The Heterogeneity of Disability Trajectories in Later Life: Dynamics of Activities of Daily Living Performance Among Nursing Home Residents

Abstract

Objective: This study investigated the variability in activities of daily living (ADL) trajectories among 6,155 nursing home residents using unique and rich observational data. Method: The impairment in ADL performance was considered as a dynamic process in a multi-state framework. Using an innovative mixture model, such states were not defined a priori but inferred from the data. Results: The process of change in functional health differed among residents. We identified four latent regimes: stability or slight deterioration, relevant change, variability, and recovery. Impaired body functions and poor physical performance were main risk factors associated with degradation in functional health. Discussion: The evolution of disability in later life is not completely gradual or homogeneous. Steep deterioration in functional health can be followed by periods of stability or even recovery. The current condition can be used to successfully predict the evolution of ADL allowing to set and target different care priorities and practices.

Keywords

ADL trajectories longitudinal analysis variability disability trajectories multi-state model nursing home resident

Introduction

Over the last several decades with a substantial increase in life expectancy, the interest on aging and health of the aged population has grown exponentially. Elderly people are at high risk of losing independence in activities of daily living (ADL) performance (Jacobs et al., 2012), and the development of ADL impairments is a strong predictor of admission to a nursing home (NH; Gaugler, Duval, Anderson, & Kane, 2007; Jette, Branch, Sleeper, Feldman, & Sullivan, 1992). Among NH residents, being functionally active and able to participate in everyday activities improves self-esteem and more generally the quality of life (Edvardsson, Petersson, Sjogren, Lindkvist, & Sandman, 2014; Murphy, O’Shea, & Cooney, 2007). Thus, the analysis of patterns of change in ADL performance and the associated demographic factors and health impairments are relevant for determining care priorities and public health expenditures (Lindholm, Gustavsson, Jönsson, & Wimo, 2013), and to identify measures able to slow down the decline in ADL capacities.

The present study aimed to shed light on the evolution of functional health in later life using secondary data from a unique and rich observational study conducted in Switzerland from 1997 to 2007 on a large sample of NH residents.

Disability is a complex and multifaceted process. As described by Verbrugge and Jette (1994), a disablement process exists. Decline in functional capacities is not completely age-related but rather affected by several different physical, psychological, and social factors, and a relevant heterogeneity in evolution of functional performance in later life exists. Guilley et al. (2008) showed that decline in ADL performance was not irreversible with several cases of recovery from situations of frailty, and identified limitations in memory, energy, and sensory capacity as predictors of developing ADL dependence. Hardy, Dubin, Holford, and Gill (2005) studied the return to independence of older persons who experienced multiple but brief episodes of disability, with participants with impaired mobility more likely to experience disability episodes. Similarly, developmental studies using cluster-based approaches have identified multiple possible disability patterns. Han et al. (2013) identified five different patterns in adults aged 70 years and older with cognitive impairment, chronic conditions and physical frailty as main risk factors related to deterioration in ADL functioning. Gill, Gahbauer, Han, and Allore (2010), using monthly observational data on community-dwelling older persons, identified five distinct disability trajectories.

Relatively few longitudinal studies have been conducted on change in ADL capacities among NH residents (Palese et al., 2016). Most have focused on specific residents such as those who have been hospitalized (Kruse, Petroski, Mehr, Banaszah-Holl, & Intrator, 2013) or died (J. H. Chen, Chan, Kiely, Morris, & Mitchell, 2007) during the follow-up, and those with middle stage or severe dementia (Carpenter et al., 2006; Helvik, Engedal, Benth, & Selbaek, 2015; Slaughter, Eliasziw, Morgan, & Drummond, 2010). Banaszak-Holl et al. (2011) found an overall trend of losing ADL capacity over time in a sample of Medicare NH residents in the United States, with physical and cognitive impairments at baseline as main predictors of a faster onset in ADL impairment. Medical conditions were instead weakly associated with rate of development of functional impairment. McConnell, Pieper, Sloane, and Branch (2002) reported a decline per year of 0.84 points in ADL (scale score range from 0 to 20 points) among NH residents. Among the characteristics considered in their study, only marital status and cognition had a significant effect on the evolution of ADL dependence. Looking at the trajectories of ADL loss in terms of three hierarchical stages―early, middle, and late loss―bladder incontinence and poor balance were associated with deterioration in the early (personal hygiene) and middle (toileting) stages of ADL loss (Wang, Chang, Elberly, Virnig, & Kane, 2010).

The complexity of disability trajectories in later life explains the need for using appropriate statistical techniques. Approaches commonly used in the medical literature include time-to-event models, such as event history analysis to analyze the time until a certain event, multistate life tables to estimate an average duration in each state and the transition probabilities among them (e.g., mortality or morbidity rates), and group-based modeling techniques that estimate a set of common disability patterns (mixture models or mixed effect models) (see Lynch & Taylor, 2016, for a review). The latter approach has been also extended to test the effect of earlier conditions on the current one (i.e., an autoregressive model). The Autoregressive Latent Trajectory (ALT) model introduced by Bollen and Curran (2004) has been used to establish the effect of earlier depressive symptoms on disability (Chen et al., 2012), test the direction of the relationship between depressive symptoms and memory loss (Zahodne, Stern, & Manly, 2014), and investigate the influence of NH contextual characteristics on the quality improvement of resident outcomes (Wan, Zhang, & Unruh, 2006). The results from these studies show a clinically and statistically significant impact of the autoregressive components. Including prior levels of the outcome variable brings better understanding of the evolution of a health condition over time.

Each of these approaches has potential limitations in fully capturing the heterogeneity in the evolution of functional health in later life. Mixed effect models for instance, although able to distinguish differences in the overall disability trend among groups, have an underlying hypothesis of gradual accumulation of functional limitation assuming a smooth functional form in the trajectories (Engelman & Jackson, 2017). Disability trajectories might instead show complex non-homogeneous behaviors (see Figure 1 in supplemental materials).

In the present study, we conducted a secondary data analysis examining changes in ADL functioning of NH residents using a person-oriented approach looking at individual pathways of change in functional health. The ADL disability is not considered a static circumstance or a single transition, but as a dynamic process (Hardy et al., 2005) in a multi-state framework. Unlike conventional multi-state models (Gill & Kurland, 2003), such states are not defined a priori but inferred from the data; that is, they are latent regimes.

To identify unobserved distinctive conditions and to analyze the heterogeneity in functional performance, we used an innovative autoregressive longitudinal mixture model called the Hidden Mixture Transition Distributional (HMTD; Bolano & Berchtold, 2016) model. HMTD is a Markov switching regime model that allows the analysis of individual trajectories in terms of change in an underlying construct to account for non-linearity in disability trajectories. Switching regime models have been applied extensively in finance and economics to analyze sudden changes in economic trends since Hamilton (1989). However, the model has not been used in research on aging thus far.

Focusing on disability in later life as a complex and multidimensional phenomenon, this study aimed to (a) shed light on the different pathways of change in functional health in later life; (b) determine whether the process of losing ADL capabilities among NH residents is irreversible; and (c) assess demographic, cognitive, and physical health factors associated with different patterns in losing ADL capacity. In particular, we identified impairments in body functions and mobility as primary factors associated with a worsening in ADL capacity (e.g., see reviews from Paterson & Warburton, 2010; Tak, Kuiper, Chorus, & Hopman-Rock, 2011; Vermeulen, Neyens, van Rossum, Spreeuwenberg, & de Witte, 2013). In a similar way to ALT models, we accounted for the effect of previous ADL scores on the current condition.

Method

Data

The data used in this study were derived from a longitudinal study conducted from 1997 to 2007 in 90 NHs in Switzerland. The residents received a first assessment (using the Resident Assessment Instrument Minimum Data Set [RAI-MDS]) shortly after their entry into the NH. Then, subsequent assessments were performed whenever the health condition underwent a significant change. A total of 21,821 residents were included in the original study, and about half received more than one assessment (N = 10,199). To analyze inter- and intra-individual variations in disability trajectories and to test the effect of previous ADL scores on the current one, this study focused on medium long-term residents defined as residents who received at least four assessments. The resulting sample included 6,155 residents (4,503 females). Because new assessments were performed each time a significant change occurred in the health of the respondents, data used in this study represent very accurately the evolution of ADLs, hence the trajectories overtime. The original sample characteristics have been described in Bürge, Berchtold, and von Gunten (2011). As the sample size was large, we fixed the significance level for all statistical tests at .001 to avoid Type I error.

Measures

Dependence in ADL was measured using seven items from the RAI-MDS: bed mobility, transfer, locomotion on unit, toilet use, personal hygiene, dressing, and eating. Each item is evaluated separately with five possible answers ranging from independence (score 0) to total dependence (score 4; Morris et al., 1994). Combining the items leads to three types of ADL summary measures: the hierarchical ADL self-performance rating scale and two additive scales—the ADL Short-Form Scale and the ADL Long-Form Scale. We focused on the latter measure because the aim of the study was to analyze any change in ADL performance and, according to Morris et al. (1994), the ADL Long-Form Scale is more successful in identifying minor incremental changes. The outcome variable, scores on the MDS-ADL Long-Form Additive Scale, ranges from 0 to 28. For convenience, in the rest of the article, we will refer to the MDS-ADL Long-Form Scale simply as ADL.

The sociodemographic and impairment factors considered in this study were the following: gender; age group (≤64, 65-79, 80 and older); Cognitive Performance Scale (CPS) score ranging from 1 to 6, with higher scores indicating a higher level of impairment (Morris et al., 1994); body mass index (BMI: underweight < 18.5, normal weight = 18.5–24.9, overweight and obese ≥ 25); visual (see well, slight difficulties, moderate difficulties, severe difficulties, or blind); urinary or fecal incontinence; balance (maintained position, unsteady, partial physical support, not able to attempt test alone); outdoors walking or wheeling without any help; and MDS Depression Rating Scale score ranging from 0 to 14 (Burrows, Morris, Simon, Hirdes, & Phillips, 2000). Given health conditions may change substantially over time influencing the developmental trajectories of functional health, the above impairments were considered as time-varying. See supplemental material for a description of the sample at baseline.

Statistical Model

Due to the variability in ADL trajectories, the choice of an accurate statistical model is not straightforward. Over the last decades, many studies on health trajectories in later life (e.g., Gill et al., 2010; Liang, Xu, Bennett, Ye, & Quinones, 2009; Zimmer, Martin, Jones, & Nagin, 2014) relied on cluster-based approaches such as Latent Growth Curve models (Bollen & Curran, 2006), Growth Mixture Modeling (e.g., Asparouhov & Muthén, 2009), and Group-Based Trajectory modeling (e.g., Nagin, 1999). The aim of these methods is to estimate developmental trajectories, one for each cluster, that approximate (“average”) the general trend of the respondents classified in that specific cluster. These approaches are particularly appealing because they are relatively easy to fit and, using an underlying continuous distribution function, they produce smooth estimated average trajectories ready to represent and interpret. However, the model assumes a gradual and smooth accumulation of functional limitation focusing on inter-individual differences in the overall disability trajectory rather than on intra-individual variations in health conditions. This might lead to an oversimplification of the complexity of evolution of health conditions in later life.

Cluster-based models have also some limitations when applied to observational data. These models mainly apply to discrete time measurements collected at fixed occasions (e.g., yearly panel data) or they are used to analyze the evolution of a characteristic of interest with respect to the age of the participants. Moreover, they do not allow directly for missing information or having different numbers of repeated measures by subject (unbalanced panel data) as common in observational studies.

In addition to the heterogeneity and potential non-gradual change in disability trajectories in later life, in a study like ours, we had to take into account the variations in age and health conditions of the NH residents at baseline (see supplemental material), and the number of assessments that varied significantly from one participant to another (from four up to 23) with relevant dropping-out (only a quarter of the sample received more than seven assessments). Moreover, the assessments were not collected on a regular basis. The average time between assessments was 166.17 days with a standard deviation of 71.56 days (Mdn = 178). The statistical approach used in this study allowed for overcoming these issues.

This study analyzed changes in ADL using the HMTD model introduced by Bolano and Berchtold (2016). The HMTD model is an autoregressive mixture-based model to analyze longitudinal data switching between alternative unobserved regimes. Such latent regimes represent different pathways of change in ADL performance an older person might go through. For each pathway, a set of risk factors is estimated. In a similar way to factor loadings in a latent class model, the estimated coefficients of the risk factors allow for the interpretation and labeling of the regimes. They show the contribution of the risk factor to the expected ADL score in a certain regime. For example, consider urinary incontinence as a risk factor represented by a dummy variable equal to 1 if the NH resident is incontinent and 0 if otherwise. Assuming that the estimated coefficient will be equal to 2, it means that, in that regime, an incontinent NH resident has on average an ADL score two points higher than a similar NH resident not incontinent.

Unlike conventional mixture-based models (e.g., Growth Mixture models), NH residents are not classified as belonging to one single pathway, rather they can move (transit) between hidden regimes at each time point allowing subjects to follow different patterns of change in functional health during their period of observation. In such a way, the model is able to better capture the heterogeneity in disability trajectory. Similar to the ALT models (Bollen & Curran, 2004), the model includes previous ADL scores as predictors of future disability level. Including previous levels of functional disabilities $(A D L_{t - 1}, A D L_{t - 2}, \dots)$ as explanatory variables may bring about a better estimation of the dynamics of functional limitations (C.-M. Chen et al., 2012), and allow for analyzing the cumulative effects of ADL performance over time. The coefficients estimated for earlier ADL scores can then be interpreted as the effect of previous ADL conditions on the current assessment. For example, if the previous ADL score $(A D L_{t - 1})$ was 10 and the associated autoregressive coefficient is estimated to be 0.9, it means that the current ADL score is estimated to be 9 (10 × 0.9), assuming no other risk factors are considered.

This approach differs from previous studies that used a conventional Markov multi-state model to analyze the transition between states of disability (e.g., Beckett et al., 1996; Chiang, 1968). Our multi-state representation of ADL impairment is not a priori defined but inferred from the data. Starting from the individual impairment levels, the HMTD model is able to estimate the underlying ADL disability pathways that characterize the evolution in functional capabilities. The different ADL regimes identified by the model are latent characteristics. NH residents, even if they share similar health levels, may belong to different latent regimes (Figure 2 in the supplemental material). As suggested by Scott, James, and Sugar (2005), unlike the classical medical perspective where genotypes are defined by an individual’s characteristics, here the latent regimes can be considered as genotypes and individual characteristics as phenotypes.

Results

Descriptive Analysis

The sample consisted of 6,155 NH residents with a mean ADL score at baseline (entry in the NH) of 9.02 points (minimum = 0, maximum = 28; Mdn = 7; SD = 8.23).

The gender difference in ADL scores was not statistically significant, with men (26.84% of the participants) being just slightly less dependent than women (mean scores were 9.35 and 8.9 points, respectively). The level of ADL impairment at baseline was also not completely related to the age of the subject although a significant age group difference appeared. The highest level of impairment was in NH residents aged less than 65 years (11.95 points). This result is not surprising as people below 65 years are admitted in this kind of residential facility only in the case of severe disability. Among residents aged 65 and older, the cohort characterized by the highest level of ADL impairment at admission was those between 65 and 79 years old (23.25% of sample) in which the ADL mean score was 9.67 points. ADL impairment seemed strongly related to a pathologically low BMI (ADL mean score = 11.12 points), impaired body function as incontinence (12.81 points), impaired physical performance as poor balancing (15.7 points), and reduced mobility outdoors (13.18 points).

ADL disability was quite stable in the short run. The ADL score did not change between two assessments in 41.05% of the cases (15,899 transitions) and, looking at small variations, the differences were an absolute value of only 1 point in 7,240 cases (18.69% of the total). Using a difference of 2 points as a criterion for a relevant change in health impairment (e.g., Carpenter, Hastie, Morris, Fries, & Ankri, 2006), evidence of decline in functional health was observed in 27.8% of the events. Among them, 1,729 changes corresponded to a very large decline of at least 10 points. In 12.46% of the cases (4,827), we observed a recovery in ADL capacities with an improvement between two assessments of at least two points.

Looking at the change in any two consecutive assessments (Table 1), a slight decrease of 1.07 points was observed at each follow-up assessment but with a relevant level of variation among participants (SD = 4.21) and no significant gender differences (p value of Welch’s t test = .017; mean change = 0.98 men, 1.1 women). However, a significant age effect was observed. Older NH residents were more likely to experience a decline in their disability condition with a mean variation between two assessments of 0.38 points for residents below age 65, 0.81 for those aged 65 to 79, and 1.21 for residents aged 80 and older. See also Figure 3 in supplemental material for a graphical representation of the distribution of the change in ADL scores between two consecutive assessments by gender and age group.

Table 1.

Change in ADL Score Between Two Assessments (Left-Hand Side) and Between Last and First Assessment (Right-Hand Side), by Gender and Age at Baseline.

	Change between two consecutive assessments			Change between first and last assessment
	M	SD	p value	M	SD	p value	Improve^a %	Decline^b %	Stable^c %
Overall (n = 6,155)	1.07	4.21		6.72	8.29		11.52	67.12	21.36
Female (n = 4,503)	1.1	4.29	.017	7.04	8.38	<.001	10.84	68.56	20.61
Man (n = 1,652)	0.98	3.39		5.84	7.97		13.38	63.19	23.43
Age group
≤64 (n = 247)	0.38	2.53	<.001	2.96	6.82	<.001	14.97	44.56	40.46
65-79 (n = 1,431)	0.81	3.7		5.47	8.17		13.69	60.06	26.24
80+ (n = 4,477)	1.21	4.45		7.32	8.31		8.11	73.22	18.67

Note. p value of the test for difference in means; Welch two-sample t test for gender difference; ANOVA for mean difference among age groups. ADL = activities of daily living; NH = nursing home.

Percentage of NH residents. Decrease of at least two points in ADL score.

Decline = increase of at least 2 points.

Stability = no variations or change in absolute value of 1 point.

Considering the entire assessment period, a general trend of decline was observed in the study with a mean change of 6.72 points from baseline to last assessment. Nevertheless, once again, a relevant variability in the patterns was observed: 67.12% of NH residents experienced degradation in ADL performance, 21.36% (n = 1,315) remained stable (i.e., the difference in ADL score between the beginning and the end of the period of assessment was less than 2 points), and, in 709 cases (11.52%), an improvement over the assessment period (mean change of at least 2 points) was observed. Female NH residents were more likely to have a decline in ADL performance (a statistically significant mean difference of 7.04 points for women and 5.84 for men). The share of NH residents who experienced a worsening in ADL capacities increased with age (Table 1).

To summarize, although there was a general trend of degradation in impairment level, quite high levels of heterogeneity were observed both in intra-individual and inter-individual changes. Some participants remained stable over time while others suffered a relevant degradation in ADL score even between two successive assessments. Within the same trajectory, important variations may exist, switching between periods of stability and periods of changing as shown in the Figure 1 in the supplemental materials. These results suggest that decline in ADL performance is not gradual and smooth over time.

Modeling Functional Disability Trajectories

The number of latent regimes that best represent the observed ADL trajectories was determined according to the Bayesian Information Criterion (BIC; Schwarz, 1978), and the distribution of the different regimes to avoid rare situations (regimes observed less than 2% of the time). A four-regime model was selected (see supplemental Material).

The estimated coefficients reported in Table 2 represent the association of each regime to a possible pathway of change in ADL capacity. The magnitude of the risk factors represents the average contribution, in points, of the correspondent risk factor to the ADL score in that regime. The coefficients for the autoregressive components (i.e., $A D L_{t - 1}$ and $A D L_{t - 2}$ ) show the effect of previous ADL score on the current level of functional performance. The four latent regimes are associated with pathways of stability or slight deterioration, relevant change, variability, and recovery from a shock. The four regimes differed in prevalence occurring from 71% of the time in Regime 1 to 3.17% in Regime 4.

Table 2.

Estimated Parameters for Each Regime.

Variable	Regime 1	Regime 2	Regime 3	Regime 4
Variable	Normative	Worsening	Variability	Recovery
Constant	−0.0896* (0.016)	0.8932* (0.106)	−0.1728 (0.123)	−0.3727 (0.124)
Constant	[<.001]	[<.001]	[.156]	[.006]
ADL (t – 1)	0.9556* (0.001)	0.0796* (0.010)	0.9780* (0.012)	0.1563* (0.014)
ADL (t – 1)	[<.001]	[<.001]	[<.001]	[<.001]
ADL (t – 2)	0.0182* (0.001)	0.1409* (0.009)	−0.2946* (0.010)	0.7429* (0.014)
ADL (t – 2)	[<.001]	[<.001]	[<.001]	[<.001]
Gender (ref. Male)
Female	−0.0064 (0.017)	−0.1647 (0.140)	−0.3976 (0.151)	−0.0329 (0.188)
Female	[.598]	[.210]	[.04]	[.976]
Age (ref. 65-79)
Less than 64	0.0161 (0.029)	0.5499 (0.490)	0.1374 (0.338)	−0.1353 (0.279)
Less than 64	[.598]	[.210]	[.518]	[.768]
80 and more	0.0240 (0.018)	−0.2430 (0.138)	0.3360 (0.157)	0.0418 (0.182)
80 and more	[.120]	[.124]	[.044]	[.854]
BMI (ref. Normal)
Overweight	0.0104 (0.016)	−0.4081 (0.143)	−0.0257 (0.158)	−0.0505 (0.157)
Overweight	[.454]	[.018]	[.692]	[.758]
Underweight	0.0568 (0.031)	0.6504 (0.209)	0.9611* (0.302)	−0.4472 (0.376)
Underweight	[.048]	[.008]	[<.001]	[.200]
Eyesight (ref. See well)
Slight difficulties	0.0716* (0.024)	0.4765 (0.174)	0.6298 (0.204)	0.0033 (0.246)
Slight difficulties	[<.001]	[.010]	[.002]	[.846]
Moderate difficulties	0.0659 (0.038)	1.3207* (0.263)	0.7003 (0.294)	0.2615 (0.362)
Moderate difficulties	[.034]	[<.001]	[.026]	[.402]
Severe difficulties or blind	0.1329* (0.044)	2.2339* (0.280)	1.3767* (0.297)	0.1862 (0.468)
Severe difficulties or blind	[<.001]	[<.001]	[<.001]	[.766]
Balance (ref. Maintained position)
Unsteady	0.0146 (0.020)	0.5965* (0.179)	−0.0879 (0.175)	0.1789 (0.157)
Unsteady	[.482]	[<.001]	[.550]	[.280]
Partial support	0.0758 (0.028)	2.0048* (0.257)	0.4072 (0.299)	0.8428 (0.269)
Partial support	[.002]	[<.001]	[.148]	[.004]
Not able to attempt the test alone	0.1861* (0.026)	5.7208* (0.295)	1.7003* (0.223)	0.6299* (0.236)
Not able to attempt the test alone	[<.001]	[<.001]	[<.001]	[<.001]
Cognitive Performance Scale (ref. Less than 3 points)
Moderate/severe impairment	0.1329* (0.022)	1.2851* (0.164)	1.4915* (0.215)	0.7773* (0.211)
Moderate/severe impairment	[<.001]	[<.001]	[<.001]	[<.001]
Depression score	0.0139* (0.004)	0.0705* (0.023)	0.1351* (0.032)	0.0161 (0.043)
Depression score	[<.001]	[<.001]	[<.001]	[.780]
Outdoors walking or wheeling (ref. Yes)
No	0.2477* (0.026)	8.3781* (0.023)	1.5309* (0.229)	0.7041 (0.236)
No	[<.001]	[<.001]	[<.001]	[.022]
Continence (urinary or fecal) (ref. Continent)
Incontinence	0.2002* (0.022)	1.9827* (0.183)	2.2203* (0.244)	0.9278* (0.235)
Incontinence	[<.001]	[<.001]	[<.001]	[<.001]
SD	1.1355* (0.123)	7.7253* (0.262)	5.6567* (0.162)	3.3782* (0.126)
SD	[<.001]	[<.001]	[<.001]	[<.001]

Note. ADL (t – 1) and ADL (t – 2) represent the first and the second lag of ADL score, that is, the ADL scores of the two previous assessments. Bootstrapped standard errors are reported in parentheses. p values for statistical significance given in brackets are based on 1,000 bootstrap replications. ADL = activities of daily living; BMI = body mass index.

Coefficient significant at the .001 level.

In Regime 1 (prevalence of 71% of the time), the current ADL score was strongly related to the value observed at the time of the previous assessment. The estimated coefficient of ADL at t – 1 of 0.96 suggests stability in functional health. Looking at the risk factors considered in the analysis, a slight deterioration in functional performance was associated with a worsening in eyesight and balance, severe cognitive impairment, incontinence, and difficulties with going outside the NH without any help. However, the average increase in ADL score implied by these risk factors is smaller than 0.2 points. We typified this as a Normative regime because it represents a condition of stability or slight deterioration in ADL impairment over time.

The estimated risk factors in Regime 2 (used 17.4% of the time) were positive and relatively high in magnitude, while the autoregressive coefficients showed only a limited influence of previous ADL scores on the current one (0.08 and 0.14). This regime represents then a significant deterioration in ADL capacity. We labeled this as a Worsening regime. The inability of going outdoors determined on average an increase in ADL score of 8.4 points; poor and severe balancing were associated with a degradation in ADL performance of 2 and 5.7 points, respectively.

In Regime 3, the coefficient estimated for the previous ADL score (ADL at t – 1) was positive and quite high (0.98), while the second autoregressive coefficient (ADL at t – 2) was negative (–0.3) suggesting a fluctuation in ADL scores. Risk factors associated with a variation in ADL in this regime were depression (increase of 0.14 points), inability of walking or wheeling outdoors alone (1.5 points), severe eyesight difficulties (1.4 points), and incontinence (2.2 points). This regime represents then a slight reduction in functional ability, as well as a possible fluctuation of ADL score over two assessments. We typified this as a Variability regime (observed 8.44% of the time).

The mean ADL level in Regime 4 was strongly related to the ADL observed two assessments before (0.743) but only poorly with the previous one (0.156). This regime represents then a Recovery after a shock (prevalence of 3.17%). Cognitive impairment, incontinence, and poor balance were poorly associated with a worsening in ADL scores in this regime with a mean effect of less than 1 point.

The estimated standard deviations (Table 2) represent the variability within each regime. They ranged from 1.1 for the Normative regime to 7.7 for the Variability regime, showing high levels of heterogeneity in the individual patterns.

One of the outputs of the HMTD is the estimation of the transition probabilities among hidden regimes (see Zucchini & MacDonald, 2009, for a complete discussion on the estimation of latent Markov models). In other words, the model predicts the likelihood of moving from one pathway to another in the following observational period (Table 3). Such probabilities can be used to predict the falling in an unwanted condition (e.g., the transition into a regime of degradation in ADL performance). Despite the transition probabilities representing the probability of moving from one regime to another at each assessment, they can also be used to describe a longer trend. In particular, the values on the diagonal in Table 3 represent the probabilities of staying in the same pathway over time (recurrence). Worsening, Recovery, and Variability regimes were poorly recurrent, indicating that it is unlikely, for instance, to have a trajectory of continuing accumulation (recurrence of Worsening regime) of ADL disability. They represent short-term conditions.

Table 3.

Transition Probabilities Among Latent Regimes.

	Normative	Worsening	Variability	Recovery
Normative	.492	.241	.205	.014
Worsening	.575	.176	.113	.135
Variability	.574	.267	.06	.099
Recovery	.564	.222	.201	.062

Whatever the current regime, the most likely transition for an NH resident was to move to a stable condition (around a 50% chance of going to the Normative regime as shown in the first column of the transition matrix). The second most likely transition was to experience degradation in ADL performance (i.e., transition to the Worsening regime, second column). Even if the Normative regime was the most common one and represents a situation of stability, NH residents being in this Normative condition were not protected against a future change in ADL capabilities. They had a 24% risk of experiencing a steep degradation in their ADL condition (transition probability from Normative to Worsening) and another 20% risk of experiencing a small deterioration (probability of moving to Variability regime). After recovery from a positive or negative shock, NH residents had more than a 40% chance of experiencing deterioration in their ADL performance (transition probability from Recovery to Worsening and from Recovery to Variability regimes is 22.2% and 20.1%, respectively). A small change in ADL score can be a pre-condition of a more severe worsening in ADL performance. The transition from the Variability regime to the Worsening regime occurred in 26.7% of the cases.

Discussion

Different Patterns of Change in ADL

This study sheds light on the variability of functional disability trajectories in later life using unique and rich observational longitudinal data on a large sample of NH residents. The secondary data analysis conducted in this study clearly shows that the process of losing independence in ADL performance is not completely gradual and continuous over time. Although there was a general trend of degradation in impairment level while aging (mean change of 6.7 points between baseline and last assessment), quite high levels of heterogeneity were observed in inter- and intra-individual changes. Some participants remained stable over time while others suffered a relevant degradation in ADL score even between two successive assessments. Within the same trajectory, important variations may exist, switching between periods of stability and periods of change. Moreover, in 12.46% of the cases, the ADL performance improved from one assessment to another by at least two points. Thus, the process of losing ADL independence is not irreversible.

Multi-state models have been used in the literature to analyze the onset of disability (e.g., Gill & Kurland, 2003; Guilley et al., 2008). However, in these studies, the states of disability have been defined a priori often using arbitrary categorizations. Conversely, in the present study, the states, or regimes, were inferred from the data. Using a switching regime model, we are then able to shed light on the ADL impairment process without imposing any theoretical constraints or arbitrary choices, and potentially reducing the measurement error (see, for instance, Van de Pol & de Leeuw, 1986; Wiggins, 1973).

We identified four possible pathways NH residents might experience: one of stability, one of recovery, and two of change. We can either have a relevant change (Worsening) or a minor incremental change characterized by fluctuations in ADL performance (Variability). Such results are fairly consistent with the literature. Liang et al. (2009), using a group-based trajectory model on changing in functional health of U.S. residents aged 50 and above and controlling for time-varying covariates, identified three trajectories of disability: health functioning, moderate decrement, and large functional decrement. Similarly, Deeg (2005), excluding the trajectories that ended by death of the subject, identified two unstable and three stable conditions. Unlike these studies where each respondent is classified in one smooth developmental disability trajectory, the HMTD used here allows for a better description of the intra-individual variations in the evolution of ADL capacities, because each respondent can be described as moving among the four disability pathways identified. Although the estimated transition probabilities show a general tendency to experience periods of stability or slight deterioration in functional health, as shown by the predominance of the Normative regime, NH residents were likely to undergo a degradation of ADL status. NH residents had almost a 50% chance of moving out of a condition of stability and experiencing a change in their ADL score in a following assessment. However, such sudden changes were not persistent and residents were likely to return to a stable condition. A period of fluctuation in ADL capacity was likely to be followed by an important degradation of the ADL status (transition probability of 27% from the Variability to the Worsening regime).

The Risk Factors

The HMTD model allows for an estimation of the effect of risk factors in each pathway separately. The main risk factors associated with a relevant degradation in ADL performance concerned impaired body function (incontinence), and impaired physical performance such as poor balance and lack of outdoors mobility (Lahmann et al., 2015). Promoting mobility and physical exercises, in particular for improving balance, can then be a way to prevent degradation in ADL performance of NH residents (Forster, Lambley, & Young, 2010; Paterson & Warburton, 2010; Peri et al., 2008; Slaughter et al., 2015).

In accordance with the results shown by Carpenter et al. (2006), we found that moderate or severe cognitive impairment was a relevant risk factor for functional decline. Severe visual deficits were also clearly related to a loss in ADL capabilities as well as a pathologically low BMI and depression. However, for the latter factor, we did not find a strong effect on change in ADL performance, which differs from other studies (e.g., C.-M. Chen et al., 2012; Taylor & Lynch, 2004). Age per se was not a risk factor associated with a change in ADL score, although fluctuations and variations in ADL performance were more common among the oldest old (Bürge, von Gunten, & Berchtold, 2013). Looking at the cross-sectional distribution of latent regimes among age groups (Figure 5 in the supplemental material), residents below age 65 were likely to remain stable over time despite quite high levels of impairment. For this group, the Normative regime was estimated to be visited in 85.6% of the cases against the 71% observed for the overall sample. On the contrary, the Variability regime appeared to be relatively more common in residents aged 80 and older (3.5% of the cases).

We accounted for the correlation over time in ADL score including the two previous levels of ADLs as explanatory variables. The two autoregressive coefficients were significant in all four estimated regimes. This means that the previous assessments can be used to predict ADL performance in the following periods (see supplemental materials for a small illustration). Care plans can then be set in advance according to current ADL capacities.

Limitations of the Study

Potential limitations concern the data collection and the follow-up of the study. Information on reasons for dropping out of the study has not been collected. It might have been not only due to death of the subject but also because of a transfer to another care facility or refusal to do further assessments. Consequently, different from other longitudinal studies on disability (e.g., Han et al., 2013; Nusselder, Looman, & Mackenbach, 2006), we could not include mortality as a (final) state. The number of assessments also varies across residents, either for dropping out or because some residents experienced less changes in their health condition than others. Nevertheless, the attrition does not affect our findings. This study focused on the changes in disability trajectories at the individual level rather than the overall developmental pattern, so having different numbers of repeated measures by individuals has no impact on identifying different pathways of disability a person might go through. The assessments were not performed on a regular time basis (for instance, once a year), but whenever the health condition of the NH residents was changing. We took into account these different time gaps in our model by adding explicitly as an additional factor, the number of days between successive assessments. This factor was never significant, so we chose to not include it in the final model. The reason why the time between assessments was not significant is mainly due to the modeling strategy. The HTMD model used here controlled for time dependence both at a visible level (evolution of observed ADL scores) and latent level (switching between latent regimes during the assessment period). As such, the number of days between assessments may no longer become relevant in identifying change in ADL score. Further analysis should be conducted to better understand the effect of the time difference between assessments on the evolution of ADL capacity. NH residents have been assessed using the RAI-MDS. The RAI-MDS is a standard tool for monitoring residents in long-term care facilities, and its validity and reliability have been discussed in the literature (see, for example, Hawes et al., 1995; Lawton et al., 1998). However, we cannot exclude that levels of ADL performance have been measured without error. Finally, this study focused on cognitive and physical impairments as risk factors associated with a deterioration in ADL capacity. Medical records on acute and chronic illness and medications were not included in the analysis. Moreover, the secondary data used in this study did not include information on NH structural characteristics such as size and staffing pattern that might affect care quality (Bostick, 2004; Wang et al., 2010) as well as psychological (e.g., sense of loneliness, coping strategies) and sociological (e.g., quality of life, frequency of contacts with peers and family members) factors related to the vulnerability process that most people face in later life (Hanappi, Bernardi, & Spini, 2015).

Practice Implications

Our findings have important practical implications for health professionals working with the elderly. First, our study shows that small changes or fluctuations in ADL performance are warning factors associated with an increasing risk of experiencing a relevant decline in ADL capacity in the near future. Therefore, assessments and interventions aimed at identifying warning factors at an early stage may be very effective for the prevention of a fast deterioration of functional health. Second, our results also indicate that a deterioration in ADL is not always irreversible. An early detection of a decline in ADL performance should trigger interventions and treatments aimed at quickly bringing the patient back to his or her previous level of functional independence. Moreover, knowing that a decline in ADL is not irreversible could provide a psychological comfort to patients, helping them to overcome this decline by scrupulously following their treatments and the recommendations of the nursing staff. Finally, the development of specific programs to promote constant physical and/or intellectual exercises could be relevant in helping older people to maintain or even improve their ADL capabilities. A decline in physical functions have been identified as one of the major risk factors associated with a deterioration in ADL performance. A systematic review conducted by Forster et al. (2010) on the effects of physical rehabilitation for older people residing in long-term care facilities found positive but small size effects. However, the studies considered differed by target group, type of interventions, intensity, and, in particular, type of NH staff involved in delivering the intervention. Such relevant differences make the results difficult to compare. We hypothesize that the involvement of specialized health professionals such as physiotherapist or rehabilitation aide, as well as trained NH staff and caregivers, might have more effective and long-term effects on preventing functional deterioration in NH residents. Thus, there is the need for programs to train NH staff to deliver physical and intellectual interventions designed for the specific needs of people in long-term care. A longitudinal randomized case-control study might be necessary to quantify the cost effectiveness of such training programs.

Conclusion

Our findings show that the process of losing independence in ADL capacity is not irreversible or continuous, and that gradual and significant heterogeneity in disability trajectories in later life exists. Functional health of NH residents were most of time stable or slightly declining, but episodes of sudden change in ADL conditions were common, in particular among the oldest-old, followed again by periods of stability or recovery. A better understanding of the variability of disability condition and the associated risk factors can inform the development of practice and policies for an early identification of warning factors of increasing risk of worsening in functional limitation and to develop specific and effective programs helping people to maintain or even improve their ADL capabilities. Further studies to determine the effectiveness of such programs with different types of NH residents and in different settings are needed.

Supplementary Material

Supplementary Material, Suppl_Materials_JAH-17-299_20171103 – The Heterogeneity of Disability Trajectories in Later Life: Dynamics of Activities of Daily Living Performance Among Nursing Home Residents

Supplementary Material, Suppl_Materials_JAH-17-299_20171103 for The Heterogeneity of Disability Trajectories in Later Life: Dynamics of Activities of Daily Living Performance Among Nursing Home Residents by Danilo Bolano, PhD, André Berchtold, PhD, and Elisabeth Bürge, MPtSc in Journal of Aging and Health

Footnotes

Acknowledgements

The authors thank Prof. Armin von Gunten for providing the data used in this article.

Declaration of Conflicting Interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This publication benefited from the support of the Swiss National Centre of Competence in Research LIVES–Overcoming vulnerability: life course perspectives, which is financed by the Swiss National Science Foundation (grant number: 51NF40-160590).

Authors’ Contribution

D. Bolano performed the statistical analysis and drafted the original manuscript. A. Berchtold contributed to writing the manuscript and supervised the statistical analysis. E. Bürge contributed to conceptualizing the paper and provided critical feedbacks on the first version of the manuscript.

ORCID iD

Danilo Bolano

Supplemental Material

Supplemental material is available for this article online.

References

Asparouhov

Muthén

(2009). Exploratory structural equation modeling. Structural Equation Modeling: A Multidisciplinary Journal, 16, 397-438. doi:10.1080/10705510903008204

Banaszak-Holl

Liang

Quinones

Cigolle

Lee

I. C.

Verbrugge

L. M.

(2011). Trajectories of functional change among long stayers in nursing homes: Does baseline impairment matter? Journal of Aging and Health, 23, 862-882.

Beckett

L. A.

Brock

D. B.

Lemke

J. H.

Mendes

Leon

C. F.

Guralnik

J. M.

Fillenbaum

G. G.

. . . Evans

D. A.

(1996). Analysis of change in self-reported physical function among older persons in four population studies. American Journal of Epidemiology, 143, 766-778.

Bolano

Berchtold

(2016). General framework and model building in the class of Hidden Mixture Transition Distribution models. Computational Statistics & Data Analysis, 93, 131-145. doi:10.1016/j.csda.2014.09.011

Bollen

K. A.

Curran

P. J.

(2004). Autoregressive Latent Trajectory (ALT) Models: A synthesis of two traditions. Sociological Methods & Research, 32, 336-383. doi:10.1177/0049124103260222

Bollen

K. A.

Curran

P. J.

(2006). Latent curve models: A structural equation perspective. Hoboken, NJ: John Wiley.

Bostick

J. E.

(2004). Relationship of nursing personnel and nursing home care quality. Journal of Nursing Care Quality, 19, 130-136.

Bürge

Berchtold

von Gunten

(2011). Gender-related ADL performance of old people recently admitted to a Swiss nursing home. A cross-sectional study. Swiss Medical Weekly, 141, w13183. doi:10.4414/smw.2011.13183

Bürge

von Gunten

Berchtold

(2013). Factors favoring a degradation or an improvement in activities of daily living (ADL) performance among nursing home (NH) residents: A survival analysis. Archives of Gerontology and Geriatrics, 56, 250-257. doi:10.1016/j.archger.2012.09.001

10.

Burrows

Morris

J. N.

Simon

S. E.

Hirdes

J. P.

Phillips

(2000). Development of a minimum data set-based depression rating scale for use in nursing homes. Age and Ageing, 29, 165-172.

11.

Carpenter

G. I.

Hastie

C. L.

Morris

J. N.

Fries

B. E.

Ankri

(2006). Measuring change in activities of daily living in nursing home residents with moderate to severe cognitive impairment. BMC Geriatrics, 6, Article 7. doi:10.1186/1471-2318-6-7

12.

Chen

C.-M.

Mullan

Y.-Y.

Griffiths

Kreis

I. A.

Chiu

H.-C.

(2012). The longitudinal relationship between depressive symptoms and disability for older adults: A population-based study. The Journals of Gerontology, Series A: Biological Sciences & Medical Sciences, 67, 1059-1067. doi:10.1093/gerona/gls074

13.

Chen

J. H.

Chan

D. C.

Kiely

D. K.

Morris

J. N.

Mitchell

S. L.

(2007). Terminal trajectories of functional decline in the long-term care setting. The Journals of Gerontology, Series A: Biological Sciences & Medical Sciences, 62, 531-536.

14.

Chiang

C. L.

(1968). Introduction to stochastic process in biostatistics. Hoboken, NJ: John Wiley.

15.

Deeg

D. J.

(2005). Longitudinal characterization of course types of functional limitations. Disability and Rehabilitation, 27, 253-261. doi:10.1080/09638280400006507

16.

Edvardsson

Petersson

Sjogren

Lindkvist

Sandman

P. O.

(2014). Everyday activities for people with dementia in residential aged care: Associations with person-centredness and quality of life. International Journal of Older People Nursing, 9, 269-276. doi:10.1111/opn.12030

17.

Engelman

Jackson

(2017). Gradual change or punctuated equilibrium? Reconsidering patterns of health in later-life (CDE Working Paper No. 2017-01). Retrieved from https://www.ssc.wisc.edu/cde/cdewp/2017-01.pdf

18.

Forster

Lambley

Young

J. B.

(2010). Is physical rehabilitation for older people in long-term care effective? Findings from a systematic review. Age and Ageing, 39, 169-175.

19.

Gaugler

J. E.

Duval

Anderson

K. A.

Kane

R. L.

(2007). Predicting nursing home admission in the U.S.: A meta-analysis. BMC Geriatrics, 7, Article 13. doi:10.1186/1471-2318-7-13

20.

Gill

T. M.

Gahbauer

E. A.

Han

Allore

H. G.

(2010). Trajectories of disability in the last year of Life. The New England Journal of Medicine, 362, 1173-1180. doi:10.1056/NEJMoa0909087

21.

Gill

T. M.

Kurland

(2003). The burden and patterns of disability in activities of daily living among community-living older person. Journal of Gerontology, Series A: Medical Sciences, 58(1), 70-75.

22.

Guilley

Ghisletta

Armi

Berchtold

Epinay

C. L.

Michel

J.-P.

de Ribaupierre

(2008). Dynamics of frailty and ADL dependence in a five-year longitudinal study of octogenarians. Research on Aging, 30, 299-317. doi:10.1177/0164027507312115

23.

Hamilton

J. D.

(1989). A new approach to the economic analysis of nonstationary time series and the business cycle. Econometrica, 57, 357-384.

24.

Han

Allore

Murphy

Gill

Peduzzi

Lin

(2013). Dynamics of functional aging based on latent-class trajectories of activities of daily living. Annals of Epidemiology, 23(2), 87-92. doi:10.1016/j.annepidem.2012.11.010.Dynamics

25.

Hanappi

Bernardi

Spini

(2015). Vulnerability as a heuristic for interdisciplinary research: Assessing the thematic and methodological structure of empirical life-course studies. Longitudinal and Life Course Studies, 6(11), 59-87. doi:10.14301/llcs.v6i1.302

26.

Hardy

S. E.

Dubin

J. A.

Holford

T. R.

Gill

T. M.

(2005). Transitions between states of disability and independence among older persons. American Journal of Epidemiology, 161, 575-584. doi:10.1093/aje/kwi083

27.

Hawes

Morris

J. N.

Phillips

C. D.

Mor

Fries

B. E.

Nonemaker

(1995). Reliability estimates for the Minimum Data Set for nursing home resident assessment and care screening (MDS). Gerontologist, 35, 172-178.

28.

Helvik

A.S.

Engedal

Benth

J.S.

Selbaek

(2015). Prevalence and Severity of Dementia in Nursing Homre Residents. Dementia and Geriatric Cognitive Disorders, 40, 166-077. doi: 10.1159/000433525

29.

Jacobs

J. M.

Maaravi

Cohen

Bursztyn

Ein-Mor

Stessman

(2012). Changing profile of health and function from age 70 to 85 years. Gerontology, 58, 313-321. doi:10.1159/000335238

30.

Jette

Branch

Sleeper

Feldman

Sullivan

(1992). High-risk profiles for nursing home admission. Gerontologist, 32(5), 167-174.

31.

Kruse

R. L.

Petroski

G. F.

Mehr

D. R.

Banaszah-Holl

Intrator

(2013). Activity of daily living trajectories surrounding acute hospitalization of long-stay nursing home residents. Journal of the American Geriatrics Society, 61, 1909-1918.

32.

Lahmann

Tannen

Kuntz

Raeder

Schmitz

Dassen

. . . Kottner

(2015). Mobility is the key! Trends and associations of common care problems in German long-term care facilities from 2008 to 2012. International Journal of Nursing Studies, 52, 167-174. doi:10.1016/j.ijnurstu.2014.07.014

33.

Lawton

M. P.

Casten

Parmelee

P. A.

Van Haitsma

Corn

Kleban

M. H.

(1998). Psychometric characteristics of the minimum data set II: Validity. Journal of the American Geriatrics Society, 46, 736-744.

34.

Liang

Bennett

J. M.

Quiñones

A. R.

(2009). Ethnicity and changing functional health in middle and late life: A person-centered approach. The Journals of Gerontology, Series B, 65, 470-481. doi:10.1093/geronb/gbp114

35.

Lindholm

Gustavsson

A. J.

Jönsson

Wimo

(2013). Costs explained by function rather than diagnosis—Results from the SNAC Nordanstig elderly cohort in Sweden. International Journal of Geriatric Psychiatry, 28, 454-462. doi:10.1002/gps.3844

36.

Lynch

S. M.

Taylor

M. G.

(2016). Trajectory models for aging research. In George

L. K.

Ferraro

K. F.

(Eds.), Handbook of aging and the social sciences (pp. 23-51). San Diego, CA: Academic Press.

37.

McConnell

E. S.

Pieper

C. F.

Sloane

R. J.

Branch

L. G.

(2002). Effects of cognitive performance on change in physical function in long-stay nursing home residents. The Journals of Gerontology, Series A: Biological Sciences & Medical Sciences, 57(12), M778-M784.

38.

Morris

J. N.

Fries

B. E.

Mehr

D. R.

Hawes

Phillips

Mor

Lipsitz

L. A.

(1994). MDS Cognitive Performance Scale. Journal of Gerontology, 49(4), M174-M182. doi:10.1093/geronj/49.4.M174

39.

Murphy

O’Shea

Cooney

(2007). Quality of life for older people living in long-stay settings in Ireland. Journal of Clinical Nursing, 16, 2167-2177.

40.

Nagin

D. S.

(1999). Analyzing developmental trajectories: A semiparametric, group-based approach. Psychological Methods, 4, 139-157.

41.

Nusselder

W. J.

Looman

W. N.

Mackenbach

J. P.

(2006). The level and time course of disability: Trajectories of disability in adults and young elderly. Disability and Rehabilitation, 28, 1015-1026. doi:10.1080/09638280500493803

42.

Palese

Menegazzi

Tullio

Zigotti Fuso

Hayter

Watson

(2016). Functional decline in residents living in nursing homes: A systematic review of the literature. Journal of the American Medical Directors Association, 17, 694-705.

43.

Paterson

D. H.

Warburton

D. E.

(2010). Physical activities and functional limitations in older adults: A systematic review related to Canada’s Physical Activity Guidelines. International Journal of Behavioral Nutrition and Physical Activity, 7, Article 38.

44.

Peri

Kerse

Robinson

Parsons

Latham

(2008). Does functionally based activity make a difference to health status and mobility? A randomised controlled trial in residential care facilities (The Promoting Independent Living Study; PILS). Age and Ageing, 37, 57-63. doi:10.1093/ageing/afm135

45.

Schwarz

G. E.

(1978). Estimating the dimension of a model. Annals of Statistics, 6, 461-464.

46.

Scott

S. L.

James

G. M.

Sugar

C. A.

(2005). Hidden Markov models for longitudinal comparisons. Journal of the American Statistical Association, 100, 359-369. doi:10.1198/016214504000001592

47.

Slaughter

S. E.

Eliasziw

Morgan

Drummond

(2010). Incidence and predictors of excess disability in walking among nursing home residents with middle stage of dementia: A prospective cohort study. International Psychogeriatrics, 23, 54-64. doi:10.1017/S1041610210000116

48.

Slaughter

S. E.

Wagg

A. S.

Jones

C. A.

Schopflocher

Ickert

Bampton

. . . Estabrooks

C. A.

(2015). Mobility of vulnerable elders study: Effect of the sit-to-stand activity on mobility, function, and quality of life. Journal of the American Medical Directors Association, 16, 138-143. doi:10.1016/j.jamda.2014.07.020

49.

Tak

Kuiper

Chorus

Hopman-Rock

(2013). Prevention of onset and progression of basic ADL disability by physical activity in community dwelling older adults: A meta-analysis. Ageing Research Reviews, 12, 329-338.

50.

Taylor

M. G.

Lynch

S. M.

(2004). Trajectories of impairment, social support, and depressive symptoms in later life. The Journals of Gerontology, Series B: Psychological Sciences & Social Sciences, 59, 238-246. doi:10.1093/geronb/59.4.S238

51.

Van de Pol

de Leeuw

(1986). A latent Markov model to correct for measurement error. Sociological Methods & Research, 15(1-2), 118-147.

52.

Verbrugge

L. M.

Jette

A. M.

(1994). The disablement process. Social Science & Medicine, 38(1), 1-14.

53.

Vermeulen

Neyens

J. C.

van Rossum

Spreeuwenberg

M. D.

de Witte

L. P.

(2011). Predicting ADL disability in community-dwelling elderly people using physical frailty indicators: A systematic review. BMC Geriatrics, 11, Article 33. doi:10.1186/1471-2318-11-33

54.

Wan

T. T.

Zhang

N. J.

Unruh

(2006). Predictors of resident outcome improvement in nursing homes. Western Journal of Nursing Research, 28, 974-993.

55.

Wang

Chang

L. H.

Eberly

L. E.

Virnig

B. A.

Kane

R. L.

(2010). Cognition moderates the relationship between facility characteristics, personal impairments, and nursing home residents’ activities of daily living. Journal of American Geriatrics Society, 58, 2275-2283.

56.

Wiggins

L. M.

(1973). Panel analysis: Latent probability models for attitude and behavior processes. Amsterdam, The Netherlands: Elsevier Scientific Publishing Company.

57.

Zahodne

L. B.

Stern

Manly

J. J.

(2014). Depressive symptoms precede memory decline, but not vice versa, in non-demented older adults. Journal of the American Geriatrics Society, 62, 130-134. doi:10.1111/jgs.12600

58.

Zimmer

Martin

L. G.

Jones

B. L.

Nagin

D. S.

(2014). Estimating late-life functional limitation trajectories and their associations with underlying onset, recovery, and mortality. Journals of Gerontology, Series B: Psychological Sciences & Social Sciences, 692(2), 275-286. doi:10.1093/geronb/gbt099

59.

Zucchini

MacDonald

I. L.

(2009). Hidden Markov models for time series. An introduction using R. New York, NY: Chapman & Hall.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.39 MB