This disagreement about a key measurement issue in geriatric assessment prompted us to ask whether the current practice of defining BADL disability as either difficulty or dependence might represent a false choice. We sought to determine whether questions about difficulty and dependence provide complementary information. We hypothesized that persons who are independent in BADLs but have difficulty would have functional profiles, physical performance scores, and rates of health care utilization and death that are between those of persons who are BADL independent and have no difficulty and persons who are BADL dependent. If this hypothesis is correct, it would suggest that clinicians and investigators could depict the continuum of disability more fully by including questions about both difficulty and dependence in their clinical practice or epidemiologic studies.

Methods

Top Methods Results Discussion Author & Article Info References

Participants

Participants were members of Project Safety, a probability sample of community-living persons 72 years of age and older in New Haven, Connecticut, in 1989. The sampling technique, described in detail elsewhere [5], was similar to that used to establish the New Haven site of the Established Populations for Epidemiologic Studies of the Elderly [6]. Of the 1436 persons originally contacted, only 44 (3%) did not meet the three eligibility criteria: the ability to speak English, Spanish, or Italian; the ability to follow simple commands; and the ability to walk across a room without the assistance of another person. Of those eligible, 1103 (79%) agreed to participate and were enrolled in the cohort. The sample for the current study comprised the 1065 participants who had complete baseline data on BADL function. The 38 persons who had incomplete BADL data did not differ significantly from those in our study sample in terms of mean age (81.2 years compared with 79.5 years), sex (76.3% female compared with 72.8% female), or ethnicity (81.6% white compared with 84.0% white).

Data Collection

Baseline interviews and assessments were completed in participants' homes by a trained nurse researcher. To assess BADL function, participants were asked two separate questions for each of six personal care tasks: bathing, dressing, transferring from bed to chair, eating, using the toilet, and grooming [1, 7]. Participants were first asked, "At the present time, do you need help (yes/no) from another person to (perform the task)?" They were then asked, "How much difficulty (none, some, a lot), on average, do you have doing this [task]?" Measures of higher-level function included 1) four instrumental activities of daily living (performing light and heavy housework, shopping, and driving a car) derived from the Older American Resource Services instrument [8]; 2) mobility, determined from the number of blocks walked on an average day; 3) physical activity, assessed with a modified version of the Yale Physical Activity Scale [9]; and 4) social activity, scored as the sum of the frequency ratings of eight groups of activities (attending events, taking trips, engaging in paid work, volunteering, visiting friends, attending religious services, participating in groups, and going to museums or shows) [10]. Physical performance was assessed with three timed tests that previous investigations [11, 12] found to be most predictive of BADL dependence at 1 year. The tests included walking back and forth over a 10-foot course, turning in a full circle, and standing up and sitting down from a hard-back chair three times with arms folded. The time (in seconds) needed to complete each task "as quickly as possible" was recorded.

During follow-up interviews at 1 and 3 years, BADL function was reassessed and participants were asked, "Do any health-care workers visit you in your home on a regular basis to take care of you (yes/no)?" The remaining outcomes were monitored over a 4-year period. Hospitalizations were ascertained from monthly surveillance of discharge records at two local hospitals, which account for more than 90% of acute care admissions for persons living in New Haven, and from Health Care Financing Administration (HCFA) data tapes, which include admissions to hospitals within and outside of New Haven. Admissions to skilled-nursing facilities were ascertained from the Connecticut Long Term Care Registry [13]. Mortality was monitored by contacts with participants or their proxies during monthly surveillance and by review of local obituaries [5].

Statistical Analysis

For our primary analysis, we classified participants into one of three BADL groups: 1) independent without difficulty if they required no personal assistance and reported no difficulty in any BADL [n = 701]; 2) independent with difficulty if they required no personal assistance in any BADL but reported some or a lot of difficulty in one or more BADLs [n = 227]; and 3) dependent if they required personal assistance in one or more BADLs (n = 137).

To test our hypothesis, we did a series of cross-sectional and longitudinal analyses. In the cross-sectional analysis, we first compared the proportion of participants in the three BADL groups who had poor higher-level function. Measures of poor higher-level function included dependence in one or more instrumental activities of daily living, no blocks walked on an average day, worst quarter of physical activity, and worst quarter of social activity. Next, for each of the timed tests of physical performance, we dichotomized the scores to distinguish participants in the worst quarter from those in the other three quarters [11, 12], and we compared the proportion of participants in the three BADL groups who scored in the worst quarter.

In the longitudinal analysis, we first compared the rates of hospitalization and regular visits by a home health care worker among the three BADL groups. We then calculated survival curves for admission to a skilled-nursing facility and for death using the product-limit method of Kaplan and Meier [14]. In both survival analyses, participants were censored at the end of the monthly surveillance (31 August 1993) if an outcome had not occurred. When the outcome was admission to a skilled-nursing facility, decedents without a previous admission to a skilled-nursing facility were censored at the time of death. Differences in survival curves were assessed by using the log-rank test statistic [15]. The Mantel-Haenszel chi-square statistic for linear trend was used for all other statistical comparisons among the three BADL groups. Finally, among participants who were BADL independent at baseline, we compared the rate of new BADL dependence over a 3-year period between those who had difficulty and those who did not. For all rates and proportions, we calculated 95% CIs.

To determine whether the postulated association between difficulty and dependence was seen for individual BADLs, we did a secondary set of analyses. Bathing was evaluated because it is generally the BADL with the highest prevalence of disability [11]. We classified participants into one of three bathing groups-independent with no difficulty (n = 812), independent with difficulty (n = 139), and dependent (n = 114)-and repeated the cross-sectional and longitudinal analyses described previously.

Results

Top Methods Results Discussion Author & Article Info References

The number and type of BADL disabilities among participants who were BADL independent with difficulty and those who were BADL dependent are shown in Table 1. Most participants in each group had only one or two BADL disabilities; disability was most common in bathing and dressing.

In the cross-sectional analysis, the proportion of participants who had poor higher-level function was lowest in those who were BADL independent without difficulty, intermediate in those who were BADL independent with difficulty, and highest in those who were BADL dependent (chi-square trend, P < 0.001). This finding was seen for each self-reported measure (Table 2, top). The results for physical performance were similar (Table 2, middle). For each of the timed tests, the proportion of participants in the quarter with the worst function increased across the three BADL groups (chi-square trend, P < 0.001). With only two exceptions, the 95% CIs for pair-wise comparisons between BADL groups in the cross-sectional analysis did not overlap; this provides even stronger evidence for the graded association among the three BADL groups.

In the longitudinal analysis, rates of hospitalization and regular visits by a home health care worker for participants who were BADL independent with difficulty were intermediate compared with corresponding rates for the other two groups (Table 2, bottom). Among persons who were BADL independent, those with difficulty were significantly more likely to develop BADL dependence over a 3-year period than those without difficulty (relative risk, 1.7 [95% CI, 1.3 to 2.2]). The cumulative probability of admission to a skilled-nursing facility increased steadily for all three BADL groups (Figure 1); however, the probability was significantly greater for participants who were BADL dependent and significantly lower for participants who were BADL independent without difficulty than for participants who were BADL independent with difficulty. The mortality curves for the BADL groups also differed significantly; the highest mortality rate was seen in participants who were BADL dependent, and the lowest mortality rate was seen in participants who were BADL independent without difficulty (Figure 2). For the combined end point of admission to a skilled-nursing facility or death, the results were similar to those of the preceding analyses (data not shown). This indicates that for admission to a skilled-nursing facility, group comparisons were not biased by censoring deaths.

View larger version (17K): [in this window] [in a new window]   Figure 1. Cumulative probability of admission to a skilled-nursing facility for the three basic activity of daily living groups. The data were calculated by using the product-limit method of Kaplan and Meier. Data on participants who died without an admission to a skilled-nursing facility were censored at the time of death. P < 0.001 by log-rank test for comparisons between curves A and B and curves A and C; P = 0.017 by log-rank test for comparison between curves B and C.

View larger version (16K): [in this window] [in a new window]   Figure 2. Mortality curves for the three basic activity of daily living groups. The data were calculated by using the product-limit method of Kaplan and Meier. P = 0.034 by log-rank test for comparison between curves A and B; P < 0.001 by log-rank test for comparison between curves A and C; P < 0.003 by log-rank test for comparison between curves B and C.

When we repeated the cross-sectional and longitudinal analyses for the three bathing groups, results were similar to those of the other analyses. In persons who were independent in bathing but had difficulty, functional profiles, physical performance scores, and rates of health care utilization and death were intermediate to those of persons who were either independent in bathing with no difficulty or dependent in bathing (data available from authors).

Discussion

Top Methods Results Discussion Author & Article Info References

In this large, community-based study, we found that in older persons who are BADL independent but have difficulty, functional profiles, physical performance scores, and rates of health care utilization and death are intermediate to those of persons who are BADL independent without difficulty and persons who are BADL dependent. Results were similar when we focused solely on bathing, the BADL with the highest prevalence of disability. These findings suggest that responses to questions about difficulty and dependence provide complementary information that together depicts the continuum of disability more fully than does the response to either question alone.

When assessing BADL disability, epidemiologic studies have generally focused on difficulty [16-19] or dependence [20-25]. Among the few studies that have rated both difficulty and dependence, two used information on dependence alone when defining BADL disability [26, 27]. To our knowledge, only one study [28] evaluated the outcomes of persons who are BADL dependent and those who are BADL independent with difficulty. In that study, Harris and colleagues, using data from the Supplement on Aging to the 1984 National Health Interview Survey, found that persons 80 years of age or older were twice as likely to die within 2 years if they were BADL dependent than if they were BADL independent with difficulty [28]. In contrast to our study, however, this study did not find a significant difference in the rate of nursing home admission between persons in these two BADL groups. One explanation may be that Harris and associates' study, unlike ours, was not specifically designed to determine whether assessing BADL disability as difficulty or dependence identifies persons who have distinct clinical courses. In our study, information on nursing home admissions and hospitalizations was obtained from state and federal registries and from active surveillance rather than from self-report; this approach minimized the possibility of inaccurate or biased recall. Furthermore, because we had information on date of nursing home admission and date of death, we could perform survival analyses for two of our primary outcomes.

The validity of our findings is strengthened considerably by three additional methodologic features. First, we had data on a rich array of both self-reported and performance-based measures. This allowed us to compare functional profiles and physical performance scores among the three BADL groups. Second, we confirmed the postulated relation between difficulty and dependence for bathing, the BADL with the highest prevalence of disability. Third, in contrast to Harris and colleagues' study, few participants in our study were lost to follow-up. Data on regular visits by a home care worker and the onset of BADL dependence were missing for only 5.9% and 2.7% of study participants, respectively. In addition, the amount of missing data did not differ among the three BADL groups. For hospitalization and admission to a skilled-nursing facility, essentially complete follow-up was ensured by our use of HCFA data tapes (complemented by our local hospital surveillance) and the Connecticut Long Term Care Registry. Finally, because our monthly surveillance system had a 99% response rate [5], the risk for undetected deaths in our study is probably small. Among 2812 participants in the Yale Health and Aging Project [6], a less aggressive surveillance system missed only 25 deaths during a 7-year period when tested against the National Death Index (unpublished data). Of note, because participants in our study were drawn from a probability sample, our findings should be generalizable to other urban populations of community-living older persons.

A potential limitation of our study is that we did not include walking as a BADL. Persons who could not walk across a room without personal assistance were not enrolled in Project Safety. However, walking was not one of Katz's original BADLs [1] and is often considered a measure of mobility rather than a personal care task [29].

Although our intent was not to evaluate transitions between functional states during the follow-up period, we did find that participants who were BADL independent with difficulty were nearly twice as likely to develop BADL dependence over a 3-year period as participants who were BADL independent without difficulty. Moreover, we previously reported that nearly 30% of Project Safety participants who were BADL dependent at either the baseline or 1-year interview recovered their independence within 2 years [30]. Taken together, these findings show that functional status and disability represent dynamic processes in older persons and should spur further research into the mechanisms of disability and recovery in this population.

Our study was prompted by recent arguments that BADL disability should be defined as difficulty rather than as dependence [4]. Advocates of the former strategy contend that dependence measures the presence of an intervention to reduce disability (that is, formal or informal care) rather than disability itself [4]. Our findings suggest that clinicians and investigators who want to measure BADL disability should assess both difficulty and dependence. When BADL disability is defined as dependence alone, the nondisabled group includes persons who are BADL independent without difficulty and those who are BADL independent with difficulty. Similarly, when BADL disability is defined as difficulty alone, the disabled group includes persons who are BADL independent with difficulty and those who are BADL dependent. Assessing both difficulty and dependence would allow older persons to be categorized into one of three disability groups: 1) BADL independent without difficulty [such persons have the best functional profiles and physical performance scores and the lowest rates of health care utilization and death], 2) BADL dependent [such persons have the worst functional profiles and physical performance scores and the highest rates of health care utilization and death], and 3) BADL independent with difficulty (such persons have intermediate functional profiles, physical performance scores, and rates of health care utilization and death). The inclusion of this latter group (which made up 21% of our study sample) as disabled largely explains why prevalence estimates of disability are considerably higher in epidemiologic studies that rate difficulty rather than dependence [2, 31]. Despite recent evidence that difficulty scales may be less reliable than dependence scales [32], it would be prudent for investigators to explicitly define both "difficulty" and "dependence" in their disability studies [2].

Our findings support the use of disability instruments [33] that offer three response options for each BADL item (independent without difficulty, independent with difficulty, and dependent). Although some disability instruments [34] now offer four response options (independent without difficulty, independent with some difficulty, independent with great difficulty, and dependent), no evidence suggests that subdividing the "difficulty" category provides additional useful information or that respondents can reliably distinguish between gradations of difficulty. On the basis of results of a recently published cross-sectional study [35], however, it may be possible to subdivide persons who are BADL independent without difficulty into two distinct groups by asking whether they have modified the methods used to complete their personal care tasks. If validated in future studies, these refinements in self-reported instruments could provide clinicians and investigators with a simple and inexpensive strategy with which to assess the broad spectrum of disability in older persons. Enhanced methods of assessing BADL function, such as ours, can advance patient care by helping clinicians to better define the effect of disease on everyday function, to determine personal care needs more accurately, to target preventive and restorative treatments more effectively, and to evaluate the effect of these interventions more completely.

In summary, in the assessment of disability in BADLs, questions about difficulty and dependence provide complementary information. The current practice of defining BADL disability as either difficulty or dependence may represent a false choice. Clinicians and investigators can depict the continuum of disability more fully by including questions about both difficulty and dependence in clinical practice or epidemiologic studies.

Dr. Robison: Braceland Center for Mental Health and Aging, Institute of Living, 400 Washington Street, Hartford, CT 06106.