- Email: [email protected]
When Did Old Age Stop Being Depressing? Depression Trajectories of Older Americans and Britons 2002–2012
Accepted Manuscript Title: When Did Old Age Stop Being Depressing? Depression Trajectories of Older Americans and Britons 2002-2012 Author: Gindo Tampubolon, Asri Maharani PII: DOI: Reference:
S1064-7481(17)30350-0 http://dx.doi.org/doi: 10.1016/j.jagp.2017.06.006 AMGP 863
To appear in:
The American Journal of Geriatric Psychiatry
Received date: Revised date: Accepted date:
1-2-2017 2-6-2017 5-6-2017
Please cite this article as: Gindo Tampubolon, Asri Maharani, When Did Old Age Stop Being Depressing? Depression Trajectories of Older Americans and Britons 2002-2012, The American Journal of Geriatric Psychiatry (2017), http://dx.doi.org/doi: 10.1016/j.jagp.2017.06.006. This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Title : When did old age stop being depressing? Depression trajectories of older Americans and Britons 2002-2012. Author : Gindo Tampubolon1 and Asri Maharani1 Corresponding author: Gindo Tampubolon Address: Humanities Bridgeford Street Building 2nd floor, Oxford Road, Manchester M13 9PL, United Kingdom Telephone: +44 161 30 66959 Email address: [email protected] Author affiliation and qualification: 1
University of Manchester
Humanities Bridgeford Street Building, Oxford Road, Manchester M13 9PL, United Kingdom Acknowledgement: This work was supported by grants from the Medical Research Council and Economic & Social Research Council (grant number G1001375/1) and the European Union’s Horizon 2020 research and innovation programme (grant number 668648) to Gindo Tampubolon. Asri Maharani was supported by SENSE-Cog project. This project has received funding from the European Union’s Horizon 2020 research and innovation programme under grant agreement No 668648. Keywords: depression, ageing, cohort effect, CES-D, English Longitudinal Study of Ageing, Health and Retirement Survey, linear mixed model
1 Page 1 of 19
Highlights
When did old age stop being depressing? Depression trajectories of older Americans and Britons 2002-2012. We investigated the implications of heterogeneous cohort composition on depression trajectories of older adults in the United States and England, using growth curve model to draw the trajectories over a decade. The findings showed for the first time that the trajectories of depression in later life are radically different for different cohorts living at the same time in the early 21st century. The trajectories of depression of older cohorts, particularly those of the pre-war cohorts in both countries and the war cohort in England, followed a U-shape. Differently, the trajectories of depression of younger cohort, particularly those of the post-war cohorts in both countries and the war cohort in the U.S., took an inverted U-shape.
Abstract OBJECTIVE This study aims to investigate the implications of the heterogeneous cohort composition on depression trajectories of older adults in the United States and England. DESIGN Growth curve models were employed to identify depressive symptom trajectories. SETTING Data spanning six waves over ten years (2002-2012) were drawn from the U.S. Health Retirement Study (HRS) and the English Longitudinal Study of Ageing (ELSA). PARTICIPANTS Community-dwelling Americans and Britons aged 50 years and older. MEASUREMENT Depressive symptoms were measured using the eight-item Center for Epidemiologic Studies Depression Scale (CES-D). RESULTS The sample included 11,919 respondents (7,095 [59.53%] women) in the U.S. and 10,606 respondents (5,802 [54.7%] women) in England aged 50 and older. Older cohorts were shown to have higher depressive symptoms than younger cohorts in the U.S. and England. The trajectories of depression of older cohorts, particularly those of the pre-war cohorts in both countries and the war cohort in England, followed a U-shape. Differently, the trajectories of depression of younger cohort, particularly those of the post-war cohorts in both countries and the war cohort in the U.S., took an inverted U-shape. CONCLUSIONS The trajectories of depression in later life between cohorts took different shapes. This finding may lead to the development of more cost-effective policies for treating depression in later life.
2 Page 2 of 19
1 Introduction Population aging changes the composition of cohorts living in old age with implications that have largely been missed. Population aging is indicated by a larger share of people aged beyond the notional threshold of old age (Bismarck’s 65 years) and by an expanding period of expectancy to enjoy this later life.1,2 These two facts change the cohort composition of older people today. In the early 21st century, sizable numbers of pre-war cohort members, war cohort members, and baby boomers intermingle. Members of each cohort can expect to live longer than their predecessors half a century before. These cohorts, with their compositions, sizes and characteristics, can be expected to remain in society for some time yet due to their increasing life expectancy. Their composition matters for policy in a particular way and for our understanding of aging today. Recently, reports have documented differential aging trajectories for different cohorts, particularly the trajectories of frailty,3 cognition4 and well-being.5 Any uniform social policy on aging may therefore be appropriate for some cohorts but not for others. Basing policy on the experiences of the largest cohort or of a synthetic one, to the unintended neglect of others, may be inappropriate not only because the size of the resulting mismatch but also because of the duration of its effect. After all, those neglected cohorts, with needs mismatched, are expected to live ever longer. Depression is common among older adults6-10 and its consequences are particularly severe. Later life depression may lead to higher suicide,11 disability,12 morbidity and mortality rates7,8,13 among older people. Preventing depression is thus an important aspect of healthy aging to mitigate the challenges associated with globally aging populations.14 Cross-sectional studies have documented a significant nonlinear association between age and depression.15-17 Depression symptoms across life exhibit a U shape: they are relatively widespread in early adulthood, decline during middle age and rise again during old age. A more recent ten-year longitudinal study reported a consistent U-shaped association between age and depression among Canadians aged 65 and older after including risk factors.18 However, a longitudinal study using a community sample of U.S. respondents aged 65 to 95 found that the association between age and higher levels of depression diminished after taking into account the risk factors included in the analysis.19
3 Page 3 of 19
In addition to documenting different trajectories of depression, the literature has also noted the well-being paradox of old age, according to which midlife adults’ well-being is lower than in earlier and later life.20,21 Though based mostly on recent cross-sectional data, this hints at the possibility that the post-war cohort may not follow their predecessors in experiencing a depressing old age. Suppose that both the pre-war and war cohorts followed the synthetic trajectories of ‘old age is depressing’,15-18 whereas the post-war cohort were more satisfied with their lot. If social policy were to consist of offering talking therapies to improve older people’s mental well-being, then a synthetic trajectory-based policy would be insensitive and expensive. Our purpose in this study was to investigate the implications of the heterogeneous cohort composition on depression trajectories from middle age (≥50 years) onwards. We took advantage of the comparability built into the design of the U.S. Health and Retirement Study (HRS) and the English Longitudinal Study of Ageing (ELSA) to answer a series of research questions: 1. Are different cohorts in America following different trajectories of depression? 2. Are the same cohorts in Britain following comparable trajectories? 3. What are the individual factors driving the different trajectories? 4. Are there similarities and dissimilarities in those factors between the two countries? This study showed for the first time that the trajectories of depression in later life are radically different for different cohorts living at the same time in the early 21st century: the U shape intermingles with the inverted U-shape. It also revealed the period when one shape transitioned to the other, perhaps marking the unique role of the war in both the United States and England. Because the war cohorts followed different trajectories in the two countries, a new finding uncovered here, it offers rather different policy implications for maintaining the well-being and health of older people. 2 Methods 2.1 Data sources To identify within-person change in depression, we fit separate growth curve analysis to longitudinal data from respondents in the U.S. Health and Retirement Study (HRS; six biannual waves from 2002 to 2012)22 and English Longitudinal Study of Ageing (ELSA; six biannual waves from 2002 to 2012).23 HRS data was harmonized with ELSA data; we chose 4 Page 4 of 19
those waves of the surveys as they had identical time frames and included the same measure of depression. The HRS data were collected using a combination of in-person and phone interviews, while all of the ELSA interviews were performed in person. HRS collected information on the demographic, socioeconomic, and health circumstances of noninstitutionalized adults aged 51 years and older in the U.S., while ELSA collected the same information from non-institutionalized adults aged 50 years and older in England. To maximize the comparability of the U.S. and English sample, we included Caucasian respondents only as less than 5 per mil of the ELSA sample were non-Caucasian. Our final sample comprised 11,919 respondents in the U.S. and 10,606 respondents in England. Appendix 1 shows the frequency of samples based on age groups and cohort to show the sufficient overlap in the ages of the samples for each cohort in each data. 2.2 Study measures 2.2.1 Depressive symptoms The presence of depressive symptoms was assessed using the eight-item version of the Center of Epidemiologic Studies Depression Scale (CES-D).24 The CES-D has gained general acceptance as a tool to screen for depressive symptoms in older people24-26 and has been widely used in studies of late-life depression.27-29 Respondents responded to the eight questions regarding their feelings during the past week by answering ‘yes’ or ‘no’ to the following items: 1) you felt depressed; 2) you felt that everything you did was an effort; 3) your sleep was restless; 4) you were happy; 5) you felt lonely; 6) you enjoyed life; 7) you felt sad; and 8) you could not ‘get going’. A depressive symptom score was assigned by totaling all item scores after reversing the questions of the positive mood (Questions 4 and 6). The possible range of scores was from 0 to 8. 2.2.2 Birth cohorts To ensure their comparability, we grouped the sample into three birth cohorts based on sociohistorical events having occurred in both the U.S. and England. The oldest cohort members (pre-war cohort) were born before the start of World War II in 1938, while the second cohort members (war cohort) were born during the war. The most recent cohort members (post-war cohort) were born in or after 1946. Although England came into the War
5 Page 5 of 19
in 1939 while the U.S. in 1941 we opted to ensure the same age range between the two data sets. 2.2.3 Covariates A growing body of literature has investigated the relationships between demographic factors (e.g. gender, education and socioeconomic status) and depression among older adults.29-31 The demographic variables used in this study were measured as follows. Age was determined by self-reported date of birth. Gender was included in the analysis with male as the reference. We categorized the levels of education completed by respondents into: less than high school (reference), high school, and some college. Marital status was measured using a four-group categorization based on individuals’ current and previous relationships: single, married or cohabiting, divorced, and widowed. We used tertiles of the aggregate of public pension wealth and private pension wealth to measure wealth,4 using the poorest tertile as the reference. The literature has reported associations between behavioral risk factors and depression.30,32 In this study, we considered smoking status using the categories current smoker, former smoker, and non-smoker (reference). In addition to smoking status, we included drinking behavior (drinking more than five days a week). Depression is known to be affected by physical and cognitive functions. Physical function in this study was measured by a summary score of (instrumental) activities of daily living, mobility, and muscle function. We assessed cognitive ability using tests of immediate and delayed recall of ten words each. As chronic conditions may be implicated in depression,28,30 it is advisable to also include those conditions in the analysis. Chronic conditions included in this study were hypertension, heart condition, diabetes, stroke, arthritis and cancer. 2.3 Statistical analysis We evaluated trajectories of depressive symptoms over twelve years using growth curve model; which is also known as random coefficients or mixed model. This model is a specific type of multilevel model in which the lowest level of observation is repeated over time using various measures that are then treated as nested within the individuals.33,34 This method is widely used in ageing studies using longitudinal data.4,5,19 This model provides information on the effects of each determinant, including random and fixed effects, adjusting for all risk 6 Page 6 of 19
factors in the model. The growth curve analysis was carried out using four sequence of models. We began with a null model in which only age and squared age were included to identify the shapes or forms of the growth model. Next, we created a baseline model including age, squared age, gender, marital status, socioeconomic determinants, lifestyle factors (smoking status, drinking behavior, and physical exercise) and the presence of chronic diseases. Cohort indicators were included in the third cohort model. Finally, cohort indicators and the interaction between cohorts, age and squared age were included in the fourth or final model. Akaike information criteria (AIC) across the four models suggested the final model has the best fit. We then generated the predicted trajectories of depressive symptoms (controlled for all covariates in the final model) with separate curves for each cohort in the U.S and England. All statistical analyses were conducted using Stata 14.0. 3 Results (Table 1 is around here) A total of 11,919 participants in the U.S. and 10,606 participants in England aged 50 years and over were initially included in this study. Table 1 shows the baseline characteristics of the American and English respondents in 2002-2003. It shows that older adults from the U.S. were, on average, less depressed than their English counterparts. There were a slightly higher proportion of male respondents in the English sample. The proportion of older adults who completed high school and/or attended college was higher in the U.S. than in England. A higher percentage of older adults in the U.S. suffered from every chronic condition included in this study. However, English adults reported less physical problems and more cognitive impairment. The percentages of English older adults who were smokers were a bit higher than those of American older adults, while the percentage of English older adults who drink regularly was slightly lower than that of American older adults. (Table 2 is around here) Table 2 displays the results from the baseline and cohort growth curve models for depressive symptoms in the U.S. and England separately. The null model for depression includes curvilinear age effects only. It showed that trajectories in later life have curvilinear or quadratic shapes both in the U.S. and England. The curvilinear shape of the trajectories of depression in later life remained in the second and final models, where behavioral risk 7 Page 7 of 19
factors, chronic conditions, and birth cohorts were included. Cohort had different effects on respondents in the two countries. Model 3 presents the results for the association of birth cohort and depression among older adults. War cohort respondents were less depressed than pre-war cohort respondents in both the U.S. and England. Respondents who were born after the war were less depressed than those who were born before the war in England, but the significance of that association did not appear in the U.S. Interpreting the coefficients of age, squared age, cohort indicators, and interaction between them in the final model should be done simultaneously. Therefore we predict the trajectories of each cohort and its interactions based on the final model (see Figure 1) to better interpret the relationships between birth cohorts and depression. (Figure 1 is around here) The predicted trajectories of depressive symptoms (based on the final model) for the three cohorts are given in Figure 1. Those estimated trajectories represent the quadratic effects within birth cohort and country. These plots show for the first time the cohort effects in depression in the U.S. (Figure 1A) and England (Figure 1B). It shows that the pre-war cohort had higher levels of depressive symptoms than the other cohorts in both the U.S. and England. In both countries, the level of depressive symptoms of the pre-war cohort increased as the respondents aged, following the “U”-shaped hypothesis. Differently, the level of depressive symptoms of the post-war cohort slightly increased in the decade after reaching the age of 50 and then decreased after the respondents reached the age of 60 in the U.S. (Figure 1A) and decreased after the respondents reached the age of 50 in England (Figure 1B). A puzzling result is shown in the war cohort. In the U.S., the war cohort anticipated the depressive symptoms trajectories of the post-war cohort, while the war cohort in England followed the trajectories of the pre-war cohort. Several potential confounders and socio-demographic characteristics show similar significant associations with depression among older adults in the U.S. and England. In both countries, females had higher levels of depressive symptoms than males. Single, divorced, and widowed respondents had higher depressive symptoms than married respondents. Higher educational attainment and better economic status were correlated with lower levels of depressive symptoms. Current and past smokers had more depressive symptoms than non-smokers. Respondents with better cognitive function had fewer depressive symptoms. The analysis on the ELSA data showed that drinking alcohol more than five days per week was associated 8 Page 8 of 19
with higher levels of depressive symptoms. Differently, the association between drinking alcohol more than five days a week and the level of depressive symptoms was not significant in the analysis on HRS data. 4 Discussion Using repeated assessment of CES-D over twelve years, we modeled the trajectories of depressive symptoms among older adults in the U.S. and England. Three major findings emerge from our study, extending the literature. First, older cohorts had higher levels of depressive symptoms than younger cohorts after adjusting for education and risk factors. Second, the distinguishing mark of later life today is heterogeneity of its trajectories, reflecting its heterogeneous cohort composition. For the pre-war cohort, depressive symptoms are associated with older age. This finding supports previous studies, which found that depressive symptoms follow a U-shaped trajectory.15-17 However, for the majority (the baby boomers), old age is not depressing, a finding that follows the inverted U-shaped trajectory of depression symptoms.20,21 Nevertheless, old age has not entirely ceased to be depressing. The striking finding is that trajectories of depression in old age are now heterogeneous. Finally, to answer the question of when old age ceased to be depressing, we need to focus on the war cohort in the two countries. In the U.S., the war cohort pattern presaged that of the baby boomers. The war cohort took a new path on which the baby boomers soon followed. In England, the war cohort stayed on the old path that the baby boomers soon abandoned. The different trajectories of the war cohort in two different countries may be explained by different effects of war in those countries. Previous studies have shown the significant effect of World War II as a life event on depressive symptoms.35 It is instructive to compare children and adults from the Midwest America and the Midlands England. The theater of war in Europe involved adults from both the Midwest and the Midlands: both were members of the pre-war cohort, and both have displayed similar trajectories of depression. But the children were different. The children from the Midlands were more directly affected by the war, and the war children of the Midlands followed the shape of trajectories earlier traced by the pre-war cohort. But the war children of the Midwest reshaped the trajectories to be followed by the post-war cohort.
9 Page 9 of 19
This contrast opens up a new way of understanding various trajectories of health and wellbeing in later life in these countries. The trajectories may be so heterogeneous that they move beyond what can be characterized as better or poorer. The shapes are fundamentally different: U-shaped versus inverted U-shaped. Moreover, explaining this difference in terms of war experience, although not a complete explanation, may be a fruitful one. We hypothesize that trajectories of health and well-being in European countries can be arrayed along a continuum marked by the patterns observed in England at one end and by those in the U.S. at another. We hypothesize, for instance, that the experiences of German and French people will reveal late-life trajectories more similar to those of the English, whereas the Swedish war experience will render its trajectories of elderly well-being more akin to the American ones. The strengths of our study include the wide range of potential covariates included in the analysis. Depression symptoms tend to be more severe among women, respondents who are not in a marriage or relationship, respondents with less education, and less wealthy respondents. These results provide additional support for the well-documented link between demographic factors and depression. A previous study using panel data from eleven European countries showed the significant relationship between physical activity and lower levels of depressive symptoms.36 Our finding that respondents with physical problems had higher levels of depressive symptoms in both countries supports previous research citing the fact that physical problems may restrict respondents’ ability to perform physical activities.37 Despite known associations between depression and chronic diseases, we found that, in both countries, heart diseases, stroke, arthritis, and cancer were significantly associated with higher depression symptoms, but that hypertension and diabetes were not. A study using data from 7,240 older women in four areas in the U.S. (Baltimore, Minneapolis, the Monongahela Valley, and Portland, Oregon) found that persistently high and increasing depressive symptoms were found in respondents with obesity, diabetes, and myocardial infarction.30 Patients with chronic diseases are likely to face numerous threats and challenges, including pain, disability, changes in lifestyle, and death. Facing such conditions may result in a variety of responses, including depression. This study is subject to some limitations. Firstly, the measure of depression used in this study (the CES-D) is a subjective measure, not a clinical diagnostic measure. However, the CES-D has been widely used and is well accepted for the modeling of depressive symptoms among older adults.24,27-29 Secondly, our results may be affected by attrition. The literature has 10 Page 10 of 19
highlighted that the attrition in longitudinal studies in the field of aging is problematic, given that individuals who drop out of the study are more likely to be frail, less healthy, older, and poorer.38,39 As those dropouts cannot be assumed to be ‘missing at random’, subsequent studies should take attrition in the data into account to avoid bias in analysis. Thirdly, there is only a limited overlap in age of assessments across the three cohorts as only the ages 61-65 are covered in all three birth cohorts. The differences across the cohorts in the shape of the depressive symptoms trajectories may thus due to the constant variation across age. Given the ongoing nature of both HRS and ELSA, further study can rectify this limitation more firmly. Finally, this study pooled males and females together in the analysis. Literature has shown that males and females had different pattern of suicide rates.40-42 Further research can fruitfully explored the interaction between gender, cohort and depression. Our study has theoretical and practical implications. In the theoretical and empirical literature, there is an unsettled debate regarding trajectories of health functioning in later life, whether they are largely marked by decline or essentially linear once comorbidities are acknowledged. The literature has begun to document improvement or deterioration across the cohort. But no study has reported that later life today brings improvements for one cohort and deterioration for another. Researchers will now be called upon to recognize the radical and heterogeneous trajectories of later life; they are radical in that not only do the levels differ, but the directions flip from upward to downward within a society. Failure to appreciate this cohort effect is bound to lead to misleading inference. Beyond the theoretical implications, this cohort effect has policy implications, leading to the consideration of more cost-effective policies. Treating depression requires considerable resources. For example, cognitive behavioral therapy, as one of the effective treatments for depression, is complex and depends on costly and highly trained professionals. 43 For a sizeable proportion of the older population, i.e. baby boomers, therapy may not be needed or can at least, perhaps, be postponed. Only a reduced number of older people need the therapy now. In short, more refined targeting (cohort vs. period) may be both possible and easy to administer.
11 Page 11 of 19
References 1. Myles J: Back to Bismarck? The public policy implications of living longer, Canadian Journal on Aging 2002, 21:325-329. 2. United
Nations:
World
Population
Ageing
2015,
New
York,
URL
http://esa.un.org/undp/wup/Highlights/WUP2014-Highlights.pdf. 3. Marshall A, Jivraj S, Nazroo J, Tampubolon G, Vanhoutte B: Does the level of wealth inequality within an area influence the prevalence of depression amongst older people? Health & Place 2014, 27:194-204. 4. Tampubolon G: Cognitive ageing in Great Britain in the new century: Cohort differences in episodic memory, PLOS ONE 2015, 10:e0144907. 5. Tampubolon G: Growing up in poverty, growing old in infirmity: The long arm of childhood conditions in Great Britain, PLOS ONE 2015, 10:e0144722. 6. Snowdon J: Is depression more prevalent in old age? Australian and New Zealand Journal of Psychiatry 2001, 35:782-787. 7. Blazer DJ: Depression in late life: Review and commentary, Journal of Gerontology 2003, 58A:249-265. 8. Alexopoulos GS: Depression in the elderly, The Lancet 2005, 365:1961-1970. 9. Büchtemann D, Luppa M, Bramesfeld A, Riedel-Heller S: Incidence of late-life depression: A systematic review, Journal of Affective Disorders 2012, 142:172-179. 10. Khalaila R: Depression statuses and related predictors in later life: A 10-year followup study in Israel, European Journal of Ageing 2016, 13:311-321. 11. Conwell Y: Suicide in later life: A review and recommendations for prevention, Suicide and Life-Threatening Behavior 2001, 31:32-47. 12. Beekman ATF, Deeg DJH, Braam AW, Smit JH, Van Tilburg W: Consequences of major and minor depression in later life: a study of disability, well-being and service utilization, Psychological Medicine 1997, 27:1397-1409. 13. Dines P, Hu W, Sajatovic M: Depression in later-life: An overview of assessment and management, Psychiatria Danubina 2014, 26:78-84. 14. Simons LA, McCallum J, Friedlander Y, Simons J: Healthy ageing is associated with reduced and delayed disability, Age and Ageing 2000, 29:143-148. 15. Newman JP: Aging and depression, Psychology and Aging 1989, 4:150-165.
12 Page 12 of 19
16. Kessler RC, Foster C, Webster PS, House JS: The relationship between age and depressive symptoms in two national surveys, Psychology and Aging 1992, 7:119126. 17. Mirowsky J, Ross CE: Age and depression, Journal of Health and Social Behavior 1992, 33:187-205. 18. Wu Z, Schimmele CM, Chappell NL: Aging and late-life depression, Journal of Aging and Health 2012, 24:3-28. 19. Yang Y: Is old age depressing? Growth trajectories and cohort variations in late-life depression, Journal of Health and Social Behavior 2007, 48:16-32. 20. Hansen T, Slagsvold B: The age and subjective well-being paradox revisited: A multidimensional perspective, Norsk Epidemiologi 2012, 22:187-195. 21. Swift HJ, Vauclair CM, Abrams D, Bratt C, Marques S, Lima ML, Revisiting the paradox of well-being: The importance of national context, Journal of Gerontology 2008, 31:1-10. 22. Sonnega A, Weir DR: The health and retirement study: A public data resource for research on aging, Journal of Open Health Data 2014, 2:1-4. 23. Marmot M, Oldfield Z, Clemens S, Blake M, Phelps A, Nazroo J, Steptoe A, Rogers N, Banks J: English Longitudinal Study of Ageing: Waves 0-6, 1998-2013. [data collection].
23rd
edition,
UK
Data
Service.
SN:
5050
(2015).
URL
http://dx.doi.org/10.5255/UKDA-SN-5050-10. 24. Karim J, Weisz R, Bibi Z, ur Rehman S: Validation of the Eight-Item Center for Epidemiologic Studies Depression Scale (CES-D) among older adults, Current Psychology 2015, 34:681-692. 25. Andresen EM, Carter WB, Malmgren JA, Patrick DL: Screening for depression in well older adults: Evaluation of a short form of the CES-D, American Journal of Preventive Medicine 1994, 10:77-84. 26. Irwin M, Artin KH, Oxman MN: Screening for depression in the older adults: Criterion validity of the 10-item Center for Epidemiological Studies Depression Scale (CES-D), JAMA Internal Medicine 1999, 159:1701-1704. 27. Penninx BWJH, Guralnik JM, Ferrucci L, Simonsick EM, Deeg DJH, Wallace RB: Depressive symptoms and physical decline in community-dwelling older persons, JAMA 1998, 279:1720-1726.
13 Page 13 of 19
28. Zivin K, Llewellyn DJ, Lang IA, Vijan S, Kabeto MU, Miller EM, Langa KM: Depression among older adults in the United States and England, The American Journal of Geriatric Psychiatry 2010, 18:1036-1044. 29. Kuchibhatla MN, Fillenbaum GG, Hybels CF, Blazer DG: Trajectory classes of depressive symptoms in a community sample of older adults, Acta Psychiatrica Scandinavica 2012, 125:492-501. 30. Byers AL, Vittinghoff E, Lui LY, Hoang T, Blazer DG, Covinsky KE, Ensrud KE, Cauley JA, Hillier TA, Fredman L, Yaffe K: Twenty-year depressive trajectories among older women, JAMA Psychiatry 2012, 69:1073-1079. 31. Sutin AR, Terracciano A, Milaneschi Y, An Y, Ferrucci L, Zonderman AB: The trajectory of depressive symptoms across the adult life span, JAMA Psychiatry 2013, 70:803-811. 32. van Milligen BAL, Lamers F, Smit JH, Penninx BWJH: Six-year trajectory of objective physical function in persons with depressive and anxiety disorders, Depression and Anxiety 2016: 1-10. 33. Laird NM, Ware JH: Random effects models for longitudinal data, Biometrics 1982, 38 (4): 963-974. 34. Raudenbush SW, Chan W: Growth curve analysis in accelerated longitudinal designs, Journal of Research in Crime and Delinquency 1992, 29 (4): 387-411. 35. Beekman ATF, de Beurs E, van Balkom AJLM, Deeg DJH, van Dyck R, van Tilburg W: Anxiety and depression in later life: Co-occurence and communality of risk factors, The American Journal of Psychiatry 2000, 157:89-95. 36. Lindwall M, Larsman P, Hagger MS: The reciprocal relationship between physical activity and depression in older European adults: A prospective cross-lagged panel design using SHARE data, Health Psychology 2011, 30:453-462. 37. Prohaska T, Belansky E, Belza B, Buchner D, Marshall V, McTigue K, Satariano W, Wilcox S: Physical activity, public health, and aging: Critical issues and research priorities, Journal of Gerontology 2006, 61B (5): S267-S273. 38. Chatfield MD, Brayne CE, Matthews FE: A systematic literature review of attrition between waves in longitudinal studies in the elderly shows a consistent pattern of dropout between differing studies, Journal for Clinical Epidemiology 2005, 58:13-19. 39. Matthews FE, Chatfield M, Freeman C, McCracken C, Brayne C, MRC CFAS: Attrition and bias in the MRC cognitive function and ageing study: An epidemiological investigation, BMC Public Health 2004, 4:1-10. 14 Page 14 of 19
40. Phillips JA, Robin AV, Nugent CN, Idler EL: Understanding recent changes in suicide rates among the middle-aged: Period or cohort effects?, Public Health Reports 2010, 125: 680-688. 41. Phillips JA, Nugent CN: Suicide and the Great Recession of 2007-2009: The role of economic factors in the 50 U.S. states, Social Science & Medicine 2014, 116: 22-31. 42. Phillips JA: A changing epidemiology of suicide? the influence of birth cohorts on suicide rates in the United States, Social Science & Medicine 2014, 114: 151-160. 43. Richards DA, Ekers D, McMillan D, Taylor RS, Byford S, Warren FC, B. Barrett B, Farrand PA, Gilbody S, Kuyken W, Mahen HO, Watkins ER, Wright KA, Hollon SD, Reed N, Rhodes S, Fletcher E, Finning K: Cost and Outcome of Behavioural Activation versus Cognitive Behavioural Therapy for Depression (COBRA): a randomised, controlled, non-inferiority trial, The Lancet 2016, 388:871-880.
Figure legend Figure 1 Predicted trajectories and 95% confidence intervals (CIs) of depressive symptoms among older adults in (A) the United States and (B) England by cohorts. Source: HRS and ELSA 2002-2012.
15 Page 15 of 19
Table 1 Baseline characteristics of the HRS Wave 6 and ELSA Wave 1 respondents, 20022003
CES-D score Age Male Cohort Pre-war War Post-war Marital status Single Married Divorced Widowed Education < High school High school College Drink regularly Yes No Smoking status Never Former Current Physical function Cognitive function Low memory Medium memory High memory Hypertension Heart condition Diabetes Stroke Arthritis Cancer
US (N=11,919) England (N=10,606) 1.37 (1.87) 1.57 (1.97) 68.23 (9.13) 64.81 (9.96) 40.47% 45.3% 66.78% 26.81% 6.42%
51.86% 27.07% 21.07%
2.31% 68.66% 8.99% 20.04%
5.03% 69.7% 8.61% 16.66%
21.04% 35.2% 43.76%
57.84% 14.56% 27.61%
90.87% 9.13%
89.1% 10.9%
41.1% 45.28% 13.62% 2.63 (3.17)
34.95% 46.94% 18.11% 2.21 (3.21)
62.79% 25.74% 11.47% 47.73% 23.35% 13.80% 7.17% 56.63% 13.98%
70.78% 22.54% 6.69% 36.52% 17.11% 6.84% 3.95% 31.5% 5.99%
16 Page 16 of 19
Table 2 Determinants of depressive symptoms among older adults in the US and England US Model 1 Age -0.133 (0.01) ‡ 2 Age 0.102 (0.007) ‡ Cohort, ref: Pre-war War Post-war War*age Post-war*age War*age2 Post-war*age2 Female Marital status, ref: Married Single Divorced Widowed Education, ref:
Model 2
England Model 3
-0.047 (0.01) ‡ -0.046 (0.011) ‡ 0.02 (0.007) ‡ 0.017 (0.007) †
Model 4
Model 1
Model 2
Model 3
Model 4
0.029 (0.023) -0.192 (0.013) ‡ -0.097 (0.013) ‡ -0.03 (0.015) † 0.15 (0.009) ‡ 0.057 (0.009) ‡
-0.108 (0.013) ‡ 0.062 (0.009) ‡
-0.07 (0.039) 0.037 (0.026)
-0.109 (0.034) ‡ -0.167 (0.042) ‡
0.158 (0.022) ‡ 0.159 (0.022) ‡
4.855 (2.277) † -3.017 (3.285) -0.131 (0.069) 0.128 (0.111) 0.000 (0.000) -0.001 (0.000) 0.159 (0.022) ‡
0.269 (0.025) ‡
0.27 (0.025) ‡
2.036 (2.871) -3.878 (3.189) -0.056 (0.086) 0.142 (0.105) 0.036 (0.065) -0.132 (0.089) 0.27 (0.025) ‡
0.308 (0.061) ‡ 0.305 (0.061) ‡ 0.493 (0.03) ‡ 0.495 (0.03) ‡ 0.551 (0.022) ‡ 0.549 (0.022) ‡
0.304 (0.061) ‡ 0.497 (0.03) ‡ 0.549 (0.022) ‡
0.334 (0.053) ‡ 0.44 (0.038) ‡ 0.68 (0.03) ‡
0.335 (0.053) ‡ 0.445 (0.038) ‡ 0.676 (0.03) ‡
0.335 (0.053) ‡ 0.446 (0.038) ‡ 0.676 (0.03) ‡
-0.345 (0.029) ‡ -0.343 (0.029) ‡ -0.458 (0.029) ‡ -0.453 (0.029) ‡
-0.343 (0.029) ‡ -0.454 (0.029) ‡
-0.143 (0.035) ‡ -0.211 (0.029) ‡
-0.137 (0.035) ‡ -0.202 (0.029) ‡
-0.137 (0.035) ‡ -0.202 (0.029) ‡
-0.106 (0.017) ‡ -0.105 (0.017) ‡ -0.147 (0.021) ‡ -0.146 (0.021) ‡ 0.168 (0.002) ‡ 0.168 (0.002) ‡
-0.105 (0.017) ‡ -0.146 (0.021) ‡ 0.168 (0.002) ‡
-0.16 (0.021) ‡ -0.282 (0.023) ‡ 0.169 (0.003) ‡
-0.161 (0.021) ‡ -0.285 (0.023) ‡ 0.168 (0.003) ‡
-0.161 (0.021) ‡ -0.284 (0.023) ‡ 0.168 (0.003) ‡
-0.082 (0.014) ‡ -0.082 (0.014) ‡ -0.106 (0.022) ‡ -0.105 (0.022) ‡
-0.081 (0.014) ‡ -0.105 (0.022) ‡
-0.136 (0.017) ‡ -0.14 (0.028) ‡
-0.13 (0.017) ‡ -0.129 (0.028) ‡
-0.13 (0.017) ‡ -0.129 (0.028) ‡
-0.15 (0.029) ‡ -0.045 (0.039)
17 Page 17 of 19
Drink regularly Smoking status, ref: Never Former Current Hypertension Heart condition Diabetes Stroke Arthritis Cancer Constant AIC
0.023 (0.024)
0.022 (0.024)
0.022 (0.024)
-0.129 (0.028) ‡
-0.13 (0.028) ‡
-0.13 (0.028) ‡
0.088 (0.022) ‡ 0.238 (0.03) ‡ 0.028 (0.017) 0.169 (0.019) ‡ 0.002 (0.022) 0.071 (0.029) † 0.068 (0.018) ‡ 0.075 (0.022) ‡ 5.582 (0.385) ‡ 3.171 (0.368) ‡
0.089 (0.022) ‡ 0.242 (0.03) ‡ 0.031 (0.017) 0.169 (0.019) ‡ 0.006 (0.022) 0.073 (0.029) † 0.071 (0.018) ‡ 0.076 (0.022) ‡ 3.327 (0.41) ‡
0.088 (0.022) ‡ 0.243 (0.03) ‡ 0.032 (0.017) 0.169 (0.019) ‡ 0.006 (0.022) 0.073 (0.029) † 0.072 (0.018) ‡ 0.076 (0.022) ‡ 0.273 (0.912)
7.584 (0.45) ‡
0.113 (0.025) ‡ 0.304 (0.034) ‡ 0.021 (0.021) 0.161 (0.026) ‡ 0.043 (0.035) 0.096 (0.048) † 0.169 (0.022) ‡ 0.132 (0.035) ‡ 4.829 (0.444) ‡
0.115 (0.025) ‡ 0.303 (0.034) ‡ 0.024 (0.021) 0.162 (0.026) ‡ 0.048 (0.035) 0.097 (0.048) † 0.175 (0.022) ‡ 0.137 (0.035) ‡ 5.429 (0.482) ‡
0.115 (0.025) ‡ 0.303 (0.034) ‡ 0.024 (0.021) 0.162 (0.026) ‡ 0.048 (0.035) 0.097 (0.048) † 0.175 (0.022) ‡ 0.137 (0.035) ‡ 3.955 (1.5) ‡
238822
238806
188490
165726
165714
165717
247071
238846
Note: Sig.: †: significant at 5% or less; ‡: significant at 1% or less. Reported are coefficients (standard errors) from the three growth curve models for each dataset. Significance of these coefficients is determined with a t-statistic. Degrees of freedom ranged from 5 to 31 for both datasets.
18 Page 18 of 19
Appendix 1 Frequency of HRS Wave 6and ELSA Wave 1 respondents, 2002-2003 based on age and cohort Age 50-55 56-60 61-65 66-70 71-75 76-80 81-85 86-90 >90
Pre-war 0 0 1,160 2,157 1,767 1,460 1,045 370 0
HRS War 0 1,544 1,310 0 0 0 0 0 0
Post-war 936 170 0 0 0 0 0 0 00
Pre-war 0 0 781 1,553 1,310 987 658 211 0
ELSA War 0 1,617 797 0 0 0 0 0 0
Post-war 2,431 261 0 0 0 0 0 0 00
19 Page 19 of 19
S1064-7481(17)30350-0 http://dx.doi.org/doi: 10.1016/j.jagp.2017.06.006 AMGP 863
To appear in:
The American Journal of Geriatric Psychiatry
Received date: Revised date: Accepted date:
1-2-2017 2-6-2017 5-6-2017
Please cite this article as: Gindo Tampubolon, Asri Maharani, When Did Old Age Stop Being Depressing? Depression Trajectories of Older Americans and Britons 2002-2012, The American Journal of Geriatric Psychiatry (2017), http://dx.doi.org/doi: 10.1016/j.jagp.2017.06.006. This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Title : When did old age stop being depressing? Depression trajectories of older Americans and Britons 2002-2012. Author : Gindo Tampubolon1 and Asri Maharani1 Corresponding author: Gindo Tampubolon Address: Humanities Bridgeford Street Building 2nd floor, Oxford Road, Manchester M13 9PL, United Kingdom Telephone: +44 161 30 66959 Email address: [email protected] Author affiliation and qualification: 1
University of Manchester
Humanities Bridgeford Street Building, Oxford Road, Manchester M13 9PL, United Kingdom Acknowledgement: This work was supported by grants from the Medical Research Council and Economic & Social Research Council (grant number G1001375/1) and the European Union’s Horizon 2020 research and innovation programme (grant number 668648) to Gindo Tampubolon. Asri Maharani was supported by SENSE-Cog project. This project has received funding from the European Union’s Horizon 2020 research and innovation programme under grant agreement No 668648. Keywords: depression, ageing, cohort effect, CES-D, English Longitudinal Study of Ageing, Health and Retirement Survey, linear mixed model
1 Page 1 of 19
Highlights
When did old age stop being depressing? Depression trajectories of older Americans and Britons 2002-2012. We investigated the implications of heterogeneous cohort composition on depression trajectories of older adults in the United States and England, using growth curve model to draw the trajectories over a decade. The findings showed for the first time that the trajectories of depression in later life are radically different for different cohorts living at the same time in the early 21st century. The trajectories of depression of older cohorts, particularly those of the pre-war cohorts in both countries and the war cohort in England, followed a U-shape. Differently, the trajectories of depression of younger cohort, particularly those of the post-war cohorts in both countries and the war cohort in the U.S., took an inverted U-shape.
Abstract OBJECTIVE This study aims to investigate the implications of the heterogeneous cohort composition on depression trajectories of older adults in the United States and England. DESIGN Growth curve models were employed to identify depressive symptom trajectories. SETTING Data spanning six waves over ten years (2002-2012) were drawn from the U.S. Health Retirement Study (HRS) and the English Longitudinal Study of Ageing (ELSA). PARTICIPANTS Community-dwelling Americans and Britons aged 50 years and older. MEASUREMENT Depressive symptoms were measured using the eight-item Center for Epidemiologic Studies Depression Scale (CES-D). RESULTS The sample included 11,919 respondents (7,095 [59.53%] women) in the U.S. and 10,606 respondents (5,802 [54.7%] women) in England aged 50 and older. Older cohorts were shown to have higher depressive symptoms than younger cohorts in the U.S. and England. The trajectories of depression of older cohorts, particularly those of the pre-war cohorts in both countries and the war cohort in England, followed a U-shape. Differently, the trajectories of depression of younger cohort, particularly those of the post-war cohorts in both countries and the war cohort in the U.S., took an inverted U-shape. CONCLUSIONS The trajectories of depression in later life between cohorts took different shapes. This finding may lead to the development of more cost-effective policies for treating depression in later life.
2 Page 2 of 19
1 Introduction Population aging changes the composition of cohorts living in old age with implications that have largely been missed. Population aging is indicated by a larger share of people aged beyond the notional threshold of old age (Bismarck’s 65 years) and by an expanding period of expectancy to enjoy this later life.1,2 These two facts change the cohort composition of older people today. In the early 21st century, sizable numbers of pre-war cohort members, war cohort members, and baby boomers intermingle. Members of each cohort can expect to live longer than their predecessors half a century before. These cohorts, with their compositions, sizes and characteristics, can be expected to remain in society for some time yet due to their increasing life expectancy. Their composition matters for policy in a particular way and for our understanding of aging today. Recently, reports have documented differential aging trajectories for different cohorts, particularly the trajectories of frailty,3 cognition4 and well-being.5 Any uniform social policy on aging may therefore be appropriate for some cohorts but not for others. Basing policy on the experiences of the largest cohort or of a synthetic one, to the unintended neglect of others, may be inappropriate not only because the size of the resulting mismatch but also because of the duration of its effect. After all, those neglected cohorts, with needs mismatched, are expected to live ever longer. Depression is common among older adults6-10 and its consequences are particularly severe. Later life depression may lead to higher suicide,11 disability,12 morbidity and mortality rates7,8,13 among older people. Preventing depression is thus an important aspect of healthy aging to mitigate the challenges associated with globally aging populations.14 Cross-sectional studies have documented a significant nonlinear association between age and depression.15-17 Depression symptoms across life exhibit a U shape: they are relatively widespread in early adulthood, decline during middle age and rise again during old age. A more recent ten-year longitudinal study reported a consistent U-shaped association between age and depression among Canadians aged 65 and older after including risk factors.18 However, a longitudinal study using a community sample of U.S. respondents aged 65 to 95 found that the association between age and higher levels of depression diminished after taking into account the risk factors included in the analysis.19
3 Page 3 of 19
In addition to documenting different trajectories of depression, the literature has also noted the well-being paradox of old age, according to which midlife adults’ well-being is lower than in earlier and later life.20,21 Though based mostly on recent cross-sectional data, this hints at the possibility that the post-war cohort may not follow their predecessors in experiencing a depressing old age. Suppose that both the pre-war and war cohorts followed the synthetic trajectories of ‘old age is depressing’,15-18 whereas the post-war cohort were more satisfied with their lot. If social policy were to consist of offering talking therapies to improve older people’s mental well-being, then a synthetic trajectory-based policy would be insensitive and expensive. Our purpose in this study was to investigate the implications of the heterogeneous cohort composition on depression trajectories from middle age (≥50 years) onwards. We took advantage of the comparability built into the design of the U.S. Health and Retirement Study (HRS) and the English Longitudinal Study of Ageing (ELSA) to answer a series of research questions: 1. Are different cohorts in America following different trajectories of depression? 2. Are the same cohorts in Britain following comparable trajectories? 3. What are the individual factors driving the different trajectories? 4. Are there similarities and dissimilarities in those factors between the two countries? This study showed for the first time that the trajectories of depression in later life are radically different for different cohorts living at the same time in the early 21st century: the U shape intermingles with the inverted U-shape. It also revealed the period when one shape transitioned to the other, perhaps marking the unique role of the war in both the United States and England. Because the war cohorts followed different trajectories in the two countries, a new finding uncovered here, it offers rather different policy implications for maintaining the well-being and health of older people. 2 Methods 2.1 Data sources To identify within-person change in depression, we fit separate growth curve analysis to longitudinal data from respondents in the U.S. Health and Retirement Study (HRS; six biannual waves from 2002 to 2012)22 and English Longitudinal Study of Ageing (ELSA; six biannual waves from 2002 to 2012).23 HRS data was harmonized with ELSA data; we chose 4 Page 4 of 19
those waves of the surveys as they had identical time frames and included the same measure of depression. The HRS data were collected using a combination of in-person and phone interviews, while all of the ELSA interviews were performed in person. HRS collected information on the demographic, socioeconomic, and health circumstances of noninstitutionalized adults aged 51 years and older in the U.S., while ELSA collected the same information from non-institutionalized adults aged 50 years and older in England. To maximize the comparability of the U.S. and English sample, we included Caucasian respondents only as less than 5 per mil of the ELSA sample were non-Caucasian. Our final sample comprised 11,919 respondents in the U.S. and 10,606 respondents in England. Appendix 1 shows the frequency of samples based on age groups and cohort to show the sufficient overlap in the ages of the samples for each cohort in each data. 2.2 Study measures 2.2.1 Depressive symptoms The presence of depressive symptoms was assessed using the eight-item version of the Center of Epidemiologic Studies Depression Scale (CES-D).24 The CES-D has gained general acceptance as a tool to screen for depressive symptoms in older people24-26 and has been widely used in studies of late-life depression.27-29 Respondents responded to the eight questions regarding their feelings during the past week by answering ‘yes’ or ‘no’ to the following items: 1) you felt depressed; 2) you felt that everything you did was an effort; 3) your sleep was restless; 4) you were happy; 5) you felt lonely; 6) you enjoyed life; 7) you felt sad; and 8) you could not ‘get going’. A depressive symptom score was assigned by totaling all item scores after reversing the questions of the positive mood (Questions 4 and 6). The possible range of scores was from 0 to 8. 2.2.2 Birth cohorts To ensure their comparability, we grouped the sample into three birth cohorts based on sociohistorical events having occurred in both the U.S. and England. The oldest cohort members (pre-war cohort) were born before the start of World War II in 1938, while the second cohort members (war cohort) were born during the war. The most recent cohort members (post-war cohort) were born in or after 1946. Although England came into the War
5 Page 5 of 19
in 1939 while the U.S. in 1941 we opted to ensure the same age range between the two data sets. 2.2.3 Covariates A growing body of literature has investigated the relationships between demographic factors (e.g. gender, education and socioeconomic status) and depression among older adults.29-31 The demographic variables used in this study were measured as follows. Age was determined by self-reported date of birth. Gender was included in the analysis with male as the reference. We categorized the levels of education completed by respondents into: less than high school (reference), high school, and some college. Marital status was measured using a four-group categorization based on individuals’ current and previous relationships: single, married or cohabiting, divorced, and widowed. We used tertiles of the aggregate of public pension wealth and private pension wealth to measure wealth,4 using the poorest tertile as the reference. The literature has reported associations between behavioral risk factors and depression.30,32 In this study, we considered smoking status using the categories current smoker, former smoker, and non-smoker (reference). In addition to smoking status, we included drinking behavior (drinking more than five days a week). Depression is known to be affected by physical and cognitive functions. Physical function in this study was measured by a summary score of (instrumental) activities of daily living, mobility, and muscle function. We assessed cognitive ability using tests of immediate and delayed recall of ten words each. As chronic conditions may be implicated in depression,28,30 it is advisable to also include those conditions in the analysis. Chronic conditions included in this study were hypertension, heart condition, diabetes, stroke, arthritis and cancer. 2.3 Statistical analysis We evaluated trajectories of depressive symptoms over twelve years using growth curve model; which is also known as random coefficients or mixed model. This model is a specific type of multilevel model in which the lowest level of observation is repeated over time using various measures that are then treated as nested within the individuals.33,34 This method is widely used in ageing studies using longitudinal data.4,5,19 This model provides information on the effects of each determinant, including random and fixed effects, adjusting for all risk 6 Page 6 of 19
factors in the model. The growth curve analysis was carried out using four sequence of models. We began with a null model in which only age and squared age were included to identify the shapes or forms of the growth model. Next, we created a baseline model including age, squared age, gender, marital status, socioeconomic determinants, lifestyle factors (smoking status, drinking behavior, and physical exercise) and the presence of chronic diseases. Cohort indicators were included in the third cohort model. Finally, cohort indicators and the interaction between cohorts, age and squared age were included in the fourth or final model. Akaike information criteria (AIC) across the four models suggested the final model has the best fit. We then generated the predicted trajectories of depressive symptoms (controlled for all covariates in the final model) with separate curves for each cohort in the U.S and England. All statistical analyses were conducted using Stata 14.0. 3 Results (Table 1 is around here) A total of 11,919 participants in the U.S. and 10,606 participants in England aged 50 years and over were initially included in this study. Table 1 shows the baseline characteristics of the American and English respondents in 2002-2003. It shows that older adults from the U.S. were, on average, less depressed than their English counterparts. There were a slightly higher proportion of male respondents in the English sample. The proportion of older adults who completed high school and/or attended college was higher in the U.S. than in England. A higher percentage of older adults in the U.S. suffered from every chronic condition included in this study. However, English adults reported less physical problems and more cognitive impairment. The percentages of English older adults who were smokers were a bit higher than those of American older adults, while the percentage of English older adults who drink regularly was slightly lower than that of American older adults. (Table 2 is around here) Table 2 displays the results from the baseline and cohort growth curve models for depressive symptoms in the U.S. and England separately. The null model for depression includes curvilinear age effects only. It showed that trajectories in later life have curvilinear or quadratic shapes both in the U.S. and England. The curvilinear shape of the trajectories of depression in later life remained in the second and final models, where behavioral risk 7 Page 7 of 19
factors, chronic conditions, and birth cohorts were included. Cohort had different effects on respondents in the two countries. Model 3 presents the results for the association of birth cohort and depression among older adults. War cohort respondents were less depressed than pre-war cohort respondents in both the U.S. and England. Respondents who were born after the war were less depressed than those who were born before the war in England, but the significance of that association did not appear in the U.S. Interpreting the coefficients of age, squared age, cohort indicators, and interaction between them in the final model should be done simultaneously. Therefore we predict the trajectories of each cohort and its interactions based on the final model (see Figure 1) to better interpret the relationships between birth cohorts and depression. (Figure 1 is around here) The predicted trajectories of depressive symptoms (based on the final model) for the three cohorts are given in Figure 1. Those estimated trajectories represent the quadratic effects within birth cohort and country. These plots show for the first time the cohort effects in depression in the U.S. (Figure 1A) and England (Figure 1B). It shows that the pre-war cohort had higher levels of depressive symptoms than the other cohorts in both the U.S. and England. In both countries, the level of depressive symptoms of the pre-war cohort increased as the respondents aged, following the “U”-shaped hypothesis. Differently, the level of depressive symptoms of the post-war cohort slightly increased in the decade after reaching the age of 50 and then decreased after the respondents reached the age of 60 in the U.S. (Figure 1A) and decreased after the respondents reached the age of 50 in England (Figure 1B). A puzzling result is shown in the war cohort. In the U.S., the war cohort anticipated the depressive symptoms trajectories of the post-war cohort, while the war cohort in England followed the trajectories of the pre-war cohort. Several potential confounders and socio-demographic characteristics show similar significant associations with depression among older adults in the U.S. and England. In both countries, females had higher levels of depressive symptoms than males. Single, divorced, and widowed respondents had higher depressive symptoms than married respondents. Higher educational attainment and better economic status were correlated with lower levels of depressive symptoms. Current and past smokers had more depressive symptoms than non-smokers. Respondents with better cognitive function had fewer depressive symptoms. The analysis on the ELSA data showed that drinking alcohol more than five days per week was associated 8 Page 8 of 19
with higher levels of depressive symptoms. Differently, the association between drinking alcohol more than five days a week and the level of depressive symptoms was not significant in the analysis on HRS data. 4 Discussion Using repeated assessment of CES-D over twelve years, we modeled the trajectories of depressive symptoms among older adults in the U.S. and England. Three major findings emerge from our study, extending the literature. First, older cohorts had higher levels of depressive symptoms than younger cohorts after adjusting for education and risk factors. Second, the distinguishing mark of later life today is heterogeneity of its trajectories, reflecting its heterogeneous cohort composition. For the pre-war cohort, depressive symptoms are associated with older age. This finding supports previous studies, which found that depressive symptoms follow a U-shaped trajectory.15-17 However, for the majority (the baby boomers), old age is not depressing, a finding that follows the inverted U-shaped trajectory of depression symptoms.20,21 Nevertheless, old age has not entirely ceased to be depressing. The striking finding is that trajectories of depression in old age are now heterogeneous. Finally, to answer the question of when old age ceased to be depressing, we need to focus on the war cohort in the two countries. In the U.S., the war cohort pattern presaged that of the baby boomers. The war cohort took a new path on which the baby boomers soon followed. In England, the war cohort stayed on the old path that the baby boomers soon abandoned. The different trajectories of the war cohort in two different countries may be explained by different effects of war in those countries. Previous studies have shown the significant effect of World War II as a life event on depressive symptoms.35 It is instructive to compare children and adults from the Midwest America and the Midlands England. The theater of war in Europe involved adults from both the Midwest and the Midlands: both were members of the pre-war cohort, and both have displayed similar trajectories of depression. But the children were different. The children from the Midlands were more directly affected by the war, and the war children of the Midlands followed the shape of trajectories earlier traced by the pre-war cohort. But the war children of the Midwest reshaped the trajectories to be followed by the post-war cohort.
9 Page 9 of 19
This contrast opens up a new way of understanding various trajectories of health and wellbeing in later life in these countries. The trajectories may be so heterogeneous that they move beyond what can be characterized as better or poorer. The shapes are fundamentally different: U-shaped versus inverted U-shaped. Moreover, explaining this difference in terms of war experience, although not a complete explanation, may be a fruitful one. We hypothesize that trajectories of health and well-being in European countries can be arrayed along a continuum marked by the patterns observed in England at one end and by those in the U.S. at another. We hypothesize, for instance, that the experiences of German and French people will reveal late-life trajectories more similar to those of the English, whereas the Swedish war experience will render its trajectories of elderly well-being more akin to the American ones. The strengths of our study include the wide range of potential covariates included in the analysis. Depression symptoms tend to be more severe among women, respondents who are not in a marriage or relationship, respondents with less education, and less wealthy respondents. These results provide additional support for the well-documented link between demographic factors and depression. A previous study using panel data from eleven European countries showed the significant relationship between physical activity and lower levels of depressive symptoms.36 Our finding that respondents with physical problems had higher levels of depressive symptoms in both countries supports previous research citing the fact that physical problems may restrict respondents’ ability to perform physical activities.37 Despite known associations between depression and chronic diseases, we found that, in both countries, heart diseases, stroke, arthritis, and cancer were significantly associated with higher depression symptoms, but that hypertension and diabetes were not. A study using data from 7,240 older women in four areas in the U.S. (Baltimore, Minneapolis, the Monongahela Valley, and Portland, Oregon) found that persistently high and increasing depressive symptoms were found in respondents with obesity, diabetes, and myocardial infarction.30 Patients with chronic diseases are likely to face numerous threats and challenges, including pain, disability, changes in lifestyle, and death. Facing such conditions may result in a variety of responses, including depression. This study is subject to some limitations. Firstly, the measure of depression used in this study (the CES-D) is a subjective measure, not a clinical diagnostic measure. However, the CES-D has been widely used and is well accepted for the modeling of depressive symptoms among older adults.24,27-29 Secondly, our results may be affected by attrition. The literature has 10 Page 10 of 19
highlighted that the attrition in longitudinal studies in the field of aging is problematic, given that individuals who drop out of the study are more likely to be frail, less healthy, older, and poorer.38,39 As those dropouts cannot be assumed to be ‘missing at random’, subsequent studies should take attrition in the data into account to avoid bias in analysis. Thirdly, there is only a limited overlap in age of assessments across the three cohorts as only the ages 61-65 are covered in all three birth cohorts. The differences across the cohorts in the shape of the depressive symptoms trajectories may thus due to the constant variation across age. Given the ongoing nature of both HRS and ELSA, further study can rectify this limitation more firmly. Finally, this study pooled males and females together in the analysis. Literature has shown that males and females had different pattern of suicide rates.40-42 Further research can fruitfully explored the interaction between gender, cohort and depression. Our study has theoretical and practical implications. In the theoretical and empirical literature, there is an unsettled debate regarding trajectories of health functioning in later life, whether they are largely marked by decline or essentially linear once comorbidities are acknowledged. The literature has begun to document improvement or deterioration across the cohort. But no study has reported that later life today brings improvements for one cohort and deterioration for another. Researchers will now be called upon to recognize the radical and heterogeneous trajectories of later life; they are radical in that not only do the levels differ, but the directions flip from upward to downward within a society. Failure to appreciate this cohort effect is bound to lead to misleading inference. Beyond the theoretical implications, this cohort effect has policy implications, leading to the consideration of more cost-effective policies. Treating depression requires considerable resources. For example, cognitive behavioral therapy, as one of the effective treatments for depression, is complex and depends on costly and highly trained professionals. 43 For a sizeable proportion of the older population, i.e. baby boomers, therapy may not be needed or can at least, perhaps, be postponed. Only a reduced number of older people need the therapy now. In short, more refined targeting (cohort vs. period) may be both possible and easy to administer.
11 Page 11 of 19
References 1. Myles J: Back to Bismarck? The public policy implications of living longer, Canadian Journal on Aging 2002, 21:325-329. 2. United
Nations:
World
Population
Ageing
2015,
New
York,
URL
http://esa.un.org/undp/wup/Highlights/WUP2014-Highlights.pdf. 3. Marshall A, Jivraj S, Nazroo J, Tampubolon G, Vanhoutte B: Does the level of wealth inequality within an area influence the prevalence of depression amongst older people? Health & Place 2014, 27:194-204. 4. Tampubolon G: Cognitive ageing in Great Britain in the new century: Cohort differences in episodic memory, PLOS ONE 2015, 10:e0144907. 5. Tampubolon G: Growing up in poverty, growing old in infirmity: The long arm of childhood conditions in Great Britain, PLOS ONE 2015, 10:e0144722. 6. Snowdon J: Is depression more prevalent in old age? Australian and New Zealand Journal of Psychiatry 2001, 35:782-787. 7. Blazer DJ: Depression in late life: Review and commentary, Journal of Gerontology 2003, 58A:249-265. 8. Alexopoulos GS: Depression in the elderly, The Lancet 2005, 365:1961-1970. 9. Büchtemann D, Luppa M, Bramesfeld A, Riedel-Heller S: Incidence of late-life depression: A systematic review, Journal of Affective Disorders 2012, 142:172-179. 10. Khalaila R: Depression statuses and related predictors in later life: A 10-year followup study in Israel, European Journal of Ageing 2016, 13:311-321. 11. Conwell Y: Suicide in later life: A review and recommendations for prevention, Suicide and Life-Threatening Behavior 2001, 31:32-47. 12. Beekman ATF, Deeg DJH, Braam AW, Smit JH, Van Tilburg W: Consequences of major and minor depression in later life: a study of disability, well-being and service utilization, Psychological Medicine 1997, 27:1397-1409. 13. Dines P, Hu W, Sajatovic M: Depression in later-life: An overview of assessment and management, Psychiatria Danubina 2014, 26:78-84. 14. Simons LA, McCallum J, Friedlander Y, Simons J: Healthy ageing is associated with reduced and delayed disability, Age and Ageing 2000, 29:143-148. 15. Newman JP: Aging and depression, Psychology and Aging 1989, 4:150-165.
12 Page 12 of 19
16. Kessler RC, Foster C, Webster PS, House JS: The relationship between age and depressive symptoms in two national surveys, Psychology and Aging 1992, 7:119126. 17. Mirowsky J, Ross CE: Age and depression, Journal of Health and Social Behavior 1992, 33:187-205. 18. Wu Z, Schimmele CM, Chappell NL: Aging and late-life depression, Journal of Aging and Health 2012, 24:3-28. 19. Yang Y: Is old age depressing? Growth trajectories and cohort variations in late-life depression, Journal of Health and Social Behavior 2007, 48:16-32. 20. Hansen T, Slagsvold B: The age and subjective well-being paradox revisited: A multidimensional perspective, Norsk Epidemiologi 2012, 22:187-195. 21. Swift HJ, Vauclair CM, Abrams D, Bratt C, Marques S, Lima ML, Revisiting the paradox of well-being: The importance of national context, Journal of Gerontology 2008, 31:1-10. 22. Sonnega A, Weir DR: The health and retirement study: A public data resource for research on aging, Journal of Open Health Data 2014, 2:1-4. 23. Marmot M, Oldfield Z, Clemens S, Blake M, Phelps A, Nazroo J, Steptoe A, Rogers N, Banks J: English Longitudinal Study of Ageing: Waves 0-6, 1998-2013. [data collection].
23rd
edition,
UK
Data
Service.
SN:
5050
(2015).
URL
http://dx.doi.org/10.5255/UKDA-SN-5050-10. 24. Karim J, Weisz R, Bibi Z, ur Rehman S: Validation of the Eight-Item Center for Epidemiologic Studies Depression Scale (CES-D) among older adults, Current Psychology 2015, 34:681-692. 25. Andresen EM, Carter WB, Malmgren JA, Patrick DL: Screening for depression in well older adults: Evaluation of a short form of the CES-D, American Journal of Preventive Medicine 1994, 10:77-84. 26. Irwin M, Artin KH, Oxman MN: Screening for depression in the older adults: Criterion validity of the 10-item Center for Epidemiological Studies Depression Scale (CES-D), JAMA Internal Medicine 1999, 159:1701-1704. 27. Penninx BWJH, Guralnik JM, Ferrucci L, Simonsick EM, Deeg DJH, Wallace RB: Depressive symptoms and physical decline in community-dwelling older persons, JAMA 1998, 279:1720-1726.
13 Page 13 of 19
28. Zivin K, Llewellyn DJ, Lang IA, Vijan S, Kabeto MU, Miller EM, Langa KM: Depression among older adults in the United States and England, The American Journal of Geriatric Psychiatry 2010, 18:1036-1044. 29. Kuchibhatla MN, Fillenbaum GG, Hybels CF, Blazer DG: Trajectory classes of depressive symptoms in a community sample of older adults, Acta Psychiatrica Scandinavica 2012, 125:492-501. 30. Byers AL, Vittinghoff E, Lui LY, Hoang T, Blazer DG, Covinsky KE, Ensrud KE, Cauley JA, Hillier TA, Fredman L, Yaffe K: Twenty-year depressive trajectories among older women, JAMA Psychiatry 2012, 69:1073-1079. 31. Sutin AR, Terracciano A, Milaneschi Y, An Y, Ferrucci L, Zonderman AB: The trajectory of depressive symptoms across the adult life span, JAMA Psychiatry 2013, 70:803-811. 32. van Milligen BAL, Lamers F, Smit JH, Penninx BWJH: Six-year trajectory of objective physical function in persons with depressive and anxiety disorders, Depression and Anxiety 2016: 1-10. 33. Laird NM, Ware JH: Random effects models for longitudinal data, Biometrics 1982, 38 (4): 963-974. 34. Raudenbush SW, Chan W: Growth curve analysis in accelerated longitudinal designs, Journal of Research in Crime and Delinquency 1992, 29 (4): 387-411. 35. Beekman ATF, de Beurs E, van Balkom AJLM, Deeg DJH, van Dyck R, van Tilburg W: Anxiety and depression in later life: Co-occurence and communality of risk factors, The American Journal of Psychiatry 2000, 157:89-95. 36. Lindwall M, Larsman P, Hagger MS: The reciprocal relationship between physical activity and depression in older European adults: A prospective cross-lagged panel design using SHARE data, Health Psychology 2011, 30:453-462. 37. Prohaska T, Belansky E, Belza B, Buchner D, Marshall V, McTigue K, Satariano W, Wilcox S: Physical activity, public health, and aging: Critical issues and research priorities, Journal of Gerontology 2006, 61B (5): S267-S273. 38. Chatfield MD, Brayne CE, Matthews FE: A systematic literature review of attrition between waves in longitudinal studies in the elderly shows a consistent pattern of dropout between differing studies, Journal for Clinical Epidemiology 2005, 58:13-19. 39. Matthews FE, Chatfield M, Freeman C, McCracken C, Brayne C, MRC CFAS: Attrition and bias in the MRC cognitive function and ageing study: An epidemiological investigation, BMC Public Health 2004, 4:1-10. 14 Page 14 of 19
40. Phillips JA, Robin AV, Nugent CN, Idler EL: Understanding recent changes in suicide rates among the middle-aged: Period or cohort effects?, Public Health Reports 2010, 125: 680-688. 41. Phillips JA, Nugent CN: Suicide and the Great Recession of 2007-2009: The role of economic factors in the 50 U.S. states, Social Science & Medicine 2014, 116: 22-31. 42. Phillips JA: A changing epidemiology of suicide? the influence of birth cohorts on suicide rates in the United States, Social Science & Medicine 2014, 114: 151-160. 43. Richards DA, Ekers D, McMillan D, Taylor RS, Byford S, Warren FC, B. Barrett B, Farrand PA, Gilbody S, Kuyken W, Mahen HO, Watkins ER, Wright KA, Hollon SD, Reed N, Rhodes S, Fletcher E, Finning K: Cost and Outcome of Behavioural Activation versus Cognitive Behavioural Therapy for Depression (COBRA): a randomised, controlled, non-inferiority trial, The Lancet 2016, 388:871-880.
Figure legend Figure 1 Predicted trajectories and 95% confidence intervals (CIs) of depressive symptoms among older adults in (A) the United States and (B) England by cohorts. Source: HRS and ELSA 2002-2012.
15 Page 15 of 19
Table 1 Baseline characteristics of the HRS Wave 6 and ELSA Wave 1 respondents, 20022003
CES-D score Age Male Cohort Pre-war War Post-war Marital status Single Married Divorced Widowed Education < High school High school College Drink regularly Yes No Smoking status Never Former Current Physical function Cognitive function Low memory Medium memory High memory Hypertension Heart condition Diabetes Stroke Arthritis Cancer
US (N=11,919) England (N=10,606) 1.37 (1.87) 1.57 (1.97) 68.23 (9.13) 64.81 (9.96) 40.47% 45.3% 66.78% 26.81% 6.42%
51.86% 27.07% 21.07%
2.31% 68.66% 8.99% 20.04%
5.03% 69.7% 8.61% 16.66%
21.04% 35.2% 43.76%
57.84% 14.56% 27.61%
90.87% 9.13%
89.1% 10.9%
41.1% 45.28% 13.62% 2.63 (3.17)
34.95% 46.94% 18.11% 2.21 (3.21)
62.79% 25.74% 11.47% 47.73% 23.35% 13.80% 7.17% 56.63% 13.98%
70.78% 22.54% 6.69% 36.52% 17.11% 6.84% 3.95% 31.5% 5.99%
16 Page 16 of 19
Table 2 Determinants of depressive symptoms among older adults in the US and England US Model 1 Age -0.133 (0.01) ‡ 2 Age 0.102 (0.007) ‡ Cohort, ref: Pre-war War Post-war War*age Post-war*age War*age2 Post-war*age2 Female Marital status, ref: Married Single Divorced Widowed Education, ref:
Model 2
England Model 3
-0.047 (0.01) ‡ -0.046 (0.011) ‡ 0.02 (0.007) ‡ 0.017 (0.007) †
Model 4
Model 1
Model 2
Model 3
Model 4
0.029 (0.023) -0.192 (0.013) ‡ -0.097 (0.013) ‡ -0.03 (0.015) † 0.15 (0.009) ‡ 0.057 (0.009) ‡
-0.108 (0.013) ‡ 0.062 (0.009) ‡
-0.07 (0.039) 0.037 (0.026)
-0.109 (0.034) ‡ -0.167 (0.042) ‡
0.158 (0.022) ‡ 0.159 (0.022) ‡
4.855 (2.277) † -3.017 (3.285) -0.131 (0.069) 0.128 (0.111) 0.000 (0.000) -0.001 (0.000) 0.159 (0.022) ‡
0.269 (0.025) ‡
0.27 (0.025) ‡
2.036 (2.871) -3.878 (3.189) -0.056 (0.086) 0.142 (0.105) 0.036 (0.065) -0.132 (0.089) 0.27 (0.025) ‡
0.308 (0.061) ‡ 0.305 (0.061) ‡ 0.493 (0.03) ‡ 0.495 (0.03) ‡ 0.551 (0.022) ‡ 0.549 (0.022) ‡
0.304 (0.061) ‡ 0.497 (0.03) ‡ 0.549 (0.022) ‡
0.334 (0.053) ‡ 0.44 (0.038) ‡ 0.68 (0.03) ‡
0.335 (0.053) ‡ 0.445 (0.038) ‡ 0.676 (0.03) ‡
0.335 (0.053) ‡ 0.446 (0.038) ‡ 0.676 (0.03) ‡
-0.345 (0.029) ‡ -0.343 (0.029) ‡ -0.458 (0.029) ‡ -0.453 (0.029) ‡
-0.343 (0.029) ‡ -0.454 (0.029) ‡
-0.143 (0.035) ‡ -0.211 (0.029) ‡
-0.137 (0.035) ‡ -0.202 (0.029) ‡
-0.137 (0.035) ‡ -0.202 (0.029) ‡
-0.106 (0.017) ‡ -0.105 (0.017) ‡ -0.147 (0.021) ‡ -0.146 (0.021) ‡ 0.168 (0.002) ‡ 0.168 (0.002) ‡
-0.105 (0.017) ‡ -0.146 (0.021) ‡ 0.168 (0.002) ‡
-0.16 (0.021) ‡ -0.282 (0.023) ‡ 0.169 (0.003) ‡
-0.161 (0.021) ‡ -0.285 (0.023) ‡ 0.168 (0.003) ‡
-0.161 (0.021) ‡ -0.284 (0.023) ‡ 0.168 (0.003) ‡
-0.082 (0.014) ‡ -0.082 (0.014) ‡ -0.106 (0.022) ‡ -0.105 (0.022) ‡
-0.081 (0.014) ‡ -0.105 (0.022) ‡
-0.136 (0.017) ‡ -0.14 (0.028) ‡
-0.13 (0.017) ‡ -0.129 (0.028) ‡
-0.13 (0.017) ‡ -0.129 (0.028) ‡
-0.15 (0.029) ‡ -0.045 (0.039)
17 Page 17 of 19
Drink regularly Smoking status, ref: Never Former Current Hypertension Heart condition Diabetes Stroke Arthritis Cancer Constant AIC
0.023 (0.024)
0.022 (0.024)
0.022 (0.024)
-0.129 (0.028) ‡
-0.13 (0.028) ‡
-0.13 (0.028) ‡
0.088 (0.022) ‡ 0.238 (0.03) ‡ 0.028 (0.017) 0.169 (0.019) ‡ 0.002 (0.022) 0.071 (0.029) † 0.068 (0.018) ‡ 0.075 (0.022) ‡ 5.582 (0.385) ‡ 3.171 (0.368) ‡
0.089 (0.022) ‡ 0.242 (0.03) ‡ 0.031 (0.017) 0.169 (0.019) ‡ 0.006 (0.022) 0.073 (0.029) † 0.071 (0.018) ‡ 0.076 (0.022) ‡ 3.327 (0.41) ‡
0.088 (0.022) ‡ 0.243 (0.03) ‡ 0.032 (0.017) 0.169 (0.019) ‡ 0.006 (0.022) 0.073 (0.029) † 0.072 (0.018) ‡ 0.076 (0.022) ‡ 0.273 (0.912)
7.584 (0.45) ‡
0.113 (0.025) ‡ 0.304 (0.034) ‡ 0.021 (0.021) 0.161 (0.026) ‡ 0.043 (0.035) 0.096 (0.048) † 0.169 (0.022) ‡ 0.132 (0.035) ‡ 4.829 (0.444) ‡
0.115 (0.025) ‡ 0.303 (0.034) ‡ 0.024 (0.021) 0.162 (0.026) ‡ 0.048 (0.035) 0.097 (0.048) † 0.175 (0.022) ‡ 0.137 (0.035) ‡ 5.429 (0.482) ‡
0.115 (0.025) ‡ 0.303 (0.034) ‡ 0.024 (0.021) 0.162 (0.026) ‡ 0.048 (0.035) 0.097 (0.048) † 0.175 (0.022) ‡ 0.137 (0.035) ‡ 3.955 (1.5) ‡
238822
238806
188490
165726
165714
165717
247071
238846
Note: Sig.: †: significant at 5% or less; ‡: significant at 1% or less. Reported are coefficients (standard errors) from the three growth curve models for each dataset. Significance of these coefficients is determined with a t-statistic. Degrees of freedom ranged from 5 to 31 for both datasets.
18 Page 18 of 19
Appendix 1 Frequency of HRS Wave 6and ELSA Wave 1 respondents, 2002-2003 based on age and cohort Age 50-55 56-60 61-65 66-70 71-75 76-80 81-85 86-90 >90
Pre-war 0 0 1,160 2,157 1,767 1,460 1,045 370 0
HRS War 0 1,544 1,310 0 0 0 0 0 0
Post-war 936 170 0 0 0 0 0 0 00
Pre-war 0 0 781 1,553 1,310 987 658 211 0
ELSA War 0 1,617 797 0 0 0 0 0 0
Post-war 2,431 261 0 0 0 0 0 0 00
19 Page 19 of 19