Stability and Change in Activities of Daily Living Among Older Mexican Americans

Transcription

1 Journals of Gerontology: Medical Sciences cite as: J Gerontol A Biol Sci Med Sci, 2016, Vol. 71, No. 6, doi: /gerona/glv172 Advance Access publication October 7, 2015 Research Article Stability and Change in Activities of Daily Living Among Older Mexican Americans Bret T. Howrey, 1 Soham Al Snih, 2 Kyu K. Jana, 1 Mary Kristen Peek, 3 and Kenneth J. Ottenbacher 2 1 Department of Family Medicine, 2 Division of Rehabilitation Sciences, and 3 Department of Preventive Medicine and Community Health, University of Texas Medical Branch, Galveston. Address correspondence to Bret T. Howrey, PhD, Department of Family Medicine, University of Texas Medical Branch, 301 University Blvd, Galveston, TX bthowrey@utmb.edu Received April 21, 2015; Accepted September 7, 2015 Decision Editor: Stephen Kritchevsky, PhD Abstract Background. Longitudinal studies of activities of daily living (ADL) in older adults have identified numerous factors associated with declining ability. Analyses based on population averages may not observe distinct subgroups whose ADL trajectories differ. Methods. We used latent class models to identify subgroups of trajectories in a sample from the Hispanic Established Populations for Epidemiologic Study of the Elderly, a population-based study of noninstitutionalized Mexican Americans aged 65 and older from five Southwestern states (n = 2584). Results. Three distinct trajectories of ADL limitations were identified and characterized as stable, delayed, and rapid ADL increase. Sex (female), diabetes, and arthritis were associated with increased odds of membership in the delayed and rapid groups compared with the stable group. Stroke had a differential magnitude of effect on ADL limitations across the stable (β = 1.11, p <.001), delayed (β = 0.52, p <.001), and rapid groups (β = 0.12, p <.05). Hip fracture was associated with increased limitations in the stable group (β = 1.27, p <.001) but not in the rapid group. Church attendance was associated with fewer limitations in all groups with a larger effect in the stable group (β = 0.87, p <.001) compared with the rapid group (β = 0.10, p <.05). Conclusions. Substantial heterogeneity exists in changes in ADL disability over time among older Mexican Americans. Attempts at maintaining function may benefit from targeting reductions in comorbidities and acute health events associated with disability. Keywords: Disability Heterogeneous transitions Trajectories Hispanic Increases in life expectancy over the last half century have corresponded with increases in life with progressive disability among older adults (1,2). Disability may be characterized as difficulty in performing normal, daily activities due to limitations imposed by health (3). Preventing disability and delaying declines in physical function in older adults can facilitate independence, improve quality of life, and reduce the burden of institutionalization and mortality (4 6). Aging is a dynamic process uniquely affecting each individual. Significant variation in self-reported health and disability within age groups has been reported, even at very old ages (85 years and older) (7). Numerous risk factors associated with disability have been identified including increased age, gender (female), diabetes, heart disease, stroke, heart attack, and hip fracture (8 11). In addition, psychiatric disorders, depression, and impaired cognitive status have been associated with increased ADL limitations (12,13). Education, on the other hand, has been associated with reductions in disability (10,14). Research on the association of social factors with disability is more mixed. In a prospective study of 954 older adults, social activity was associated with a 43% decreased risk of developing ADL limitations and a 31% reduced risk of mobility disability (15). A review from 1999 found that low frequency of social contacts was associated with declines in functional status in three studies but was not ratable in four others (13). A study in Norway found that increases in social activity were associated with decreases in ADL limitations while increased social networks were not (16). A study The Author Published by Oxford University Press on behalf of The Gerontological Society of America. All rights reserved. For permissions, please journals.permissions@oup.com. 780

2 Journals of Gerontology: MEDICAL SCIENCES, 2016, Vol. 71, No of older adults in Taiwan found that leisure activities were positively associated with maintained functional status, whereas reduced activities and decreased social networks were associated with chronic disability (17). Similarly, a Swedish study found evidence of subjective support and reduced disability but only among chronically disabled older adults (18). It has been reported that Hispanic elders have greater rates of ADL disability than non-hispanic Whites (19), whereas others have shown that Hispanics are more likely to report similar ADL restrictions but increased limitations in instrumental ADL compared with non-hispanic Whites (20). Higher rates of physical impairment at older ages among Hispanics are of particular importance as Hispanics represent the largest and fastest growing minority group in the United States (21). A difficulty in the identification of predictors of good functional status over the long-term among Hispanics is a lack of longitudinal data on under-represented minorities with relevant predictors. An additional challenge in identifying factors associated with decline or maintenance of ability is the heterogeneous composition of older adult populations encompassing broad variations in health status, mobility, and social support (7,22). We address part of these challenges by using data from the Hispanic Established Population for the Epidemiological Study of the Elderly, an ongoing 20-year study of older Hispanics, and employ latent class models to identify subgroups (or classes) of participants with distinct trajectories of change in disability over time. In this research, we describe the psychosocial, health, and lifestyle factors associated with maintenance of ADL ability in Mexican-origin elders. We hypothesize that, given the reported heterogeneity of disability among older adults, there will be subgroups of distinct trajectories of disability among older Mexicanorigin adults that differ from the overall group mean due to heterogeneity of characteristics within the sample. We also hypothesize that the effects of both psychosocial and physical health factors on ADL change will vary by trajectory membership due to the expected differences in trajectory group characteristics. Materials and Methods Sample The study sample was drawn from participants in the Hispanic Established Populations for Epidemiologic Study of the Elderly (H-EPESE). The H-EPESE is an ongoing population-based study of 3,050 noninstitutionalized Mexican Americans aged 65 and older at baseline ( ) from five Southwestern states (Texas, California, New Mexico, Colorado, and Arizona). Eight waves of data have been collected ( n = 3,050; n = 2,438; n = 1,981; n = 1,682; n = 1,167; n = 921; n = 659; and n = 444). Respondents were interviewed in Spanish or English. Details regarding the methods have been described elsewhere (23,24). Our sample includes all respondents aged years who had complete data for included variables at baseline regardless of disability status (n = 2,584). Excluded subjects had higher rates of stroke, hip fracture, depression, and cognitive impairment than those of the main analytical sample. In addition, because the presence of disability at initial assessment might predispose a participant to increased limitations over time, we also examined a subset of respondents who had no ADL limitations at baseline (n = 2,230). We refer to this subset as disability-free in our results. All research protocols and informed consents were approved by the University s Institutional Review Board. Dependent Variable Disability status was ascertained through self-report of seven items on a modified version of the Katz ADL scale (25). The ADL scale assessed the degree to which the respondent needed assistance in walking, transferring from a bed to a chair, grooming, dressing, bathing, feeding oneself, and toileting. The items were scored as 0 for no help needed and 1 if they needed help or were unable to perform the activity. The individual item scores were then summed into a single measure of ADL limitations (range 0 7). Covariates Cognition Cognitive function was measured using the Mini-Mental State Examination (MMSE) (26) during each interview. The MMSE is composed of responses to a standard battery of memory and reasoning problems that include orientation, attention, short-term recall, language, and the ability to follow simple instructions. Although the common cutoff for impairment on the MMSE is 24, significant performance differences between Hispanics and non-hispanic Whites have been noted (27) suggesting a modified cut point. The MMSE score was dichotomized to reflect cognitive impairment (MMSE < 22) which reflects the lower educational attainment and distribution of MMSE scores in this population and represents the lower bound of normal cognitive function (12). Health Conditions We included major health conditions associated with functional decline (11 13). Hypertension, heart attack, stroke, diabetes, arthritis, and hip fracture were ascertained through self-report by asking the respondents (1) Has a doctor ever told you that you had... during the baseline interview and (2) Since we last spoke... told by a doctor that you... during follow-up. Answers were coded as yes or no. We also included four weight categories based on body mass index (BMI) (28): underweight (BMI < 18.5), overweight (BMI 25 and BMI < 30), and obese (BMI 30) compared with normal weight (BMI 18.5 and BMI < 25). Psychosocial Factors We included two psychosocial factors in the analyses. We included depressed mood because it is associated with disability (29). Depressive symptoms were assessed through the Center for Epidemiologic Studies Depression Scale (CES-D) (30) which has a range from 0 to 60, with higher scores reflecting increased depressive symptoms. A cut point of 16 was used to denote depressive symptoms. Church attendance was also included as a measure of social integration. Attendance was coded as one or more times per month, yes or no. Other Covariates Covariates included gender (female vs male), marital status (currently married vs not), language of interview (Spanish vs English), nativity (U.S. born vs foreign born), high school education (vs less than high school), and current smoking status (yes vs no). Financial strain was assessed through the respondent s reporting a great deal of difficulty meeting monthly payments and coded yes or no. Statistical Analyses All analyses were performed for the complete cohort as well as for the disability-free subset. Demographic characteristics were examined in total and disability-free subset. Next, latent class mixture

3 782 Journals of Gerontology: MEDICAL SCIENCES, 2016, Vol. 71, No. 6 models were used to develop trajectories of disability over time (31). This approach views the population under study as a heterogeneous mixture within which a finite number of distinct subgroups can be identified. The models were developed using the Stata procedure for trajectory modeling (32). ADL disability was treated as a zeroinflated Poisson distribution because it is a count variable with a high proportion of zero-level responses. Change in ADL limitations was modeled as a function of age. We examined models with one to five trajectories and tested functional forms from intercept only through cubic. In selecting the final models, we first compared Bayesian Information Criterion between models with different numbers and forms of trajectories. The model with lower absolute Bayesian Information Criterion was deemed superior. The average posterior probability of group assignment (.70 being acceptable) and group size (minimum 10% of sample per group being acceptable) was then assessed. We also considered the degree to which age was deterministic of group assignment by examining the probability of group membership by year of age, and we examined the difference in age distribution by group finding no significant differences between groups. This approach was applied separately for the complete sample and the disability-free subset. Final models meeting the selection criteria contained three linear trajectories for both the main sample and the disability-free subsample. The latent class model is divided into two parts that are then modeled simultaneously. The first part used a multinomial logit to determine the odds of group membership. This model included timestable covariates measured at baseline as predictors of trajectory group membership: female gender, Spanish interview, U.S. born, high school education or higher, over weight and obese (compared with normal or low weight), diabetes, and arthritis. Interaction effects were tested for Diabetes*Hypertension and Diabetes*Arthritis. Smoking, nativity, language of interview (Spanish), Diabetes*Hypertension effect, and Diabetes * Arthritis effects were nonsignificant and were dropped from the final multinomial models. The second part of the modeling strategy examined change in ADLs over time and incorporated time-varying covariates measured at each assessment period into the equations for each trajectory: marital status, monthly church attendance, financial strain, hypertension, heart attack, stroke, hip fracture, CES-D > 16, and MMSE < 22. In these models, covariates are included in the equations for each trajectory. In addition, interactions were tested between acute health events: Heart attack*stroke, Heart attack*hip fracture, and Stroke*Hip fracture. These interactions were not significant and were dropped from the analysis. Differences between the trajectory slopes were compared with each other and with a single equation model with Wald tests. A particular concern in analyzing panel data is participant attrition, especially data spanning 20 or more years. Data that are missing at random are generally tolerated well by the maximum likelihood approach resulting in unbiased estimates. In the context of our sample, intermittent missing data are considered missing at random. Data missing from a given measurement period forward (2 or more) is considered lost-to-follow-up and affected 8.3% of the total sample. Mortality was the largest contributor to subject attrition. By the end of the follow-up period, 91% of those aged 75 and older at baseline had died. Because attrition due to death is not ignorable, we used a generalization of the trajectory model that allows the probability of dropout to vary as a function of prior observations (33). This allowed us to calculate trajectory-specific attrition rates due to death and make adjustments to trajectory group membership probabilities. In addition, we performed sensitivity analyses re-estimating the models in two ways. First, we model trajectories of attrition combined with death as part of a joint model. Second, we incorporated death into a logit model specifying any ADL limitation or death as the outcome. The results of these models help assess the relative impact of mortality and attrition on between-group differences. All analyses were performed using Stata13 mp software (StataCorp College Station, TX). Results Baseline characteristics of the total sample and the disability-free subsample (no ADL limitations at baseline) are shown in Table 1. The total sample had a mean age of 73 years (SD 5.9) and was mostly women (58%). The subsample was similar in all characteristics. Average age at each assessment period (baseline to wave 9) was 73, 75, 77, 79, 82, 84, 87, and 89 years with an average attrition of 23% between assessments for the complete sample and 22% between assessments for the disability-free subset. Figure 1 displays estimated ADL limitations from a model based on a single population average and the three distinct trajectory groups (stable, delayed disability, rapid disability) for the total sample (left panel) and the subsample (right panel). The stable group starts at an average ADL score of 0 and shows little increase in limitations until age 82. The delayed disability group starts with a predicted average ADL limitation of 0.23 and shows little increase until age 82 at which point predicted ADL limitations increase to 3. The predicted average number of ADL limitations for the rapid disability group was 1.03 at baseline and increased steadily until age 89. Estimates based on a single population average ranged from 0.26 ADL limitations at baseline to 1.78 limitations at last followup. The mean posterior probabilities for group assignment in the total sample were acceptable for the stable (.84), delayed (.75), and rapid disability (.87) groups. Wald tests comparing slopes showed Table 1. Baseline Characteristics of Total Sample and Subsample With no ADL Limitations From the H-EPESE (1993) Total Sample Mean (SD) or Percent Subsample Mean (SD) or Percent N 2,584 2,380 Age (years) (5.87) (5.7) Female High school Overweight Obese Diabetes Arthritis Married Church: monthly Financial strain Hypertension Heart attack Stroke Hip fracture CES-D > MMSE < Notes: CES-D = Center for Epidemiologic Studies Depression Scale; H- EPESE = Hispanic Established Populations for Epidemiologic Study of the Elderly; MMSE = Mini-Mental State Examination.

4 Journals of Gerontology: MEDICAL SCIENCES, 2016, Vol. 71, No significant differences between the total sample and the stable trajectory (Χ 2 = 2,467.66, p <.001), the delayed trajectory (Χ 2 = , p <.001), and the rapid trajectory groups (Χ 2 = 12.49, p <.001). The right panel of Figure 1 shows estimated ADL limitations for the disability-free subsample. Because these are estimates from the regression equations, values at baseline are not equal to zero. The stable group showed little change until age 89 with a final predicted number of ADL limitations of 0.8. The rapid disability group climbs steadily to four ADL limitations by age 89. The estimates based on a single population average ranged from 0.2 at baseline to 1.8 limitations by age 89. The delayed disability group follows the single population estimates until age 87 and reaches 3.1 ADL limitations by age 89. The plotted confidence intervals suggest that the three trajectory groups are distinct from each other until age 89 when the rapid and delayed groups overlap. The mean posterior probabilities for group assignment in the disability-free subsample were.85 in the stable group,.77 in the delayed group, and.91 in the rapid disability group. Table 2 presents the results of the multinomial logit models predicting odds of group membership (stable ADL group is referent) in the fully adjusted models for the full sample (left panel) and the disability-free subsample (right panel). Women were more than twice as likely to be in the rapid group and 78% more likely to be in the delayed disability group compared with the stable group. Diabetes was also associated with increased odds of membership in the delayed disability group (odds ratio [OR] 1.81, 95% confidence interval [CI] ) and rapid disability group (OR 2.44, 95% CI ) compared with the stable group. Obesity was associated with increased odds of membership in the delayed group (OR 2.74, 95% CI ) but was not associated with odds of membership in the rapid group. In the disability-free subsample, diabetes was associated with an increase in odds of membership in the rapid disability group (OR 1.98, 95% CI ) and obesity was associated with increased odds of membership in both the rapid group and the delayed group compared with the stable group (OR 1.98, 95% CI and OR 2.55, 95% CI , respectively). The results of the fully adjusted zero-inflated Poisson models are shown in Table 3 for the main sample (left panel) and the disabilityfree subsample (right panel). The β coefficients indicate the estimated change in the log count of ADL limitations associated with the presence of each covariate at each measurement point. The exponentiated β coefficients are interpreted as change in ADL limitations associated with each covariate. In the main sample, stroke was associated with increased limitations across all trajectory groups. This was highest among the stable trajectory group where stroke was associated with an increase of nearly three ADL limitations (β = 1.11, p <.001). In the stable trajectory group, heart attack, hip fracture, a CES-D score > 16, and an MMSE score < 22 were associated with increased limitations. In the delayed disability group, financial strain and hip fracture both were associated with increases in ADL limitations (β = 0.25, p <.001 and β = 0.27, p <.01, respectively). Church attendance was associated with decreases in ADL limitations in the stable (β = 0.87, p <.001), delayed (β = 0.45, p <.001), and rapid Figure 1. Estimated activities of daily living (ADL) limitations based on a single population average (double line) and for the stable (solid line), delayed (dashed line), and rapid (dotted line) ADL limitation groups in the main sample (left panel) and the disability-free subsample (right panel). Table 2. Multinomial Logit Models Predicting Trajectory Group Membership Into Delayed and Rapid Trajectory Groups Compared With the Stable Group for the Main Sample (left panel) and the Disability-free Subsample (right panel) in the H-EPESE ( ) Main Sample Subsample Delayed Rapid Delayed Rapid OR (95% CI) OR (95% CI) OR (95% CI) OR (95% CI) Female 1.78 ( ) 2.24 ( ) 1.21 ( ) 2.47 ( ) High school 0.98 ( ) 0.78 ( ) 1.08 ( ) 0.84 ( ) Over weight 1.23 ( ) 0.81 ( ) 1.30 ( ) 0.99 ( ) Obese 2.74 ( ) 1.37 ( ) 2.55 ( ) 1.81 ( ) Diabetes 1.81 ( ) 2.44 ( ) 1.79 ( ) 1.98 ( ) Arthritis 1.59 ( ) 1.94 ( ) 1.53 ( ) 1.89 ( ) Notes: CI = confidence interval; H-EPESE = Hispanic Established Populations for Epidemiologic Study of the Elderly; OR = odds ratio.

5 784 Journals of Gerontology: MEDICAL SCIENCES, 2016, Vol. 71, No. 6 Table 3. Zero-Inflated Poisson Models (β coefficients) Predicting Increases in ADL Limitations as a Function of Age for Each Trajectory Group for the Main Sample and Disability-free Subsample in the H-EPESE ( ) Main Sample Subsample Stable Delayed Rapid Stable Delayed Rapid Married Church 0.87*** 0.45*** 0.10* 0.89*** 0.67*** 0.12** Financial strain *** Heart attack 0.58*** ** Stroke 1.11*** 0.52*** 0.12* 1.15*** 0.52*** 0.18** Hip fracture 1.27*** 0.27** *** 0.47** 0.03 Hypertension MMSE < *** 0.51*** 0.18*** 1.28*** 0.54*** 0.20*** CES-D > *** 0.44*** *** 0.45*** 0.11* Notes: ADL = activities of daily living; CES-D = Center for Epidemiologic Studies Depression Scale; H-EPESE = Hispanic Established Populations for Epidemiologic Study of the Elderly; MMSE = Mini-Mental State Examination. *p.05. **p.01. ***p.001. groups (β = 0.10, p <.05). In the disability-free subsample (Table 3, right panel), stroke, cognitive impairment, and depression were associated with increased ADL limitations in each trajectory group, whereas church attendance was associated with reduced limitations in each group. Hip fracture was associated with significant increases in ADL limitations in the stable and delayed group (β = 1.57, p <.001 and β = 0.47, p <.01, respectively) but not in the rapid group. We also performed sensitivity analyses to examine the impact of attrition and mortality. The probabilities of death at next assessment by age at prior assessment for each trajectory group for both the main sample and the disability-free subsample are shown in Supplementary Figure 1. In both the main sample and the disabilityfree subsample, the probability of death increases with age with no significant difference between trajectory groups, suggesting that differences in functional assessments between groups are not likely due to selective mortality. As a further test, we combined mortality and attrition into a single indicator and modeled a joint trajectory of ADL limitations. These models showed a low joint probability of membership in the rapid limitation group conditional on low attrition in the main sample and in the disability-free subsample. There were no differences in group memberships conditional on higher rates of attrition. Two additional sets of models examined binary outcomes: any ADL limitation as well as any ADL limitation or dead combined into a single variable. Both sets of models identified three trajectories analogous to the main analysis. These sensitivity analyses suggest that although death and attrition may contribute to some of the variation between groups, differential mortality and assessment were not solely responsible for differences in ADL limitations between trajectory groups. Discussion Substantial heterogeneity in disability over time was evident in our examination of the H-EPESE. Using latent class mixture models, we tested our hypothesis that subgroups with distinct patterns of change would be discernable and found that three distinct trajectory groups were in both the total sample and the disability-free subset: stable, delayed increase, and rapid increase in ADL limitations. In our analysis, none of the estimated trajectories closely approximated the overall population average for the duration of the study. It has been reported that Hispanics experience more activity limitations and spend longer lives in disability than non-hispanic Whites (19,34,35). We found that only a subset of Mexican-origin older adults, the rapid ADL limitation group, experienced sustained periods of disability (one or more ADL limitations) from age 73 until age 87. In contrast, the stable trajectory group experienced no disability (less than one ADL limitation) over the same range. It is possible that previous reports of average effects are driven by the presence of ADL limitations in a subset of the Hispanic population and are not representative of the population as a whole. Indeed, when the stable and delayed trajectory groups are combined, our results suggest that many Mexican-origin older adults (about 80 %) experience few ADL limitations until about age 82. An important aspect of our analysis is the ability to describe subgroups at particular risk of ADL limitations and to identify risk factors. The rapid ADL increase group had relatively high rates of chronic conditions (obesity, diabetes, and arthritis). Importantly, the presence of diabetes at baseline was associated with a more than twofold increase in odds of membership in the rapid group. Given the relatively high prevalence of diabetes among Mexican-origin older adults, these results suggest that, by age 73, a substantial proportion is at risk of rapid increase in ADL limitations over the next 10 years. The degree to which this increased risk was due to duration of disease exposure or poorly controlled disease could not be ascertained with the current data. Previous reports on the H-EPESE population found that increased limitations were associated with increased age, low education (less than high school), marital status (not married), diabetes, stroke, and depressive symptoms (11,12). Our results suggest that heterogeneity within the sample limits the broad interpretation of previous findings. For example, similar to other reports, we found that depressive symptoms increase the risk of activity limitations; however, this effect was only observed in the stable and delayed disability groups and not in the rapid disability group in the main sample. Heart attack also had a differential impact across groups only having a significant effect in the stable group in both the main and subsample. Population average approaches are unable to observe these subgroup variations and may lead to biased interpretations. The results of the disability-free subsample analysis were consistent with those of the main sample. The presence of different trajectories in the subsample provides support for the findings from the main sample and further suggests that trajectories of limitations were not merely extensions of ADL limitations at study entry.

6 Journals of Gerontology: MEDICAL SCIENCES, 2016, Vol. 71, No The association of stroke with increased ADL limitations was expected as stroke is a leading cause of disability (36) and a leading contributor to years lived with disability (37). Although stroke was a significant predictor of increased ADL limitations in all groups in our analyses, the results showed a gradient in the magnitude of effect with the largest impact in the stable trajectories and the smallest in the rapidly increasing groups. This gradient suggests that in the relatively healthy, stable groups, a single event contributes to a greater loss of function than in groups that have more competing conditions. The lack of significant contributors to ADL limitations in the rapid disability group was unexpected. In both the total sample and disability-free subsample, the baseline profile of the rapid disability group, including older age, high rates of diabetes, arthritis, stroke, hip fracture, depressive symptoms, and impaired cognition, suggests that this group is already frail. It is possible that, given the groups general frailty, there is little that individual events contribute to disability over time. While an association of hip fracture with increased ADL limitations was expected, the effect was confined to the stable and delayed groups. Although many recover independence within 6 months, as many as 30% of those experiencing fracture will be in a nursing home at 12 months (38) after the fracture. Additionally, rehabilitation following hip fracture can be slow, particularly in elderly patients (39). It is possible that differential access to hospital and rehabilitation across the groups contributed to this observed variation in effect. It is important to note that the magnitude of effect of hip fracture in the stable group was quite large, an estimated increase of 3.5 limitations, underscoring the importance of prevention. Although low levels of education have been associated with poor health outcomes, increased mortality, and increased functional limitations (40), we found no indication of an education effect in our analysis. It is possible that the levels of education in this sample were below the threshold where any benefit could be observed. Although many of the covariates in our analysis were associated with functional decline (increased ADL limitations), a notable exception was religious attendance. The association of monthly church attendance with reduced ADL limitations persisted across trajectory groups in both the main sample and the disability-free subsample. Religious attendance has been found to have a protective effect on activity limitations in older adults in other studies (41,42). The reasons for this positive effect are unclear. Although those with functional limitations may find religiosity a source of coping with disability, it is also possible that regular attendance and associated increased use of prayer may influence ADLs by promoting feelings of optimism and increasing motivation to maintain function. Similarly, regular attenders of church services may be more engaged in their community and more likely to adhere to better health practices and receive more encouragement to remain physically active than nonattenders. Limitations There are several important limitations to this study. The presence of health conditions in our study was self-reported and may not accurately reflect the health of the individual respondents. However, self-report has been found to be relatively reliable in other research (43). We were also unable to assess the severity of any health factor such as stroke, heart attack, hypertension, or diabetes. We were also unable to assess the degree to which a respondent may have received rehabilitation therapy following a potentially disabling event such as stroke or hip fracture. Another limitation was due to the rather coarse measure of time in the H-EPESE study. Assessments were 2 to 3 years apart, which limited our ability to capture more nuanced variations in ADLs, particularly patterns of recovery or decreases in limitations that may occur in the interim between assessments. Conclusions The identification of heterogeneous subgroups highlights the differential effect of predictive factors in groups with different trajectories of disability. Sustained, good physical function among older Mexican-origin adults is particularly challenged by risk factors such as diabetes, heart attack, and stroke. While many factors contributed to increased limitations, the large magnitude of effect of these acute health events stand out as potential targets for interventions. The protective effect of church attendance warrants further exploration. Supplementary Material Please visit the article online at org/ to view supplementary material. Funding This work was supported by grants from the National Institutes of Health (grant numbers R01 AG010939, R01 AG017638, R24 HD065702, and P30 AG024832). Conflict of Interest No potential conflicts of interest exist for any of the authors prior to the month of this submission. References 1. Brault MW. Americans with Disabilities, 2010: Household Economic Studies. Current Population Reports. Washington, DC: U.S. Department of Commerce, Economics and Statistics Administration, U.S. Census Bureau; Murray CJ, Abraham J, Ali MK, et al. The state of US health, : burden of diseases, injuries, and risk factors. JAMA. 2013;310: doi: /jama Verbrugge LM, Jette AM. The disablement process. Soc Sci Med. 1994;38:1 14. doi: / (94) Scott WK, Macera CA, Cornman CB, Sharpe PA. Functional health status as a predictor of mortality in men and women over 65. J Clin Epidemiol. 1997;50: doi: /s (96) Portegijs E, Buurman BM, Essink-Bot ML, Zwinderman AH, de Rooij SE. Failure to regain function at 3 months after acute hospital admission predicts institutionalization within 12 months in older patients. J Am Med Dir Assoc. 2012;13:569.e1 569.e7. doi: /j.jamda Young Y. Factors associated with permanent transition from independent living to nursing home in a continuing care retirement community. J Am Med Dir Assoc. 2009;10: doi: /j. jamda Lowsky DJ, Olshansky SJ, Bhattacharya J, Goldman DP. Heterogeneity in healthy aging. J Gerontol A Biol Sci Med Sci. 2014;69: doi: /gerona/glt Peek MK, Ottenbacher KJ, Markides KS, Ostir GV. Examining the disablement process among older Mexican American adults. Soc Sci Med. 2003;57: doi: /s (02) Gill TM, Gahbauer EA, Lin H, Han L, Allore HG. Comparisons between older men and women in the trajectory and burden of disability over

7 786 Journals of Gerontology: MEDICAL SCIENCES, 2016, Vol. 71, No. 6 the course of nearly 14 years. J Am Med Dir Assoc. 2013;14: doi: /j.jamda Chiu CJ, Wray LA. Physical disability trajectories in older Americans with and without diabetes: the role of age, gender, race or ethnicity, and education. Gerontologist. 2011;51: doi: /geront/gnq Otiniano ME, Du XL, Ottenbacher K, Markides KS. The effect of diabetes combined with stroke on disability, self-rated health, and mortality in older Mexican Americans: results from the Hispanic EPESE. Arch Phys Med Rehabil. 2003;84: doi: /s (02) Raji MA, Al Snih S, Ray LA, Patel KV, Markides KS. Cognitive status and incident disability in older Mexican Americans: findings from the Hispanic established population for the epidemiological study of the elderly. Ethn Dis. 2004;14: Stuck AE, Walthert JM, Nikolaus T, Büla CJ, Hohmann C, Beck JC. Risk factors for functional status decline in community-living elderly people: a systematic literature review. Soc Sci Med. 1999;48: doi: / S (98) Lunney JR, Lynn J, Foley DJ, Lipson S, Guralnik JM. Patterns of functional decline at the end of life. JAMA. 2003;289: doi: / jama James BD, Boyle PA, Buchman AS, Bennett DA. Relation of late-life social activity with incident disability among community-dwelling older adults. J Gerontol A Biol Sci Med Sci. 2011;66: doi: /gerona/glq Eide AH, Røysamb E. The relationship between level of disability, psychological problems, social activity, and social networks. Rehabil Psychol. 2002;47: doi: / Yu HW, Chen DR, Chiang TL, Tu YK, Chen YM. Disability trajectories and associated disablement process factors among older adults in Taiwan. Arch Gerontol Geriatr. 2015;60: doi: /j.archger Fauth EB, Zarit SH, Malmberg B, Johansson B. Physical, cognitive, and psychosocial variables from the Disablement Process Model predict patterns of independence and the transition into disability for the oldest-old. Gerontologist. 2007;47: doi: /geront/ Hayward MD, Hummer RA, Chiu CT, González-González C, Wong R. Does the Hispanic paradox in U.S. adult mortality extend to disability? Popul Res Policy Rev. 2014;33: doi: /s Carrasquillo O, Lantigua RA, Shea S. Differences in functional status of Hispanic versus non-hispanic White elders: data from the Medical Expenditure Panel Survey. J Aging Health. 2000;12: doi: / Brown A, Patten E. Statistical Portrait of Hispanics in the United States, Hispanic Trends Project. Washington, DC: Pew Research Center; Montero-Odasso M, Bergman H, Béland F, Sourial N, Fletcher JD, Dallaire L. Identifying mobility heterogeneity in very frail older adults. Are frail people all the same? Arch Gerontol Geriatr. 2009;49: doi: /j.archger Markides KS, Rudkin L, Angel RJ, Espino DV. Health Status of Hispanic Elderly. In: Martin L, Soldo B, eds. Racial and Ethnic Differences in the Health of Older Americans. Washington, DC: National Academies Press; 1997: Markides KS, Stroup-Benham CA, Black SA, Satish S, Perkowski LC, Ostir G. The Health of Mexican American Elderly: Selected Findings from the Hispanic EPESE. Planning services for minority elders in the 21st century. New York, NY: Springer; 1999: Branch LG, Katz S, Kniepmann K, Papsidero JA. A prospective study of functional status among community elders. Am J Public Health. 1984;74: doi: /ajph Folstein MF, Folstein SE, McHugh PR. Mini-mental state. A practical method for grading the cognitive state of patients for the clinician. J Psychiatr Res. 1975;12: Mulgrew CL, Morgenstern N, Shetterly SM, Baxter J, Barón AE, Hamman RF. Cognitive functioning and impairment among rural elderly Hispanics and non-hispanic whites as assessed by the Mini-Mental State Examination. J Gerontol B Psychol Sci Soc Sci. 1999;54:P223 P230. doi: / geronb/54b.4.p World Health Organization. Obesity: preventing and managing the global epidemic. World Health Organization; Raji MA, Ostir GV, Markides KS, Goodwin JS. The interaction of cognitive and emotional status on subsequent physical functioning in older Mexican Americans: findings from the Hispanic established population for the epidemiologic study of the elderly. J Gerontol A Biol Sci Med Sci. 2002;57:M678 M682. doi: /gerona/57.10.m Radloff LS. The CES-D scale a self-report depression scale for research in the general population. Appl Psychol Meas. 1977;1: doi: / Nagin DS. Analyzing developmental trajectories: a semiparametric, groupbased approach. Psychol Methods. 1999;4: Jones BL, Nagin DS. A note on a stata plugin for estimating group-based trajectory models. Sociol Methods Res. 2013;42: Haviland AM, Jones BL, Nagin DS. Group-based trajectory modeling extended to account for nonrandom participant attrition. Sociol Methods Res. 2011;40: Hummer RA, Benjamins MR, Rogers RG. Racial and ethnic disparities in health and mortality among the US elderly population. In: Cohen B, Bulato RA, Anderson NB, eds. Critical Perspectives on Racial and Ethnic Differences in Health in Late Life. Washington, DC: National Academies Press; 2004: Markides KS, Eschbach K, Ray LA, Peek MK. Census disability rates among older people by race/ethnicity and type of Hispanic origin. In: Agel JL, Whitfield KE, eds. The Health of Aging Hispanics. New York: Springer; 2007: Centers for Disease Control and Prevention. Prevalence and most common causes of disability among adults United States, MMWR Morb Mortal Wkly Rep. 2009;58: U. S. Burden of Disease Collaborators. The State of US Health, : burden of diseases, injuries, and risk factors. JAMA. 2013;310: doi: /jama Leibson CL, Tosteson AN, Gabriel SE, Ransom JE, Melton LJ. Mortality, disability, and nursing home use for persons with and without hip fracture: a population-based study. J Am Geriatr Soc. 2002;50: Arinzon Z, Fidelman Z, Zuta A, Peisakh A, Berner YN. Functional recovery after hip fracture in old-old elderly patients. Arch Gerontol Geriatr. 2005;40: doi: /j.archger Freedman VA, Martin LG. The role of education in explaining and forecasting trends in functional limitations among older Americans. Demography. 1999;36: doi: / Park NS, Klemmack DL, Roff LL, et al. Religiousness and longitudinal trajectories in elders functional status. Res Aging. 2008;30: Hybels CF, Blazer DG, George LK, Koenig HG. The complex association between religious activities and functional limitations in older adults. Gerontologist. 2012;52: doi: /geront/gnr Zelinski EM, Crimmins E, Reynolds S, Seeman T. Do medical conditions affect cognition in older adults? Health Psychol. 1998;17: doi: /