| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
|
|
|
|||||||
|
|||||||||
1 University of Kansas Medical Center, Kansas City.
2 University of Pittsburgh and Pittsburgh VA Healthcare System, Pennsylvania.
3 Merck Research Laboratories, Blue Bell, Pennsylvania.
4 National Institute on Aging, Bethesda, Maryland.
Address correspondence to Stephanie Studenski, MD, MPH, 3471 Fifth Avenue, Suite 500, Pittsburgh, PA 15213. E-mail: studenskis{at}msx.dept-med.pitt.edu
 |
Abstract |
|---|
 Top Abstract MethodsResultsDiscussionReferences |
|---|
Methods. This prospective cohort study was based in two health care systems. Data were collected during home visits at baseline and every 3 months for 1 year. Subsequent deaths occurring within 5 years were ascertained using the National Death Index.
Results. Of 439 older adults, 88 (20%) died within the subsequent 5 years. The optimal magnitude of decline to predict 5-year mortality was 0.1 meters/second for gait speed, 1 point for the Short Physical Performance Battery, and 0.05 points for Euroqol. Independent contributions were found for decline in gait speed (p =.001 to.002), Short Physical Performance Battery (p =.014 to.026), global health (p <.001), and activities of daily living (p =.005 to.019). More than one half of the episodes of decline were transient. Persistent decline in 1 year consistently predicted death, and transient decline in gait speed and global health increased mortality risk compared with no change.
Conclusions. A decline in gait speed of 0.1 m/s or 1 point in the Short Physical Performance Battery within 1 year increased the subsequent 5-year mortality rate. Transient declines in gait speed and self-reported health are as common as persistent declines and affect mortality risk.
The purpose of this study was to determine the magnitude of decline in measures of self-reported health, self-reported functional status, and physical performance that best predict mortality risk within 5 years. We initially assessed decline using a priori clinical estimates, performed sensitivity analyses, assessed for threshold effects, and evaluated the influence of baseline status on the predictive capacity of decline. We assessed the independent contribution of each measure when considered in groups. Finally, we distinguished effects of persistent versus transient change on mortality risk.
|
METHODS |
|---|
|
TopAbstractMethodsResultsDiscussionReferences |
|---|
Participants
We recruited participants from a veteran's affairs hospital–based ambulatory care site and a Medicare health management organization serving a common geographic area. Eligible persons were 65 years or older, living in the community within 20 miles, and received care in the same health care system for 1 year or more. Mental status eligibility was based on ability to maintain a utilization diary and communicate self-reported health status. We excluded persons with Mini-Mental State Examination (15) scores 
16, but those with scores of 16–23 were eligible if a caregiver could help maintain the diary. We excluded persons who could not walk at least 4 meters and those considered too fit or too frail (determined by a gait speed of more than 1.3 m/s or less than 0.2 m/s).
Of the 572 persons screened, 492 (86%) entered the study. During the following 12 months, 18 died, 20 changed provider systems, 13 withdrew, and 2 moved out of the area, resulting in a final sample of 439 participants (72% Medicare health management organization; 28% Veteran's Affairs) for the study.
Measures
Participants had baseline assessments of demographic characteristics and cognition by the Mini-Mental State Examination (15), and they self-reported any comorbid conditions (16). Physical performance, health status, and self-reported functional status were assessed at a home visit at baseline and at 3, 6, 9, and 12 months. We assessed health status using the Euroqol (17) and the five-level global health item from the Medical Outcomes Study SF-36 (18). We assessed functional status using the activity of daily living items (ADL) from the National Health Interview Survey (19) and the physical function index of the SF-36 (18). We measured physical performance using the Short Physical Performance Battery (SPPB) (1,20) and usual gait speed (21). Table 1 gives the possible ranges for these measures. Whether participants were hospitalized was ascertained using their diaries and hospital records. Inter-rater and test-retest reliability for the measures used in this study was excellent, with intraclass correlations generally more than 0.9 (11).
|
We determined participants' decline from baseline during any quarterly follow-up visit. For classification of transience and persistence, we determined decline during the first three quarters, and then we classified participants' declines as "persistent" and "transient" based on whether the decline was still present at the end of the fourth quarter. We determined time to death using the Social Security Death Index (14). Administrators of the death index claim that all deaths are confirmed by a family member or by death certificate.
Statistical Analyses
We compared the numbers of deaths in those with and without decline in the six main measures of function and health. We followed participants who survived 12 months from baseline and had data collected during that interval for an additional 5 years. We used Cox proportional hazards models (22) to estimate hazard ratios, plus their 95% confidence intervals and probability values, to compare the rates of death across groups. We adjusted the models for age, sex, baseline status, comorbid burden, and ensuing hospitalization within the first year of follow-up. We considered interactions between decline and ensuing hospitalization and stratification of analyses by whether hospitalized. We fit the same models using ordinal scores for three levels of change (none, transient, and persistent), including only those participants with complete data for the entire first year of follow-up. We used SAS version 8.02 software for all analyses (23). We ran similar analyses post hoc, with larger and smaller values for decline in the six measures than those selected a priori, to determine whether the magnitude-of-change estimates were optimal. We conducted sensitivity analyses using receiver operator characteristic curves to determine whether outcome was more sensitive to a threshold baseline value than to a fixed amount of change (24). When we observed similarities in rates of decline and effects of decline on death between participants from the two facilities, we combined data to improve statistical power.
|
RESULTS |
|---|
|
TopAbstractMethodsResultsDiscussionReferences |
|---|
The proportion of participants who ever met criteria for meaningful change ranged from 13% for declines in self-reported global health to 39% for decreases of 10 points in physical function by the SF-36 (Table 2). In 5 years, 88 (20%) of the participants had died. Baseline status on all six measures independently predicted death, as did age, sex, and comorbid burden. Receiver-operator curves of baseline status for each of the six variables plotted against deaths did not show threshold effects; the increase in mortality rate with worsening baseline scores was gradual and continuous (data not shown). Persons who met a priori criteria for decline in gait speed, self reported global health, and new dependence in ADL by the National Health Interview Survey items were at greatest risk for death (Table 3). Sensitivity analyses run with larger and smaller values for change in the four continuous measures (gait speed, SPPB, Euroqol, and SF-36) confirmed accurate clinical estimates for change in gait speed. SF-36 physical function change had no value for predicting death. Changes of 1 point on the SPPB (rather than 2 points) and 0.05 point on the Euroqol (rather than 0.1 point) were slightly better discriminators than the original estimates and remained independent predictors of death in adjusted analyses (Table 3). For SPPB and Euroqol, the frequency of decline increased with the smaller estimates of change to 55% with a 1-point decline in SPPB and to 42% with a 0.05-point decline on the Euroqol. We used the optimal estimates of change for all subsequent analyses.
|
|
=.05, which prevents us from making a stronger conclusion about differential associations between decline and survival across hospitalized and nonhospitalized groups. Change in gait speed, SPPB, self-reported global health, and self-reported functional status consistently remained independent predictors of death after accounting for baseline status of the measure of change, change in other measures, age, sex, ensuing hospitalization, and comorbid burden (Tables 4, 5, and 6).
|
|
|
|
|
|
|
DISCUSSION |
|---|
|
TopAbstractMethodsResultsDiscussionReferences |
|---|
Rates of change in measures of function among older adults are beginning to be described. In a cohort of disabled older women, decline in walking speed was estimated to range from 0.027 to 0.065 m/s per year depending on the participants' ages and initial abilities. For SPPB, declines were 0.3–0.9 points per year (25). The magnitude of change found to predict death in the current study (0.1 m/s for gait speed and 1 point for SPPB) was greater than typical change in older women and thus may discriminate beyond aging and time alone. When assessed weekly, physical performance measures tend to show small fluctuations that correspond only modestly with concurrent symptoms or reports of function (26–29).
In a prospective study of community-dwelling elderly persons, performance change occurring within 1 year correlated with concurrent self-reported change in disability, but change in performance within 1 year did not predict disability at 3 years, after accounting for disability status at the end of the first year (30). The authors suggested that this unexpected finding might be due to an inadequate frequency of assessment or to the effect of assigning a missing value to a failure to complete a performance test. Self-reported decline in self-care and lower body function in a period of 2 years predicted mortality risk during the ensuing 2 years after adjustment for many predisposing, enabling, need, and utilization factors (31). The pattern of changes in self-care activities in two consecutive 2-year time periods discriminated both final functional status and mortality risk (32). Our study appears to be the first to define a magnitude of meaningful change in addition to a general prognostic effect on survival. We also found that performance change predicts death within 5 years even after accounting for self-reported changes in ADLs.
Transient states occurring within periods of time longer than weeks and shorter than years may be meaningful. Gill and colleagues (33) found the cumulative frequency of self-reported disability to be nearly twice the prevalence of self-reported disability and noted that the difference between the two estimates increased over time. These investigators recently reported that short-term self-reported disability is predictive of the development of subsequent and persistent disability or death (hazard ratio, 
2.1) among nondisabled community-dwelling elders (34). In their words, "given the dynamic nature of disability, more frequent assessments of functional status could lead to more accurate estimates of active life expectancy and an improved understanding of the course and overall burden of disability." In the current study, one half or more of the declines were transient and would not have been detected with annual screening, even though the transient declines in gait speed and self-rated health had independent prognostic value.
The design of this study offers advantages for the assessment of the meaning of change occurring over time in older adults. It used a community-based population of older adults recruited from primary care offices, making results generalizable to providers and systems of care for older adults. Follow-up and retention rates were very high, minimizing the effects of missing data. Participating older adults were assessed in person every 3 months, allowing for frequent monitoring of status and detection of transient effects.
The current study has limitations as well. We used the National Death Index to confirm deaths. Random missing deaths would tend to dilute the effects we found. False-positive reports of death are unlikely because the National Death Index requires confirmation by someone known to the deceased. Our self-reported measure of comorbidity, rather than direct assessment of physiologic deficits, may be considered a limitation, although self-reported morbidity is commonly used in studies of aging (3,31). More careful assessment of subclinical impairments may well identify specific causative factors better than simple performance measures. In a study of 5-year mortality, Fried and colleagues (35) found 20 factors that predicted death. Although baseline self-reported functional status remained an independent predictor of death, after accounting for the other physiologic, behavioral, and demographic indicators, baseline physical performance did not. The discrepancy between Fried's finding of no independent effect of baseline physical performance on mortality rate and reports both here and in previous studies of major effects may be due to some combination of the analytic strategy, the mix of covariates, and the careful detection of preclinical disease. Nevertheless, even if root causes of death are truly due to subclinical physiologic disturbances, performance change may be useful for clinical screening and monitoring of change, whereas physiologic indicators may be more appropriate for differential diagnosis and preventive interventions.
Our study did not detect effects on mortality rate of change in the SF-36 physical function measure. Although widely used, other investigators have also noted problems with this measure to detect change in older populations (36,37). The SF-36 measure may not be sensitive to change in more functionally limited populations and may suffer from lack of precision in persons who cannot accurately describe their current level of difficulty in activities they do not perform, such as walking 1 mile or climbing several flights of stairs. Finally, our study did not determine whether improvement in similar measures affects mortality rate or whether change in these measures affects other outcomes such as disability or health care services utilization.
We found that changes in self-reported status and physical performance simultaneously contribute to survival. Performance and self-reported measures each have special advantages and limitations (38,39). Self-reports can be obtained without face-to-face contact and is more likely to integrate over time and represent the person's own perspective. Performance may be more sensitive to preclinical disability and may allow for finer levels of discrimination. Both may be affected by mood or expectations. In terms of special issues related to change, self-reports might be vulnerable to problems with "response shift," in which a person's experiences alter his or her opinion of previous and current status. Performance might be more vulnerable to problems with missing data if a person becomes unable to perform the test and the test does not include a scorable status for "can't do" (39). Thus, the findings we describe here would not point to the superiority of measuring change using either performance or report alone but would suggest that the measures have complementary value and might well be used together.
Conclusion
Change in both physical performance and self-reported health during a period of 12 months affects survival. A magnitude of decline that affects mortality rate is proposed and tested. Transient declines, in addition to persistent change, affect survival. Gait speed appears to be a consistent and meaningful indicator of change. Monitoring both gait speed and self-reported health in usual health care settings might help to detect persons with degrees of decline who are at greater risk for death and warrant preventive interventions.
|
Acknowledgments |
|---|
Dr. Perera is now with the University of Pittsburgh, Pennsylvania.
|
Footnotes |
|---|
Received February 1, 2004
Accepted May 11, 2004
|
References |
|---|
|
TopAbstractMethodsResultsDiscussionReferences |
|---|
This article has been cited by other articles:
|
|
 |
|
  D. K. Miller, F. D. Wolinsky, E. M. Andresen, T. K. Malmstrom, and J. P. Miller Adverse Outcomes and Correlates of Change in the Short Physical Performance Battery Over 36 Months in the African American Health Project J. Gerontol. A Biol. Sci. Med. Sci., May 1, 2008; 63(5): 487 - 494. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 D. Sumukadas MBBS MD, M. D. Witham MBChB PhD, A. D. Struthers MBChB MD, and M. E.T. McMurdo MBChB MD Effect of perindopril on physical function in elderly people with functional impairment: a randomized controlled trial Can. Med. Assoc. J., October 9, 2007; 177(8): 867 - 874. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
C. Rosano, H. J. Aizenstein, S. Studenski, and A. B. Newman A Regions-of-Interest Volumetric Analysis of Mobility Limitations in Community-Dwelling Older Adults J. Gerontol. A Biol. Sci. Med. Sci., September 1, 2007; 62(9): 1048 - 1055. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
The LIFE Study Investigators* [*See Appendix for L Effects of a Physical Activity Intervention on Measures of Physical Performance: Results of the Lifestyle Interventions and Independence for Elders Pilot (LIFE-P) Study J. Gerontol. A Biol. Sci. Med. Sci., November 1, 2006; 61(11): 1157 - 1165. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
| All GSA journals | The Gerontologist |
| Journals of Gerontology Series B: Psychological Sciences and Social Sciences | |
