Factors Associated With Recovery of Prehospital Function Among Older Persons Admitted to a Nursing Home With Disability After an Acute Hospitalization

Abstract

Background

The objective of this study was to identify the factors associated with recovery of prehospital function among older persons admitted to a nursing home with disability after an acute hospitalization.

Methods

The analytic sample included 292 participants of an ongoing cohort study who had one or more admissions to a nursing home with disability after an acute hospitalization during nearly 10 years of follow-up, yielding a total of 364 “index” nursing home admissions. Information on nursing home admissions, hospitalizations, and disability in essential activities of daily living was ascertained during monthly telephone interviews. Data on potential predictors of functional recovery were collected during comprehensive assessments, which were completed every 18 months for 90 months. Participants were considered to have recovered if they were discharged home within 6 months of their nursing home admission at (or above) their prehospital level of function.

Results

Recovery of prehospital function was observed for 115 (31.6%) of the 364 index nursing home admissions. In the multivariate analysis, the strongest associations were observed for the best category of performance, relative to the poorest category, for gross motor coordination (hazard ratio [HR] 13.5, 95% confidence interval [CI] 4.02–45.0) and manual dexterity (HR 10.0, 95% CI 2.94–34.3). Only two other factors were independently associated with recovery of prehospital function: not cognitively impaired (HR 3.0, 95% CI 1.46–6.14) and no significant weight loss (HR 1.96, 95% CI 1.06–3.63).

Conclusions

In the setting of an acute hospitalization leading to a nursing home admission with disability, the likelihood of recovering prehospital function is low. The factors associated with recovery include faster performance on tests of gross motor coordination and manual dexterity and the absence of cognitive impairment and significant weight loss.

ACUTE illnesses and injuries leading to hospitalization are potent precipitants of functional decline and disability among community-living older persons (1–4). Following an acute hospitalization, nearly 50% of newly disabled older persons are transferred to a nursing home (1), usually for Medicare-reimbursed subacute care (5).

Subacute care in a nursing home is designed to optimize functional status and address ongoing medical conditions. In a recently completed study (6), we found that the functional trajectories among older persons admitted to a nursing home with disability after an acute hospitalization are generally poor, with only about a third of participants returning home at their prehospital level of function. Why some of these disabled older persons recover and others do not is not known. In prior studies of functional recovery, salient predictors have included younger age, fewer depressive symptoms, higher education, physical capacity, cognitive status, social support, and habitual physical activity (7–9).

In the current study, we set out to identify the factors associated with recovery of prehospital function among older persons admitted to a nursing home with disability after an acute hospitalization. We used data from a unique longitudinal study that includes monthly assessments of disability for up to 9.8 years, along with nearly complete ascertainment of nursing home admissions and hospitalizations, on a large cohort of community-living older persons.

METHODS

Study Population

Participants were members of the Precipitating Events Project, a longitudinal study of 754 community-living persons, aged 70 years or older, who were nondisabled (i.e., required no personal assistance) at baseline in four essential activities of daily living—bathing, dressing, walking inside one's home, and transferring from a chair (10). Exclusion criteria included significant cognitive impairment with no available proxy (11), inability to speak English, diagnosis of a terminal illness with a life expectancy less than 12 months, and a plan to move out of the New Haven area during the next 12 months.

The cohort was assembled between March 1998 and October 1999; complete details have been provided elsewhere (10,12). Eligibility was determined during a screening telephone interview and was confirmed during an in-home assessment. Persons who were physically frail, as denoted by a timed score of greater than 10 seconds on the rapid gait test (i.e., walk back and forth over a 10-foot [3-m] course as quickly as possible), were oversampled to ensure a sufficient number of participants at increased risk for disability (13,14). Only 4.6% of the 2,753 health plan members who were alive and could be contacted refused to complete the screening telephone interview, and 75.2% of the 1,002 eligible members agreed to participate in the project. Persons who refused to participate did not differ significantly from those who were enrolled in terms of age or sex. The study protocol was approved by the Yale Human Investigation Committee, and all participants provided verbal informed consent.

Data Collection

Data on potential predictors of functional recovery were collected during comprehensive home-based assessments, which were completed by trained nurse researchers at baseline and every 18 months for 90 months. Telephone interviews were completed monthly for up to 9.8 years, with a completion rate of 99%.

Potential predictors.—

In addition to demographic factors, we considered potential predictors from four domains that have been linked to functional recovery in prior studies (7–9). The demographic factors included age, sex, race/ethnicity, living situation, and education. The health-related factors included nine self-reported, physician-diagnosed chronic conditions; number of prescription medications; corrected near vision, assessed with a Jaeger card and calculated as the percentage of visual impairment (15); hearing, as assessed with a handheld audioscope (16); and self-report of a 10-pound weight loss in the past year (17). The cognitive–psychosocial factors included cognitive status, as assessed by the Folstein Mini-Mental State Examination (MMSE) (18); depressive symptoms, as assessed by the Center for Epidemiological Studies-Depression scale (19); and social support, as assessed by a modified version of the Medical Outcomes Survey Social Support Survey (8). The habitual factors included smoking status (current vs other); physical activity, as assessed by the Physical Activity Scale for the Elderly (20); and body mass index (BMI) based on self-reported height and weight. Although assessed at baseline and 18 months, alcohol use was not assessed thereafter and, hence, was not evaluated as a potential predictor.

The physical capacity factors included slow gait speed, as previously defined; qualitative gait and balance, as assessed by a modified version of the Performance-Oriented Mobility Assessment (21); the ability to rise from a standard chair in a single attempt with arms folded (13); three-timed chair stands (13); manual dexterity, as assessed by a pegboard test (22); gross motor coordination, as assessed by having the participant alternatively tap his/her index finger between two circles on a paper and his/her nose 10 times (23); grip strength (dominant hand), as assessed by the average of three readings using a Jamar hydraulic handheld dynamometer (17); and a modified version of the Short Physical Performance Battery (SPPB) (24) that included the standard balance maneuvers but substituted three-timed chair stands (instead of five) and timed rapid gait (back and forth over a 10-foot course) instead of timed usual gait (over a 4-m course without a turn). Quartile cut-points for these latter two tasks were determined from the first 356 enrolled participants, who had been selected randomly from our source population (10). As per convention (24), scores for the balance, chair stands, and gait tasks ranged from 0 to 4, with 0 indicating the inability to complete the task and 4 the highest level of performance; and a composite SPPB score (range 0–12) was calculated by adding the three scores. For all potential predictors, the amount of missing data was less than 1% in the baseline assessment and less than 5% in all subsequent assessments. Additional operational details are provided in Table 1.

Table 1.

Characteristics and Bivariate Associations of Potential Predictors of Recovery of Prehospital Function Among 292 Community-Living Persons Admitted to a Nursing Home With Disability After an Acute Hospitalization

Characteristic	Operational Details	n (%) or Mean ± SD*	Hazard Ratio†	95% CI	p Value
Demographic
Age <85 y		174 (59.6)	1.54	0.87–2.73	.14
Male		105 (36.0)	1.56	0.86–2.81	.14
Non-Hispanic white		260 (89.0)	0.84	0.33–2.19	.73
Lives with others		144 (49.3)	1.08	0.62–1.87	.80
High school education or greater		184 (63.0)	1.83	1.02–3.25	.04
Health related
Chronic conditions	9 self-reported, physician diagnosed‡	2.4 ± 1.3	1.04	0.87–1.24	.70
Prescription medications		5.7 ± 2.9	1.03	0.94–1.13	.55
Visual impairment	%, assessed with a Jaeger card
None or mild	<6	143 (49.0)	1.21	0.66–2.21	.55
Moderate	6–26	55 (18.8)	0.81	0.38–1.71	.57
Severe	>26	94 (32.2)	Reference
Hearing impairment	Tones missed out of 4§
None or mild	0–1	76 (26.0)	1.46	0.67–3.17	.34
Moderate	2–3	109 (37.3)	1.31	0.68–2.52	.42
Severe	4	107 (36.6)	Reference
No significant weight loss	≤10 pounds in past y	198 (67.8)	1.91	1.06–3.44	.03
Cognitive–psychosocial
Not cognitively impaired	Score on Folstein MMSE ≥24	232 (79.5)	3.44	1.69–7.02	.001
Not depressed	Score on CES-D <20	215 (73.6)	1.38	0.74–2.56	.31
Social support	MOS: 0 (low) to 28 (high)	21.1 ± 6.0	1.01	0.96–1.06	.82
Habitual
Not current smoker		268 (91.8)	1.07	0.41–2.79	.89
Physical activity not low	PASE score ≥64 for men and ≥52 for women‖	91 (31.2)	1.48	0.81–2.71	.20
Body mass index	Self-reported height and weight¶, kg/m2
Normal or underweight	<25	138 (47.3)	Reference
Overweight	25–29.9	91 (31.2)	1.11	0.57–2.17	.75
Obese	≥30	63 (21.6)	2.18	1.13–4.22	.02
Physical capacity#
Gait speed not slow	≤10 s on rapid gait test	198 (67.8)	1.42	0.71–2.86	.32
Gait and balance	Modified POMA: 0 (poor) to 12 (good)**	5.8 ± 1.7	1.24	1.05–1.46	.01
Able to rise from chair	In single attempt with arms folded	138 (47.3)	1.22	0.69–2.17	.50
Timed chair stands††, s	Up and down three times
<7.6		20 (6.9)	6.04	2.23–16.3	<.001
7.6–9.0		33 (11.3)	2.22	0.86–5.77	.10
9.1–11.9		77 (26.4)	2.11	1.13–3.97	.02
≥12.0		162 (55.5)	Reference
SPPB	0 (low) to 12 (high)	4.7 ± 2.5	1.22	1.08–1.39	.002
Manual dexterity††, s	9-hole pegboard test
<21.8		19 (6.5)	26.1	9.36–73.0	<.001
21.8–24.3		35 (12.0)	4.04	1.62–10.0	.003
24.4–27.5		50 (17.1)	1.94	0.90–4.19	.09
≥27.6		188 (64.4)	Reference
Gross motor coordination††, s	10 finger taps as described in the text
<8.8		27 (9.3)	25.8	9.20–72.1	<.001
8.8–10.3		47 (16.1)	2.96	1.37–6.38	.01
10.4–12.4		74 (25.3)	1.96	0.97–3.97	.06
≥12.5		144 (49.3)	Reference
Grip strength not poor‡‡	Handheld dynamometer	74 (25.3)	0.96	0.49–1.85	.89

Notes: CES-D = Center for Epidemiological Studies-Depression scale; CI = confidence interval; MMSE = Mini-Mental State Examination; MOS = Medical Outcomes Survey Social Support Scale; PASE = Physical Activity Scale for the Elderly; POMA = Performance-Oriented Mobility Assessment; SPPB = Short Physical Performance Battery.

^*Values are provided from the comprehensive assessment that was completed immediately prior to the first index nursing home admission.

^†Hazard ratios were estimated using recurrent event Cox regression counting process models, adjusted for the number of months from the preceding comprehensive assessment to the index nursing home admission. Among the 292 participants, there were 364 index nursing home admissions. Hazard ratios for continuous variables are per one unit increase.

^‡Hypertension, myocardial infarction, heart failure, stroke, diabetes mellitus, arthritis, hip fracture, chronic lung disease, and cancer (other than minor skin cancers).

^§Based on 1000 and 2000 Hz measurements for the left and right ears.

^‖These gender-specific cut-points denote the worse quintile of scores among the first 356 enrolled participants, who had been selected randomly from our source population of health plan members, as previously described (17).

^¶Body mass index was categorized into three groups based on published cut-points, as previously described (25).

^#Participants were asked to perform each of the timed tests as quickly (and safely) as possible.

^**Includes 5 gait items—step continuity and symmetry, path deviation, turning, and missed steps—and the three standard tasks of standing balance from the Established Population for Epidemiologic Studies of the Elderly battery, as previously described (21).

^††Because standard cut-points have not been established for these timed tests, quartile scores were calculated based on the first 356 enrolled participants and subsequently applied to the entire cohort.

^‡‡Value less than or equal to the gender- and body mass index–specific cut-points provided by Fried and coworkers (26).

Assessment of disability.—

Complete details regarding the assessment of disability, including formal tests of reliability and accuracy, are provided elsewhere (11). During the monthly telephone interviews, participants were assessed for disability using standard questions that were identical to those used during the screening telephone interview (11). For each of the four essential activities of daily living, we asked, “At the present time, do you need help from another person to (complete the task)?” Participants who needed help with any of the tasks were considered to be disabled. Given the frequency of assessments, it was not feasible to ascertain less severe forms of disability, including the use of assistive equipment or degree of difficulty. The reliability of our disability assessment was substantial (kappa = 0.75) for reassessments completed within 48 hours and excellent (kappa = 1.0) for reassessments performed the same day (11). For participants with significant cognitive impairment, the monthly telephone interviews were completed with a designated proxy. The accuracy of these proxy reports was also found to be excellent, with kappa = 1.0 (11). As described in an earlier report (27), we used multiple imputation with 100 random draws per missing observation to address the small amount of missing monthly data on disability.

Ascertainment of hospitalizations and nursing home admissions.—

Information on hospitalizations and nursing home admissions was obtained during the monthly telephone interviews. Participants were asked whether they had stayed at least overnight in a hospital since the last interview, that is, during the past month. The accuracy of these reports was high, with kappa = 0.94 (28). Participants who were hospitalized were asked to provide the primary reason for their admission. These reasons were subsequently grouped into distinct diagnostic categories using a revised version of the protocol described by Ferrucci and coworkers (1,4). Percent agreement, relative to an independent review of hospital records, was 82%.

Participants were also asked whether they had been admitted to a nursing home during the past month; if yes, the interviewer noted whether the participant was currently in a nursing home. The accuracy of this information was almost perfect, with kappa = 0.96 (25).

Assembly of Analytic Sample

To be included, participants had to be newly admitted to a nursing home with disability after an acute hospitalization, as determined during the first monthly interview that was completed during the nursing home stay (6). In addition, participants had to be living in the community during the preceding comprehensive assessment since data on the potential predictors were collected at this time. Of the 754 participants, 296 (39.3%) had a nursing home admission with disability after an acute hospitalization. Of these, 3 participants were residents of a nursing home during the preceding comprehensive assessment and 1 subsequently dropped out of the study, leaving 292 (38.7%) participants in the analytic sample.

Recovery of Prehospital Function

Participants were considered to have recovered if they were discharged home within 6 months of their nursing home admission at (or above) their prehospital level of function. This was determined by comparing the number of disabled activities of daily living during the first monthly interview that was completed after discharge from the nursing home with that during the monthly interview that was completed immediately prior to the hospitalization. Of note, only 3 (1.0%) participants were discharged home after 6 months in the nursing home; and, of these, none had regained their prehospital function.

Statistical Analysis

To make full use of the longitudinal data and enhance power, we allowed participants to contribute more than one nursing home admission to the analysis. If a participant had more than one qualifying admission within an 18-month interval, only the first was selected because the candidate predictors could have subsequently changed. Among the 292 participants in the analytic sample, 51 had two qualifying admissions, whereas 6 had three and 3 had four, yielding a total of 364 qualifying admissions. To enhance clarity, these qualifying admissions will subsequently be referred to as “index” nursing home admissions. Prior work suggests that most of these admissions included Medicare-reimbursed subacute care with rehabilitation services (5).

We determined the characteristics of participants in the analytic sample, using counts (percentages) and means (± standard deviations [SDs]), as indicated; and we calculated the frequency distributions of the primary reasons for the acute hospitalizations leading to the index nursing home admissions. Next, we calculated the median time to the index nursing home admissions from the preceding comprehensive assessment, the mean number of disabled activities of daily living immediately prior to hospitalization and upon admission to the nursing home, respectively, and the percentage of admissions resulting in recovery of prehospital function.

To evaluate the bivariate and multivariate relationships between the candidate predictors and functional recovery, we used a recurrent event Cox regression counting process model (29), adjusting for the number of months from the preceding comprehensive assessment to the index nursing home admission. This counting process method accounts for the effects of chronological time and the sequence of the recurrent outcome. No interactions were found between the candidate predictors and chronological time, that is, the effects of the predictors were stable over the follow-up period, or between the candidate predictors and the sequence of the recurrent outcome, that is, the effects of the predictors were stable over the first, second, and so forth, occurrence of functional recovery. To account for the correlation among observations within individuals, we used the robust sandwich variance estimators for standard errors of the coefficients (30). The exact method was used to handle tied outcome times, that is, when two or more participants left the nursing home in the same month (31). To create a parsimonious model, we used a hierarchical selection process (32). First, we evaluated the bivariate association between each candidate predictor and functional recovery. Because the candidate predictors, other than sex, race/ethnicity, and education, could change over time, the bivariate and multivariate models used time-dependent variables. Only factors with a p value less than or equal to .30 were considered further. Next, we sequentially evaluated the nonparametric Spearman's rank order correlations (33) among the remaining factors, first within each of the previously described domains and then overall. When the correlation coefficient was greater than 0.35, denoting potential collinearity, we chose the factor that was most explanatory, as determined by the smallest −2 log likelihood statistic (−2LL). We then sequentially evaluated the impact of each of the remaining factors on the overall model fit through a series of recurrent event Cox models (29).

To assess each factor's contribution to the model fit, we used a chi-square distribution with degree of freedoms equaling the number of parameters for the added factor, based on the difference in the −2LL between the models with and without the factor. After fitting a separate model for each factor, we added the factor leading to the greatest reduction in the −2LL to the overall model. We continued this process iteratively until no factor significantly improved the overall model fit based on the −2LL criterion. To determine the effect of prehospital functional status on the risk estimates, we reran the final model after adding the number of disabled activities of daily living during the monthly interview that was completed immediately prior to the hospitalization. To address the possibility of a floor effect, we reran the final model again after omitting the 13 observations in which disability was present in all four activities of daily living (i.e., complete disability) immediately prior to the hospitalization.

All statistical tests were two tailed, and p < .05 was considered to indicate statistical significance. All analyses were performed using SAS version 9.1.3 (SAS Institute, Cary, NC).

RESULTS

Information on the characteristics of participants in the analytic sample is provided in Table 1 (first data column). During the comprehensive assessment that immediately preceded the first index nursing home admission, about 60% of participants were younger than 85 years, about a third were men, nearly half lived with others, and about two thirds had at least a high school education. On average, participants had more than two chronic conditions and were taking about six prescription medications.

The most common reasons for the acute hospitalizations leading to the index nursing home admissions were fall-related injury (18.1%), infection (16.8%), arthritis (gout, hip and knee replacement, spinal stenosis, etc.; 10.7%), cardiac (coronary heart disease, congestive heart failure, arrhythmia, etc.; 10.2%), stroke (5.5%), and other medical conditions (25.8%). The median time to the index nursing home admission from the preceding comprehensive assessment was 10 (interquartile range [IQR] 5–14) months. The mean number of disabled activities of daily living was 0.7 (SD 1.1) immediately prior to the hospitalization and 3.0 (SD 1.2) upon admission to the nursing home. The distribution of the number of disabled activities of daily living immediately prior to hospitalization was none (60.7%), 1 (17.6%), 2 (14.3%), 3 (3.9%), and 4 (3.6%). The median duration of the index nursing home admissions was 2 months (IQR 1–4).

Recovery of prehospital function was observed for 115 (31.6%) of the 364 index nursing home admissions. The bivariate associations of the potential predictors of recovery are provided in Table 1 as hazard ratios (HRs) accompanied by 95% confidence intervals and p values. For each of the first four domains, only one factor was statistically associated with recovery at p < .05: education (demographic), no significant weight loss (health-related), not cognitively impaired (cognitive–psychosocial), and BMI/obese (habitual). In contrast, for the physical capacity domain, several factors were statistically associated with recovery, including gait and balance, timed chair stands, SPPB, manual dexterity, and gross motor coordination. Of these, gait and balance and the SPPB were omitted from the multivariate analyses to reduce potential collinearity, as described in the Methods section.

The results of the multivariate analyses are provided in Table 2. Four factors were retained in the final model: gross motor coordination, manual dexterity, not cognitively impaired, and no significant weight loss. The strongest associations were observed for the best category of performance, relative to the poorest category, for gross motor coordination and manual dexterity, respectively. The results did not change substantively after prehospital functional status was added (model 1) or after omission of the 13 observations with complete disability immediately prior to the hospitalization (model 2), although the HRs were generally a bit higher than for the model without prehospital functional status.

Table 2.

Factors Associated in Multivariate Analysis With Recovery of Prehospital Function Among Participants Admitted to a Nursing Home With Disability After an Acute Hospitalization^*

	Model Without Prehospital Functional Status			Model 1 With Prehospital Functional Status†			Model 2 With Prehospital Functional Status‡
Factor‡	Hazard Ratio	95% CI	p Value	Hazard Ratio	95% CI	p Value	Hazard Ratio	95% CI	p Value
Gross motor coordination, s
<8.8	13.5	4.02–45.0	<.001	14.4	3.95–52.2	<.001	14.5	3.93–53.7	<.001
8.8–10.3	2.23	1.06–4.71	.04	2.65	1.27–5.52	.01	2.83	1.34–5.96	.01
10.4–12.4	1.82	0.92–3.60	.08	1.89	0.94–3.80	.07	2.09	1.03–4.26	.04
≥12.5	Reference			Reference			Reference
Manual dexterity, s
<21.8	10.0	2.94–34.3	<.001	14.0	3.81–51.3	<.001	13.0	3.53–47.8	<.001
21.8–24.3	1.95	0.73–5.23	.19	2.20	0.79–6.11	.13	2.20	0.78–6.21	.14
24.4–27.5	1.58	0.75–3.33	.23	1.62	0.78–3.38	.20	1.69	0.81–3.52	.16
≥27.6	Reference			Reference			Reference
Not cognitively impaired	3.00	1.46–6.14	.003	3.43	1.62–7.29	.001	4.10	1.82–9.27	.001
No significant weight loss	1.96	1.06–3.63	.03	1.86	1.00–3.46	.05	1.93	1.03–3.62	.04

Notes: CI = confidence interval.

^*Hazard ratios were estimated using recurrent event Cox regression counting process models, adjusted for the number of months from the preceding comprehensive assessment to the index nursing home admission. Prehospital functional status was defined as the number of disabled activities of daily living during the monthly interview immediately prior to the hospitalization. Among the 292 participants, there were 364 index nursing home admissions.

^†Model 1 includes all 364 index nursing home admissions, whereas model 2 omits the 13 admissions in which disability was present in all four activities of daily living (i.e., complete disability) immediately prior to the hospitalization.

^‡Factors are listed in the order they entered the multivariate models following a forward selection procedure based on the likelihood ratio test between two successively fitted models, as described in the Methods section. Age, sex, education, physical activity, body mass index, and timed chair stands were evaluated in the multivariate analysis but did not meet the criterion for inclusion in the final model.

DISCUSSION

In this prospective cohort study, which included monthly assessments of disability for nearly 10 years, we found that faster performance on tests of gross motor coordination and manual dexterity and the absence of cognitive impairment and significant weight loss were independently associated with recovery of prehospital function among older persons admitted to a nursing home with disability after an acute hospitalization. Because the tests of gross motor coordination and manual dexterity largely assess psychomotor speed (34,35), our results highlight the prognostic importance of cognition to recovery of prehospital function after a serious disabling event.

Prior studies of discrete disabling events, such as stroke or hip fracture, have also found strong associations between cognition and the likelihood of functional recovery (36–39). Although we were unable to determine the precise cause of our participants’ disability, the most common diagnostic categories for hospitalization leading to the index nursing home admissions were fall-related injury, infection, and arthritis, which included joint replacements (among other conditions). Cognition has previously been linked to the likelihood of functional recovery after hospitalization for a diverse array of conditions (9,40). Prior studies, however, have typically evaluated only measures of global cognition, such as the MMSE, or a diagnosis of dementia as potential predictors of functional recovery (8,9,36–40). To our knowledge, few other studies of functional recovery have evaluated measures of psychomotor speed.

Psychomotor slowing has been observed in persons with depression (41) and dementia (42,43) and may reflect damage to precortical neurons (44). In the current study, depression was not associated with the likelihood of functional recovery, and the effect of psychomotor speed persisted in models that included a measure of global cognitive function. Whether psychomotor speed is modifiable is uncertain but should be the focus of future studies.

As one of the best indicators of an underlying disease process (45), weight loss is a well-established risk factor for the onset and progression of disability in older persons (46). Few prior studies, however, have evaluated the effect of weight loss on the likelihood of functional recovery. Weight loss often leads to protein energy malnutrition, which, in turn, is associated with an array of adverse consequences (45,47,48). Depending on its etiology, weight loss may or may not be amenable to intervention (49).

Our study has several strengths. First, functional status was reassessed at 1-month intervals. This allowed us to accurately determine the onset of disability and reduces the possibility that episodes of recovery followed by recurrent disability and/or death were missed (6), a phenomenon that occurs commonly in studies with longer assessment intervals (8,11,50). Secondly, the extended duration of follow-up allowed us to identify a large number of index nursing home admissions, which enhanced our power to detect clinically significant HRs for recovery of prehospital function. Thirdly, potential predictors of recovery were ascertained prior to the onset of disability, ensuring that these factors reflected participants’ prehospital status. Finally, the nearly complete ascertainment of nursing home admissions, hospitalizations, and disability, coupled with the high reliability and accuracy of these assessments and the low rate of attrition, greatly enhance the validity of our results.

Despite these strengths, several limitations warrant comment. First, although we updated information on the potential predictors every 18 months, it is possible that some of the predictors changed over shorter periods of time, leading to some misclassification of risk. Secondly, the associations found in our epidemiological study support but do not prove causation. Randomized trials will be needed to determine if interventions directed at the identified predictors can alter the likelihood of recovery. Thirdly, our study did not include data on the receipt of rehabilitation. However, most new admissions to a nursing home following an acute hospitalization among disabled older persons include Medicare-reimbursed subacute care with rehabilitation services (5). Furthermore, although the benefit of rehabilitation after a stroke has been well established, there is relatively little evidence to support the effectiveness of rehabilitation after many other disabling events and essentially no evidence in a subacute care setting (51). Fourthly, although information was available on the reason for hospitalization, data were not available to classify the severity of hospitalization, which is strongly associated with functional outcomes (2). Fifthly, because our participants were members of a single health plan in a small urban area, the generalizability of our results might be questioned. The high participation rate, which was greater than 75%, enhances the generalizability of our results (52). Moreover, our study population reflects the demographic characteristics of older persons in New Haven County, which are comparable to the United States as a whole, with the exception of race (New Haven County has a larger proportion of non-Hispanic whites in this age group than the United States, 91% vs 84%) (53). Although we initially oversampled persons who were physically frail, we have previously demonstrated that the distribution of functional trajectories did not differ significantly on the basis of physical frailty (6). Finally, utilization of subacute rehabilitation has varied over the course of our study (5) and may differ in our cohort from other populations of older persons. Although these variations and differences would influence estimates for the incidence of subacute nursing home admissions, they would not necessarily affect estimates for associations between candidate predictors and functional recovery, which was the focus of our study.

In summary, in the setting of an acute hospitalization leading to admission to a nursing home with disability, the likelihood of recovering prehospital function is low. The factors associated with recovery include faster performance on tests of gross motor coordination and manual dexterity and the absence of cognitive impairment and significant weight loss. Assessment of these factors may help clinicians to better focus their efforts in promoting functional recovery after a major disabling event.

FUNDING

The work for this report was funded by grants from the National Institute on Aging (R37AG17560, R01AG022993). The study was conducted at the Yale Claude D. Pepper Older Americans Independence Center (P30AG21342). T.M.G. is the recipient of a Midcareer Investigator Award in Patient-Oriented Research (K24AG021507) from the National Institute on Aging.