HomeLarge Type Edition
HOME ARCHIVE SEARCH TABLE OF CONTENTS
This Article
Right arrow Abstract Freely available
Right arrow Full Text (PDF)
Services
Right arrow Download to citation manager
Citing Articles
Right arrow Citing Articles via HighWire
PubMed
Right arrow PubMed Citation
The Journals of Gerontology Series A: Biological Sciences and Medical Sciences 60:1013-1016 (2005)
© 2005 The Gerontological Society of America

Estimates of Active and Disabled Life Expectancy Based on Different Assessment Intervals

Thomas M. Gill1,, Heather Allore1, Susan E. Hardy1, Theodore R. Holford2 and Ling Han1

Departments of 1 Internal Medicine
2 Epidemiology and Public Health, Yale University School of Medicine, New Haven, Connecticut.

Address correspondence to Thomas M. Gill, MD, Yale University School of Medicine, Dorothy Adler Geriatric Assessment Center, 20 York Street, New Haven, CT 06504. E-mail: gill{at}ynhh.org


   Abstract
Top
 Abstract
Methods
 Results
 Discussion
 References
 
Background. Although disability in activities of daily living (ADLs) is a highly dynamic process, analytic strategies for estimating active and disabled life expectancy have assumed stability in ADL function between periodic surveys spanning 12–24 months or have used interval estimation or instantaneous rates based on long assessment intervals. We performed a prospective cohort study to compare estimates of active and disabled life expectancy based on traditional assessment intervals of 1–2 years with those based on more frequent assessments at 1-month intervals.

Methods. Participants included 754 initially nondisabled community-dwelling persons, aged 70 years or older, who were interviewed monthly for 4 years to ascertain ADL disability. Estimates of active and disabled life expectancy were calculated using an increment–decrement life table for assessment intervals of 1 month, 1 year, and 2 years.

Results. For each of five age groups, the monthly assessment strategy yielded the highest values for active life expectancy and the lowest values for disabled life expectancy. The 95% confidence intervals for these values, however, overlapped the corresponding point estimates for the annual and biennial strategies.

Conclusions. Accurate estimates of active and disabled life expectancy may be obtained from epidemiologic studies that assess ADL function no more frequently than every other year.

ACTIVE life expectancy, defined as the projected amount of remaining time free of disability in activities of daily living (ADLs), is often used by policy makers to forecast the functional health of older persons (1). To date, analytic strategies for estimating active life expectancy have assumed stability in ADL function between periodic surveys spanning 12–24 months (1–4) or have used interval estimation (i.e., estimating transition probabilities for shorter periods of time based on longer observational periods) or instantaneous rates (i.e., calculating the probability of an event at a single point in time) based on long assessment intervals (5–7). Recent evidence, however, has demonstrated that disability for many older persons is a highly dynamic process characterized by multiple transitions into and out of disability over the course of only 2 years (8,9). It is not known whether relatively infrequent assessments of ADL function lead to inaccurate estimates (i.e., too high or too low) of active life expectancy. The same is true for estimates of disabled life expectancy, defined as the projected amount of remaining time disabled in ADLs.

The objective of the current study was to compare estimates of active and disabled life expectancy based on traditional assessment intervals of 1 or 2 years with those based on more frequent assessments at 1-month intervals. We used data from a unique longitudinal study that includes monthly assessments of ADL function for 4 years among a large cohort of community-dwelling older persons, with little missing data and few losses to follow-up.


   METHODS
Top
 Abstract
 Methods
Results
 Discussion
 References
 
Study Population
Participants were members of the Precipitating Events Project, a longitudinal study of 754 community-dwelling persons, aged 70 years or older, who were nondisabled (i.e., required no personal assistance) in four key ADLs—bathing, dressing, walking inside the house, and transferring from a chair. The assembly of the cohort, which took place between March 1998 and October 1999, has been described in detail elsewhere (10,11). In brief, potential participants were identified from a computerized list of 3157 age-eligible members of a large health plan in greater New Haven, Connecticut. Eligibility was determined during a screening telephone interview and was confirmed during an in-home assessment. Persons were excluded based on the following criteria: significant cognitive impairment with no available proxy (8), inability to speak English, diagnosis of a terminal illness with a life expectancy less than 12 months, and plan to move out of the New Haven area during the next 12 months. Only 4.6% of the 2753 health plan members who were alive and could be contacted refused to complete the screening telephone interview, and 75.2% of the eligible members agreed to participate in the project, which was approved by the Human Investigation Committee at Yale University. Persons who declined to participate did not differ from those who were enrolled in terms of age or sex.

Data Collection
The baseline assessments were completed by trained research nurses using standard instruments. Data were collected on demographic characteristics, cognitive status as assessed by the Mini-Mental State Examination (12), and 13 self-reported, physician-diagnosed chronic conditions: hypertension; myocardial infarction; congestive heart failure; stroke; diabetes; arthritis; hip fracture; fracture of wrist, arm, or spine since age 50; amputation of leg; chronic lung disease; cirrhosis or liver disease; cancer (other than minor skin cancers); and Parkinson's disease. Collection of these baseline data was 100% complete.

Complete details regarding the assessment of disability, including formal tests of reliability and accuracy, are provided elsewhere (8). During monthly telephone interviews, participants were assessed for disability using standard questions that were identical to those used during the screening telephone interview (8). For each of the four key ADLs, 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. Conversely, those who did not need help were considered to be nondisabled (or independent). Participants were not asked about eating, toileting, or grooming. The incidence of disability in these three ADLs is low among nondisabled, community-dwelling older persons (13,14). Furthermore, it is highly uncommon for disability to develop in these ADLs without concurrent disability in bathing, dressing, walking, or transferring (13–15). Among a subgroup of 91 participants who were interviewed twice within a 2-day period by different interviewers, we found that the reliability of our disability assessment was substantial (16), with Kappa = 0.75 for disability in one or more of the four ADLs. Kappa was 1.0 for the 18 paired interviews that were completed independently by different interviewers on the same day. For participants with significant cognitive impairment, the monthly telephone interviews were completed with a designated proxy. The accuracy of these proxy reports for disability, compared to reports from cognitively intact participants, was excellent, with Kappa = 1.0 (8). Deaths were ascertained by review of the local obituaries and/or from an informant during a subsequent telephone interview.

Statistical Analysis
We calculated active and disabled life expectancy using increment–decrement life tables (2,5,17). Because age is the strongest demographic determinant of disability and life expectancy, we report results separately for the following five age groups: 70–74 years, 75–79 years, 80–84 years, 85–89 years, and 90 years or older. We do not report results separately for men and women because of stochastic variability inherent in small samples and because our intent was not to calculate population-based estimates of active and disabled life expectancy, but rather to compare estimates using different assessment intervals. We chose not to model transitions between the four possible states (independence, disability, death, and missing) as a Markov chain because recent evidence suggests that the underlying assumption regarding independence of transitions over time may not be valid (11). Also, our sample size was not sufficiently large to implement a Markov chain model for the resulting 16 possible transitions for each of the five age groups.

We calculated mortality rates and probabilities of independence and disability for each age group using an incidence-density approach. For the monthly assessment strategy, we counted the total number of deaths and person-months during each year of follow-up, computed mortality rates per person-month, converted these results to annual mortality rates, and entered these rates into the increment–decrement life table to compute age-specific probabilities of death and, subsequently, life expectancy (17). Next, during each year of follow-up we counted the total number of months of independence and disability, respectively, and divided these results by the total number of person-months. These proportions were subsequently used to partition total life expectancy into active life expectancy and disabled life expectancy in the increment–decrement life table. The aforementioned procedures allowed us to simultaneously account for the effects of advancing age and time. For example, a participant who was 73 years at baseline and had 4 years of follow-up would contribute 2 person-years to each of the first (70–74 years) and second (75–79 years) age groups.

For the annual assessment strategy, we assumed that participants were interviewed only every 12 months, so that independence or disability was determined each year based solely on the interviews completed at months 12, 24, 36, and 48, respectively. If a participant was disabled (or independent) at month 12, for example, he or she was considered to be disabled (or independent) for all 12 months during the prior year (5). For nondecedents who did not complete an annual interview (<1% for each), ADL status for the prior year was set to missing. Decedents received no credit towards active or disabled life expectancy for the time from the last completed annual interview to the time of death (6). A similar set of procedures was used for the biennial assessment strategy except that independence or disability was determined every other year based solely on the interviews completed at months 24 and 48, respectively. When calculating the age-specific probabilities of death, the time from the last completed annual (or biennial) interview to the time of death was not included in the denominator to ensure that life expectancy equaled the sum of active and disabled life expectancy (5). We chose not to estimate active and disabled life expectancy from a single 4-year interval for two reasons. First, an increment–decrement life table cannot be implemented from only a single transition. Second, most prior studies have included assessment intervals of 1–2 years (1–4).

Because the increment–decrement life table method does not provide estimates of variance, we used bootstrapping methodology, as suggested by Land and colleagues (5), to calculate 95% confidence intervals for active and disabled life expectancy. We generated 1000 pseudo-samples of 754 participants, calculated an increment–decrement life table for each sample, and computed means and 95% confidence intervals based on the observed upper and lower 2.5% tails from each of the distributions. The 95% confidence intervals were used to compare the estimates of active and disabled life expectancy for the different assessment strategies. Customized code was written to construct the increment–decrement life tables and bootstrapped samples using SAS version 8.2 (18).


   RESULTS
Top
 Abstract
 Methods
 Results
Discussion
 References
 
The baseline characteristics of the study participants are shown in Table 1. During the 4-year follow-up period, 132 (17.5%) participants died after a median follow-up of 26 months, and 30 (4.0%) dropped out of the study after a median follow-up of 21 months. Data were otherwise available for 99.1% of the 32,337 monthly telephone interviews.


View this table:
[in this window]
[in a new window]
  		Table 1. Baseline Characteristics of Study Participants.

 
Table 2 provides mean values for active and disabled life expectancy, with 95% confidence intervals, by age group for each of the three assessment intervals. As expected, values for active life expectancy decreased with advancing age. For disabled life expectancy, values ranged from 1.7 to 2.1 for the youngest three age groups, were modestly lower for the 85–89 age group, and were considerably lower for the 90+ age group. For each age group, the monthly assessment strategy yielded the highest values for active life expectancy and the lowest values for disabled life expectancy. The 95% confidence intervals for these values, however, overlapped the corresponding point estimates for the annual and biennial strategies, indicating that the small differences in active and disabled life expectancy were not statistically significant.


View this table:
[in this window]
[in a new window]
  		Table 2. Active and Disabled Life Expectancy According to Age and Assessment Interval.

 

   DISCUSSION
Top
 Abstract
 Methods
 Results
 Discussion
References
 
In this prospective cohort study, we found that estimates of active and disabled life expectancy are relatively insensitive to the length of the assessment interval, at least for intervals up to 2 years. Assessing ADL function every month instead of every 1–2 years resulted in small, nonsignificant increases in active life expectancy and small, nonsignificant decreases in disabled life expectancy.

Accurate estimates of active and disabled life expectancy are important for forecasting future medical and long-term care costs and for planning effectively for services to meet the future needs of older Americans (3,19). Estimates of active and disabled life expectancy were originally calculated using a single-decrement table, which treated disability as a permanent condition (1), but are now routinely calculated using more sophisticated multistate methods (2,4,17), which allow for the possibility of recovery from disability. Nonetheless, most analytic strategies for estimating active life expectancy have assumed stability in ADL function between periodic surveys spanning 12–24 months (1–4). Because this assumption is no longer tenable (8,9), we set out in the current study to compare estimates of active and disabled life expectancy based on traditional assessment intervals of 1 or 2 years with those based on more frequent assessments at 1-month intervals. Our "null" findings bolster the validity of current estimates of active and disabled life expectancy and suggest that ADL function need not be assessed more frequently than every other year to yield accurate estimates. While periodic surveys spanning 12–24 months will undoubtedly miss many intermittent episodes of disability, this "undercounting" appears to be more than offset by "overcounting" when disability is presumed to persist for the duration of the 12- to 24-month assessment interval. This phenomenon is shown in Table 3, which provides data on the number of person-months of disability for assessment intervals of 1 month and 1 year. At each year of follow-up, disability for the 1-year assessment interval is undercounted because participants who are independent, dead, or missing at the annual assessment accrue no person-months of disability, and is overcounted because participants who are disabled at the annual assessment accrue 12 months of disability.


View this table:
[in this window]
[in a new window]
  		Table 3. Person-Months of Disability for Assessment Intervals of 1 Month and 1 Year According to Year of Follow-Up and Disability Status at Relevant Annual Assessment.

 
Our null results are consistent with those from prior studies that have used interval estimation or instantaneous rates based on assessment intervals ranging from 1 to 82 months (6,7). These studies, along with others (4,5), have relied on Markov chain models, which assume that the probability of a subsequent state is contigent only on the current state. Because the likelihood of disability is highly dependent on the occurrence and duration of prior disability episodes (11,20), we chose to estimate active and disabled life expectancy using an increment–decrement life table. In this strategy, age-specific rates of independence, disability, and death are calculated. The underlying assumption, which is similar to that for a Markov chain model, is that the rates for all persons in a specific age group are the same. Although methods have been proposed by Land and colleagues to estimate increment–decrement life tables with multiple covariates (5), we chose not to do so because our intent was not to calculate population-based estimates of active and disabled life expectancy, but rather to compare estimates using different assessment intervals. Also, the methods proposed by Land and colleagues rely, in part, on the use of Markov regression models.

We recognize potential limitations to our study. First, disability at baseline was an exclusion criterion. When we reran our analyses, using the 12-month interview as "baseline" (prevalence of disability = 10.2%), our results did not change appreciably (data available upon request). Second, our participants were members of a single health plan in a small urban area. According to the 2000 census (http://factfinder.census.gov), the demographic characteristics of persons aged 65 years or older are comparable for New Haven county and the United States, with the exception of race. New Haven county has a larger proportion of non-Hispanic whites relative to the United States (91.1% vs 83.7%). Despite this modest difference, we can think of no reason why our null findings should not be applicable to other populations of older persons. Third, eating, toileting, and grooming were omitted from our disability assessment. Although these omissions could lead to an underestimate of disability severity, they would have had little effect on our ascertainment of disability and, hence, on our estimates of active and disabled life expectancy. Indeed, despite the exclusion of persons who were nondisabled at baseline, our age-specific estimates of active life expectancy are only modestly higher than those that have been reported in three distinct populations of community-dwelling older persons, each of which included persons with disability at baseline (2).

Our study included monthly assessments of ADL function for 4 years on a large cohort of community-dwelling older persons, with a high participation rate, little missing data, and few losses to follow-up. To our knowledge, comparable data are available in no other study. Based on our results, we conclude that accurate estimates of active and disabled life expectancy may be obtained from epidemiologic studies that assess ADL function at intervals as long as 12–24 months.


   Acknowledgments
 
The work for this report was funded by grants from the National Institute on Aging (R01AG17560), the Paul Beeson Physician Faculty Scholar in Aging Research Program, the Robert Wood Johnson Foundation, and the Patrick and Catherine Weldon Donaghue Medical Research Foundation. The study was conducted at the Yale Claude D. Pepper Older Americans Independence Center (P30AG21342). Dr. Gill is the recipient of a Midcareer Investigator Award in Patient-Oriented Research (K24AG021507) from the National Institute on Aging.

We thank Denise Shepard, BSN, MBA, Shirley Hannan, RN, Andrea Benjamin, BSN, Martha Oravetz, RN, Alice Kossack, Barbara Foster, Shari Lani, Alice Van Wie, and the late Bernice Hebert, RN, for assistance with data collection; Evelyne Gahbauer, MD, MPH, for data management and programming; Wanda Carr and Geraldine Hawthorne for assistance with data entry and management; Peter Charpentier, MPH, for development of the participant tracking system; and Joanne McGloin, MDiv, MBA, for leadership and advice as the Project Director.


   Footnotes
 
Decision Editor: John E. Morley, MB, BCh

Received April 13, 2004

Accepted May 7, 2004


   References
Top
 Abstract
 Methods
 Results
 Discussion
 References
 

  1. Katz S, Branch LG, Branson MH, Papsidero JA, Beck JC, Greer DS. Active life expectancy. N Engl J Med. 1983;309:1218-1224.[Abstract]
  2. Branch LG, Guralnik JM, Foley DJ, et al. Active life expectancy for 10,000 Caucasian men and women in three communities. J Gerontol Med Sci. 1991;46:M145-M150.
  3. Manton KG, Stallard E, Liu K. Forecasts of active life expectancy: policy and fiscal implications. J Gerontol. 1993;48:(Special Issue): 11-26.
  4. Ferrucci L, Izmirlian G, Leveille S, et al. Smoking, physical activity, and active life expectancy. Am J Epidemiol. 1999;149:645-653.[Abstract/Free Full Text]
  5. Land KC, Guralnik JM, Blazer DG. Estimating increment-decrement life tables with multiple covariates from panel data: the case of active life expectancy. Demography. 1994;31:297-319.[Medline]
  6. Lievre A, Brouard N, Heathcote C. The estimation of health expectancies from cross-longitudinal surveys. Math Popul Stud. 2003;10:211-248.
  7. Laditka SB, Wolf DA. New methods for analyzing active life expectancy. J Aging Health. 1998;10:214-241.[Abstract/Free Full Text]
  8. Gill TM, Hardy SE, Williams CS. Underestimation of disability among community-living older persons. J Am Geriatr Soc. 2002;50:1492-1497.[Medline]
  9. Gill TM, Kurland B. The burden and patterns of disability in activities of daily living among community-living older persons. J Gerontol Med Sci. 2003;58:70-75.
  10. Gill TM, Desai MM, Gahbauer EA, Holford TR, Williams CS. Restricted activity among community-living older persons: incidence, precipitants, and health care utilization. Ann Intern Med. 2001;135:313-321.[Abstract/Free Full Text]
  11. Gill TM, Kurland BF. Prognostic effect of prior disability episodes among nondisabled community-living older persons. Am J Epidemiol. 2003;158:1090-1096.[Abstract/Free Full Text]
  12. Folstein MF, Folstein SE, McHugh PR. "Mini-mental state." A practical method for grading the cognitive state of patients for the clinician. J Psychiatr Res. 1975;12:189-198.[Medline]
  13. Gill TM, Williams CS, Tinetti ME. Assessing risk for the onset of functional dependence among older adults: the role of physical performance. J Am Geriatr Soc. 1995;43:603-609.[Medline]
  14. Gill TM, Richardson ED, Tinetti ME. Evaluating the risk of dependence in activities of daily living among community-living older adults with mild to moderate cognitive impairment. J Gerontol Med Sci. 1995;50A:M235-M241.[Abstract]
  15. Rodgers W, Miller B. A comparative analysis of ADL questions in surveys of older people. J Gerontol Psychol Sci. 1997;52:(special issue): 21-36.
  16. Landis JR, Koch GG. The measurement of observer agreement for categorical data. Biometrics. 1977;33:159-174.[Medline]
  17. Laditka SB, Hayward MD. The evolution of demographic methods to calculate health expectancies. In: Robine JM, Jagger C, Mathers CD, Crimmins EM, Suzman RM, eds. Determining Health Expectancies. Hoboken, NJ: John Wiley & Sons, Ltd; 2003:222–234.
  18. SAS Software System,, Version 8.2. Cary, North Carolina: SAS Institute, Inc; 2000.
  19. Guralnik JM, Alecxih L, Branch LG, Wiener JM. Medical and long-term care costs when older persons become more dependent. Am J Public Health. 2002;92:1244-1245.[Free Full Text]
  20. Hardy SE, Gill TM. Recovery from disability among community-dwelling older persons. JAMA. 2004;291:1596-1602.[Abstract/Free Full Text]



This article has been cited by other articles:


Home page
 Journals of Gerontology Series A: Biological Sciences and Medical SciencesHome page
T. M. Gill and E. A. Gahbauer
Evaluating Disability Over Discrete Periods of Time
J. Gerontol. A Biol. Sci. Med. Sci., June 1, 2008; 63(6): 588 - 594.
[Abstract] [Full Text] [PDF]

This Article
Right arrow Abstract Freely available
Right arrow Full Text (PDF)
Services
Right arrow Download to citation manager
Citing Articles
Right arrow Citing Articles via HighWire
PubMed
Right arrow PubMed Citation

HOME ARCHIVE SEARCH TABLE OF CONTENTS