or number of walking days were computed, as were standard deviations for change scores. Then cross-lagged panel analyses were conducted to estimate prospective associations between walking frequency and depressive at the four measurement times. Four models were created following previously Table 1. Characteristics of VoisiNuAge Participants Included in the Analyses Characteristic Entire Sample, n = 498 Women, n = 262 Men, n = 236 Age at inception, mean SD Education, years, mean SD Number of chronic illnesses, mean SD a Potentially clinically depressed (GDS 11), n (%) 40 (8.0) 28 (10.7) 12 (5.1) 48 (9.6) 32 (12.2) 16 (6.8) 55 (11.0) 32 (12.2) 23 (9.7) 52 (10.4) 34 (13.0) 18 (7.6) GDS score, mean SD Days of walking in previous week, mean SD SD = standard deviation; GDS = Geriatric Scale. a Arthritis/rheumatism, glaucoma/ocular disease, edema, asthma, emphysema/chronic bronchitis, high blood pressure, heart trouble, circulatory problems in arms or legs, diabetes mellitus, ulcers (of the digestive systems), other digestive problems (vomiting, constipation, diverticulosis), liver or gallbladder disease, kidney disease, urinary problems (prostate), osteoporosis, cancer, anemia, thrombosis/cerebral hemorrhage/cerebrovascular accident, Parkinson s disease, thyroid and gland problems, skin disorders, epilepsy, other diseases (specified). outlined guidelines. 11 Model 1 is a base model examining simultaneously whether walking frequency predicted itself at subsequent time points (walking frequency at predicting walking frequency at,, and ; walking frequency at predicting walking frequency at and ; walking frequency at predicting walking frequency at ) and whether depressive predict themselves at subsequent time points. Model 2 investigated whether walking frequency predicted future depressive and consisted of the base model with the addition of constraints for walking frequency predicting depressive at subsequent time points (walking frequency at predicting depressive at, walking frequency at predicting depressive at, and walking frequency at predicting depressive at ). Model 3 investigated whether depressive predicted future walking frequency. It also consisted of the base model with the addition of constraints for depressive predicting walking frequency at subsequent time points. Model 4 investigated bidirectional associations between walking frequency and depressive. It also consisted of the base model with the addition of constraints for walking frequency and depressive predicting each other at subsequent time points.

4 JAGS DECEMBER 2013 VOL. 61, NO. 12 LONGITUDINAL ASSOCIATIONS WALKING AND DEPRESSION 2075 The overarching objective of the cross-lagged panel analyses was to determine whether Model 2, 3, or 4 provided a significantly better fit of the data than the base model or either of Model 2 or 3 in the case of Model 4 (see below). A chi-square difference test (D v 2 ) was used to establish fit improvement. Models 2 and 3 were compared with the base model (Model 1), and Model 4 was compared with the base model if Models 2 and 3 did not significantly improve the fit of the data or Models 2 or 3 if either of these models provided significantly improved fit over the base model. Thus, if Model 2 provided significant fit improvement, walking frequency would be thought to predict future depressive. If Model 3 provided significant fit improvement, then depressive would be thought to predict future walking frequency. If Model 4 provided significant fit improvement, then walking frequency and depressive would both be thought to have effects on the other variable (bidirectional relationship) at subsequent time points. In addition, data fit was considered good if the models met the following criteria: ratio of v 2 to degrees of freedom (df) of less than 2.0, Comparative Fix Index (CFI) greater than 0.95, Normed Fit Index (NFI) greater than 0.95, Root Mean Square Error of Approximation (RMSEA) of 0.06 or less, 20 and Tucker-Lewis Index (TLI) of 0.95 or greater. 21 Path coefficients (from walking frequency predicting subsequent depressive or depressive predicting subsequent walking frequency) were also examined in models that significantly improved fit. These path coefficients are comparable with regression analysis coefficients such that the direction and statistical significance can be interpreted in the same way. frequency and depressive were adjusted for age and education at and for chronic illnesses at every time point. frequency and depressive symptom error terms were allowed to covary at every time point. Analyses were performed in the entire sample and sex-stratified subsamples. All analyses were performed using PASW statistical software (version 18, SPSS Inc., Chicago, IL) and AMOS 19 (Arbuckle, JL, IBM SPSS, Chicago, IL). RESULTS Participant Characteristics When comparing VoisiNuAge participants who had dropped out, died, and other VoisiNuAge respondents, those who died were significantly older (P =.04) and walked less often outside their home (P =.047) than those dropping out, and those who dropped out reported significantly more depressive at than the remaining VoisiNuAge participants (P =.02). Of the remaining participants, those who were excluded from the analyses because of incomplete data were significantly older (P <.001); reported more chronic illnesses at (P <.001), (P =.001), and (P =.04); had higher depression scores at the four measurement points (P.001); and walked less often outside their home at (P =.01) and (P =.01) than those included in the analyses. Characteristics of respondents included in the analyses are shown in Table 1. The mean age of participants at cohort inception was (range 68 84). Women represented 52.6% of the final sample. The number of chronic illnesses reported increased slightly over time (with the exception of ). Table 1 also presents participant characteristics stratified according to sex. Women reported more chronic illnesses and depressive but fewer days of walking at every time point than men (t-tests not computed). Change in and over Time The number of participants categorized as potentially clinically depressed according to GDS cutoff scores was fairly stable across time, as were mean depression scores (Table 1), although 38.4% of participants reported a decrease in depressive at time points after (40.7% of women, 35.9% of men), 41.9% of participants showed an increase (41.6% of women, 42.1% of men), and 19.8% of participants reported no change (17.7% of women, 22.0% of men). The average standard deviation in depressive symptom change was 3.0 (3.1 in women, 2.9 in men). On average, 30.6% of participants reported a decrease in walking frequency after (31.9% of women, 29.1% of men), 20.6% of participants reported an increase (20.9% of women, 20.2% of men), and 48.9% of participants reported no change (47.2% of women, 50.7% of men). The average standard deviation in change in walking frequency was 2.4 (2.5 in women, 2.4 in men). These results suggest that a substantial proportion of participants reported change in depressive and walking frequency over time, indicating that a small proportion of participants did not drive the cross-lagged panel analyses results. Cross-Lagged Panel Analysis Results Square root transformations were applied to depressive symptom scores, years of education, and number of chronic illnesses to improve the normality of distributions. Table 2 presents fit indices for the entire and sex-stratified subsamples. In the entire sample, overall fit of the base model (Model 1) was excellent (v 2 /df = 1.54, CFI = 0.99, NFI = 0.98, RMSEA = 0.03, TLI = 0.99). The model in which depressive predicted walking frequency at subsequent time points (Model 3) provided a significantly better fit of the data than the base model (D v 2 (3) = 15.13, P.002). For Model 3, the fit indices were v 2 /df = 1.34, CFI = 1.00, NFI = 0.98, RMSEA = 0.03, and TLI = 0.99 (Table 2). Path coefficients from depressive to walking frequency and from depressive to walking frequency were statistically significant and negative (b = 0.23, P =.008 and b = 0.19, P =.02, respectively), and a trend toward significant associations was observed from depressive to walking frequency (b = 0.17, P =.07) (Figure 2). Overall, this model supports the idea that more depressive were related to fewer days of walking in the future. An increase of 1 point on the square rooted depressive symptom scale (which corresponds to an increase of approximately 5.5 points on the original scale) was associated with a 21% decrease

5 2076 JULIEN ET AL. DECEMBER 2013 VOL. 61, NO. 12 JAGS Table 2. Cross-Lagged Panel Analyses Results Examining Longitudinal Associations Between Frequency and Depressive Symptoms in VoisiNuAge Participants Included in the Analyses Model v 2 (df) v 2 / df Comparative Fix Index Normed Fit Index Root Mean Square Error of Approximation Tucker- Lewis Index Model Comparisons v 2 Difference Test (df) Total (n = 498) Model 1: Base model (54) Model 2: frequency (51) vs (3) predicting depressive Model 3: Depressive (51) vs (3) a predicting walking frequency Model 4: Bidirectional (48) vs (3) Women (n = 262) Model 1: Base model (54) Model 2: frequency (51) vs (3) predicting depressive Model 3: Depressive (51) vs (3) a predicting walking frequency Model 4: Bidirectional (48) vs (3) Men (n = 236) Model 1: Base model (54) Model 2: frequency (51) vs (3) predicting depressive Model 3: Depressive (51) vs (3) predicting walking frequency Model 4: Bidirectional (48) vs (6) Df = degrees of freedom; v 2 = chi-square. Models control for age and education measured at and for number of chronic illnesses at every time point. a P <.01. (mean of significant path coefficients) in walking frequency 1 year later. The models in which walking frequency predicted depressive at subsequent time points.17 a.23 c.19 b Figure 2. Best-fitting model (Model 3) for the cross-lagged panel analyses results among 498 older adults from the VoisiNuAge Study. For greater ease of presentation, associations between walking at and, and, and and and associations between depressive at and, and, and and are not shown. Models control for age and education measured at and for chronic illnesses at every time point (not shown). refers to walking frequency, and depression refers to depressive. P < a.10, b.05, c.01. Full graphs for all models are available from the first author upon request. (Model 2) and the bidirectional model (Model 4) did not provide significant fit improvement (Table 2). Further analyses were performed in women and men separately. In women, the overall fit of the base model (Model 1) was excellent (v 2 /df = 1.44, CFI = 0.99, NFI = 0.96, RMSEA = 0.04, TLI = 0.99) (Table 2). The model in which depressive predicted walking frequency at subsequent time points (Model 3) provided a significant fit improvement over the base model (D v 2 (3) = 14.40, P.002) (Table 2). The fit indices were v 2 / df = 1.24, CFI = 0.99, NFI = 0.97, RMSEA = 0.03, and TLI = 0.99 (Table 2). The path coefficient from depressive to walking frequency was statistically significant (b = 0.34, P =.004), and a trend toward significant associations was found from depressive to walking frequency (b = 0.24, P =.06) (for the path from depressive to walking frequency, b = 0.19, P =.10) (Figure 3). This model suggests that an increase of approximately 5.5 points on the original depressive symptom scale is associated with a 34% decrease in walking frequency from to. Models 2 and 4 did not provide significantly better fit (Table 2). In men, the overall fit of the base model (Model 1) was excellent (v 2 /df = 1.40, CFI = 0.99, NFI = 0.96, RMSEA = 0.04, TLI = 0.98), but none of the other models provided significant fit improvement (Table 2).

6 JAGS DECEMBER 2013 VOL. 61, NO. 12 LONGITUDINAL ASSOCIATIONS WALKING AND DEPRESSION 2077 DISCUSSION.24 a b The aim of this study was to investigate longitudinal associations between walking frequency and depressive in older adults in an effort to determine which variable is the stronger prospective predictor of the other. In the entire sample and in the subsample of women, the models in which depressive predicted walking frequency at subsequent time points provided significant fit improvement over the base model; higher depressive scores were or tended to be related to fewer walking days at subsequent time points. An increase of approximately 5.5 points on the original depressive symptom scale was associated with a 21% decrease in walking frequency 1 year later (for and ) in the entire sample and with a 34% decrease in walking frequency 1 year later (for ) in the subsample of women. These associations were independent of age, education, and number of chronic illnesses and were not driven by a small subset of participants reporting change in depressive or walking frequency across time. The models in which walking frequency predicted depressive at subsequent time points and the bidirectional models did not significantly improve fit, so it was concluded that the longitudinal association between walking frequency and depressive is one in which greater depressive predict reduced walking frequency later on and therefore that depressive are a plausible cause of walking frequency because of time precedence rather than vice versa. This causal effect seems more prominent in women than men. Because no fit improvement or significant associations were found in the subsample of men, it was also concluded that findings observed in women mainly explained the results in the entire sample. Underreported depressive in men, who are traditionally less likely to disclose having mental health problems, may explain the sex differences. Alternatively, negative life events or depressive mood may affect women s health behaviors such as walking more than men. Figure 3. Best-fitting model (Model 3) for the cross-lagged panel analyses results in a subsample of 262 women in the VoisiNuAge Study. For greater ease of presentation, associations between walking at and, and, and and and associations between depressive at and, and, and and are not shown. Model controls for age and education measured at and for chronic illnesses at every time point (not shown). refers to walking frequency, and depression refers to depressive. P < a.10, b.01. Results pertaining to prospective associations between depressive and walking frequency in the entire sample are consistent with a study reporting that walking was not a significant predictor of future depression. 8 Similarly, another study reported that walking did not predict future depressive but that depressive predicted subsequent walking. 10 Conversely, another study reported that walking predicted future depression in older Japanese-American men, 9 which differences in samples or assessment of walking (distance vs frequency) might explain. In the current analyses, path coefficients between depressive and future walking frequency were statistically significant or close to statistical significance in the entire sample and the subsample of women. A lack of power due to smaller sample size may explain trends, so results should be replicated in larger samples, and sex differences should be further investigated. Strengths of this study include its longitudinal design and the use of cross-lagged panel analyses that allowed for insights into the direction of associations between variables. One limitation of the current study is that respondents may not be representative of the overall population of older adults because they lived in urban areas and appeared to be more educated and wealthier than the general population of older adults. Participants included in the analyses reported fewer depressive than those excluded, and average depression scores were low, so the strength of the associations reported may be underestimated in comparison with the general population of older adults. Atypical external (and personal) conditions such as weather and health might have affected walking frequency in the previous week. 22 Finally, walking frequency was focused on without taking into account duration of walking episodes because, in the response options for the measure of walking, the shortest duration category was 1 hour per day, which was too broad within the context of a population survey for older adults to adequately ascertain different durations of walking. Replication of results with a more-sensitive assessment of walking activity is warranted. In conclusion, although significant associations between walking frequency and subsequent depressive were not found, it seems premature to conclude that promoting walking is not a relevant strategy to protect against depressive. There are also ample data to recommend walking for physical health reasons. Future research could investigate prospective associations between meeting recommendations for physical activity levels and depressive in older people. ACKNOWLEDGMENTS Conflict of Interest: The editor in chief has reviewed the conflict of interest checklist provided by the authors and has determined that the authors have no financial or any other kind of personal conflicts with this paper. This work was supported by a postdoctoral fellowship from the Institut de Recherche en Santé Publique de l Université de Montréal to DJ, the Canadian Institutes of Health Research (Grants MOP and MOP-62842), and the Fonds de la Recherche en Santé du Québec

7 2078 JULIEN ET AL. DECEMBER 2013 VOL. 61, NO. 12 JAGS (Grants to LR and to YK). LG holds a Canadian Institutes of Health Research/Centre de Recherche en Prevention de l Obésité Applied Public Health Chair on Neighbourhoods, Lifestyle, and Healthy Body Weight. Author Contributions: All authors: conception, design, analyses, interpretation of data, drafting of the manuscript or revising it critically for important intellectual content. Sponsor s Role: The sponsors had no role in conception, design, analyses, interpretation of data, or in the drafting, review, or approval of the manuscript. REFERENCES 1. Fiske A, Wetherell JL, Gatz M. in older adults. Annu Rev Clin Psychol 2009;5: Blazer DG. in late life: Review and commentary. J Gerontol A Biol Sci Med Sci 2003;58A: Langa KM, Valenstein MA, Fendrick AM et al. Extent and cost of informal caregiving for older Americans with of depression. Am J Psychiatry 2004;161: Reid L, Planas LG. Aging, health, and depressive : Are women and men different? J Womens Health 2002;11: Julien D, Gauvin L, Richard L et al. The role of social participation and walking in depression among a population-based sample of urban-dwelling older adults: Results from the VoisiNuAge study. Can J Aging 2013;32: Mobily KE, Rubenstein LM, Lemke JH et al. and depression in a cohort of older adults: The Iowa 65+ Rural Health Study. J Aging Phys Act 1996;4: Yoshiuchi K, Nakahara R, Kumano H et al. Yearlong physical activity and depressive in older Japanese adults: Cross-sectional data from the Nakanojo Study. Am J Geriatr Psychiatry 2006;14: Morgan K, Bath PA. Customary physical activity and psychological wellbeing: A longitudinal study. Age Ageing 1998;27(Suppl 3): Smith TL, Masaki KH, Fong K et al. Effect of walking distance on 8-year incident depressive in elderly men with and without chronic disease: The Honolulu-Asia Aging Study. J Am Geriatr Soc 2010;58: Perrino T, Mason CA, Brown SC et al. The relationship between depressive and walking among Hispanic older adults: A longitudinal, crosslagged panel analysis. Aging Ment Health 2010;14: Martens MP, Haase RF. Advanced applications of structural equation modeling in counseling psychology research. Couns Psychol 2006;34: Gaudreau P, Morais JA, Shatenstein B et al. Nutrition as a determinant of successful aging: Description of the Quebec longitudinal study NuAge and results from cross-sectional pilot studies. Rejuvenation Res 2007;10: Payette H, Gueye NDR, Gaudreau P et al. Trajectories of physical function decline and psychological functioning: The Québec Longitudinal Study on Nutrition and Successful Aging (NuAge). J Gerontol B Psychol Sci Soc Sci 2011;66B:i82 i Daniel M, Kestens Y. MEGAPHONE: Montreal Epidemiological and Geographic Analysis of Population Health Outcomes and Neighbourhood Effect. Montreal: Centre de Recherche du Centre Hospitalier de l Université de Montréal, Yesavage JA, Brink TL, Rose TL et al. Development and validation of a geriatric depression screening scale: A preliminary report. J Psychiatr Res 1983;17: Washburn RA, Smith KW, Jette AM et al. The Physical Activity Scale for the Elderly (PASE): Development and evaluation. J ClinEpidemiol 1993;46: Craig CL, Marshall AL, Sj ostr om M et al. International Physical Activity Questionnaire: 12-country reliability and validity. Med Sci Sports Exerc 2003;35: Gauvin L, Richard L, Kestens Y et al. Living in a well-serviced urban area is associated with maintenance of frequent walking among seniors in the VoisiNuAge study. J Gerontol B Psychol Sci Soc Sci 2012;67B: Brown SC, Mason CA, Perrino T et al. Longitudinal relationships between neighboring behavior and depressive in Hispanic older adults in Miami, Florida. J Community Psychol 2009;37: Ullman JB. Structural Equation Modeling. Using Multivariate Statistics, 5th ed. Boston: Pearson Education Inc., Hu L, Bentler PM. Cutoff criteria for fit indexes in covariance structure analysis: Conventional criteria versus new alternatives. Struct Equation Model Multidiscip J 1999;6: Schuit AJ, Schouten EG, Westerterp KR et al. Validity of the Physical Activity Scale for the Elderly (PASE): According to energy expenditure assessed by the doubly labeled water method. J Clin Epidemiol 1997;50: