Received 24th January 2000; returned for revisions 22nd March 2000; revised manuscript accepted 9th November 2000.

Transcription

1 Clinical Rehabilitation 2002; 16: Correspondence of the Functional Independence Measure (FIM) self-care subscale with real-time observations of dementia patients ADL performance in the home Ellen M Cotter Georgia Southwestern State University, Americus, Georgia, Louis D Burgio University of Alabama, Tuscaloosa, Alabama, Alan B Stevens, David L Roth University of Alabama at Birmingham and Laura N Gitlin Thomas Jefferson University, Philadelphia, Pennsylvania, USA Received 24th January 2000; returned for revisions 22nd March 2000; revised manuscript accepted 9th November Objective: To determine if caregivers of dementia patients can validly report (a) the level of patient activities of daily living (ADL) dependence and (b) the amount of time they spend providing ADL assistance. Design: Seven ADLs were assessed through caregiver report using the Functional Independence Measure (FIM) and then videotaped. Videotapes were coded using a computer-assisted data collection system that documented the nature and duration of caregiver assistance provided during ADLs. Setting: The study took place in the homes of the caregiver patient dyads. Subjects: Twenty-one people with dementia and their primary caregivers participated in the study. Main outcome measures: Data analysis examined (a) the degree to which caregiver-reported FIM scores corresponded to FIM scores derived from the observational data and (b) the extent to which caregivers estimates of ADL assistance time corresponded to the assistance time observed during ADL interactions. Results: For all ADLs, caregiver-reported FIM scores were correlated signi cantly with observation-derived FIM scores (r s values ranged from to 0.909), and means were similar. Assistance durations were also signi cantly correlated for some ADLs, but means were signi cantly different, indicating poor correspondence. Conclusions: Although caregivers of dementia patients can describe the nature of their ADL assistance with reasonable accuracy, they consistently Address for correspondence: Ellen M Cotter, Department of Psychology and Sociology, Georgia Southwestern State University, 800 Wheatley Street, Americus, GA , USA. emc@canes.gsw.edu or Louis D. Burgio, University of Alabama, Applied Gerontology Program, Box , Tuscaloosa, AL , USA. lburgio@sw.ua.edu. Arnold / cr465oa

2 FIM correspondence 37 overestimate their ADL assistance time, suggesting that caregiver reports of ADL assistance duration be used with caution. Additionally, the degree to which the assistance provided corresponds to the assistance actually needed needs to be addressed in future studies. Introduction Methods The accurate assessment of activities of daily living (ADLs) is vital for documenting functional ability and decline. However, clinicians often rely on proxy ratings of ADL performance, and previous research has suggested that proxy reports of ADL performance may not accurately re ect the patient s true abilities or what actually occurs during an ADL interaction. 1 3 Similarly, although caregiver estimates of ADL assistance time are often used to determine the indirect costs of caregiving, these reports have never been compared with objective assistance times, raising the same questions about validity and bias. Because gures regarding the prevalence of ADL impairment and indirect costs of caregiving are sometimes used to guide health care policy, accurately documenting the validity of these reports is crucial for determining the economic impact of dementia and other conditions on families. Although it is likely that caregiver ratings of ADL performance and estimates of ADL assistance time would be inaccurate, it should not be assumed that caregivers would be equally biased in both types of ratings. In other words, the nature and the duration of assistance, while correlated, represent different types of appraisals, with the result that a caregiver may accurately estimate one but not the other. Thus, it is necessary to examine both the nature and the duration of ADL assistance separately when attempting to determine the validity of caregiver reports. The purpose of this study is twofold: to determine the validity of caregiver-reported functional dependence ratings and to determine the validity of caregiver-reported ADL assistance durations. Caregiver ratings of ADL performance were obtained for seven basic ADLs that were then directly observed in patients homes for comparison to the caregiver-reported FIM scores. The correspondence between dementia caregivers estimates of the time required to provide ADL assistance and the actual time spent assisting with ADLs in the home was also examined. Subjects and setting Subjects were taken from a general pool of people referred to the University of Alabama at Birmingham (UAB) site of the National Institutes of Health-funded REACH (Resources for Enhancing Alzheimer s Caregiver Health) cooperative agreement, a multisite study of psychosocial interventions to improve the health and well-being of caregivers of dementia patients. Recruitment sources included the Visiting Nurses Association, the UAB Alzheimer s Family Care Program, and the UAB Geriatric Primary Care, Geriatric Assessment, and Alzheimer s Disease Center Memory Disorders Clinics, as well as support groups, health fairs, and newspaper and radio advertisements. Approval of the procedures and consent forms was obtained from the UAB Institutional Review Board prior to initiation of the study. Data were obtained from 21 caregiver patient dyads for this ancillary study. In each dyad, the patient was an adult with dementia and the caregiver was a co-residing relative of the patient. A minimum of one reported ADL impairment was necessary for inclusion. Caregivers in this sample ranged in age from 43 to 84 years (M = years). Women constituted 76% of the sample (n = 16), and 71% of the sample (n = 15) were white. Sixty-two per cent of the caregivers were either the daughter (n = 8; 38%) or the wife (n = 5; 24%) of the patient. On average, the caregivers had lived with the patients for approximately 22 years (range 1 58 years). Approximately 86% of the caregivers (n = 18) had at least a high school education. The patients in this sample were mostly women (67%; n = 14), with an average age of years (range years). The average Mini-Mental State Exam (MMS 4 ) score was 7.84 (range 0 24), representing severe cognitive impairment. It should be noted that because two patients refused to complete the MMSE, the reported mean score of 7.84 is based on n = 19.

3 38 EM Cotter et al. Paper-and-pencil measures Each patient completed the MMSE. 4 Caregivers provided data regarding ADL performance using the Functional Independence Measure (FIM 5,6 ). This instrument was modi ed slightly in that in addition to the six self-care subscale items (bathing, dressing upper body, dressing lower body, eating, grooming, and toileting), one of the mobility subscale items (bed/chair/ wheelchair transfer) was also measured, forming a list of items comparable to those on other standardized ADL assessments. These seven items were administered and scored by a certi ed FIM rater. Because these data were obtained through caregiver interview rather than through direct observation, the FIM items were administered using a procedure comparable to the one used for telephone administration of the FIM, which has been shown to have good agreement with observation-based FIM scores. 7 9 Additionally, for each ADL, the caregivers were asked to report how much time in minutes, on average, was spent each time they helped with that ADL. Although not a part of the FIM protocol, it was felt that soliciting this type of information would provide useful information regarding the validity of caregiver reports of ADLs. Observational measure ADL performance was videotaped using a micro-video camera designed by the Southern Electronics Corporation in Opelika, AL, USA. The cylindrical camera (Watec WA502A with zoom pinhole lens) measured 15 cm 3.5 cm 3 cm. The 1.2 cm 0.8 cm microphone (model SOU 101AUD) was connected to a Panasonic AG505 VCR/monitor by a 240-cm cable. The videotaped ADL interactions were later coded on a laptop computer running the Portable Computer Systems for Observational Research software (Communitech International, DeKalb, IL, USA). This software assigns keys on the laptop computer to record behaviours of the investigator s speci cation and has been used in previous observational studies both in the nursing home and in the community. 10,11 The observation system was designed to assess ADL-related activity, caregiver physical assistance, caregiver verbal interaction, assistive device use, disruptive behaviour exhibited by the patient, and patient participation in the ADL. All behaviour codes were de ned to match as closely as possible any speci cations given in the FIM. Procedure After the initial telephone screening, the investigator met with eligible dyads in the home to discuss the informed consent process, administer the paper-and-pencil measures, and determine an observation schedule. The caregiver gave informed consent for the patient as well as himself or herself prior to beginning any data collection. Although the patients were not able to provide informed consent due to their cognitive impairment, they were told about the nature and purpose of the study and given the opportunity to refuse participation at any time. Although attempts were made to observe each dyad s ADLs on three separate days to obtain a representative sample of patient abilities, various factors such as caregiver time limitations resulted in ADLs sometimes being observed fewer than three times per dyad. Out of a possible 63 observation times per ADL across all dyads, bathing was observed 52 times; dressing above the waist, 56 times; dressing below the waist, 58 times; eating, 55 times; grooming, 53 times; toileting, 53 times; and transferring, 58 times. Each dyad s observations were completed within a four-week period to ensure that no signi cant declines in patient functioning occurred. During observation, the caregiver and patient were encouraged to ignore the observer and conduct ADLs as usual. To decrease observer reactivity, the investigator attempted to remain out of the room during ADL performance. However, due to spatial limitations, this arrangement was not always possible. When the observer was unable to videotape the ADLs from another room or out of sight of the dyad, she sat or stood in an unobtrusive location and did not initiate verbal interaction or eye contact with the dyad. Subjects objected to the video camera on only ve different occasions, in which cases videotaping of that speci c ADL was either ceased or modi ed such that the caregiver and patient were comfortable with the proceedings. Following the observation sessions, two independent, trained raters coded the videotaped

4 ADLs. To prepare the tapes for coding, the time counter on the VCR was used to identify the time span during which each performance of each ADL occurred, and these time spans were noted on an external log. An ADL session was coded by advancing the videotape to the start of the ADL as documented on the log and activating the keys on the observational system in a predetermined order, with a set interval during the press of each key. Once these start-up keys were activated, the elements of the ADL interaction were coded as they occurred. Although it was initially intended to assess inter-rater reliability during 15% ( minutes) of the total observation time, this gure was actually closer to 18% ( minutes) due to the need to obtain a representative sample of data from each dyad. Inter-rater reliability was calculated using Cohen s kappa, a measure of reliability that adjusts for chance agreements based on a second-by-second comparison of the les of the two observers. 10 Kappas for the codes ranged from 0.67 to 0.88, with a mean of 0.78, indicating acceptable inter-rater reliability for all codes. Data analysis The computer-assisted data collection system generates a stream of data for each dyad during each block of observation time. This data stream consists of a behaviour code and that code s start and stop time, both recorded in seconds. The percentage of total time during which a behaviour code was observed during each ADL was calculated for each dyad by adding the amount of time a particular behaviour was recorded during the observation sessions and dividing this sum by the total duration of the observation sessions. This mean amount of time during which a particular behaviour was recorded was computed for each dyad and entered into analysis. Based on the presence or duration of relevant events (e.g. verbal prompts or assistive device use; see Table 1), the observational data were converted into ordinal designations corresponding directly to the FIM categories. The data used to create the observation-derived FIM scores represent the averaged per centages, durations and occurrence rates across all observations of an ADL for a dyad. After the observation-derived FIM correspondence 39 FIM scores were computed, they were correlated with the caregiver-reported FIM scores using Spearman s rho. Wilcoxon signed-ranks tests were conducted to examine mean differences. To examine the correspondence between caregiver-reported ADL assistance durations and the assistance durations observed during the ADL interactions, a variable called observed caregiver assistance duration was created by averaging for each subject and each ADL the durations of the keys on the observational system that corresponded to the different types of caregiver assistance denoted in the FIM: supervision, setup, hands-on assistance, and verbal prompting. Any overlap of verbal prompting with supervision, setup and hands-on assistance was identi ed and accounted for in the raw data. The observed caregiver assistance durations were correlated with the caregiver-reported assistance durations using Pearson correlation coef cients. Wilcoxon signed-ranks tests were performed on the caregiver-reported and observed assistance duration means for each ADL. The decision tree format of the FIM s scoring system presented a unique situation for analyses of assistance duration. Speci cally, caregivers who reported a FIM score of 6 or 7, denoting patient independence in that ADL, were not asked the follow-up question pertaining to ADL assistance duration. These data are not missing in the traditional sense, because patient independence connotes that the caregiver did not assist with the ADL and therefore spent no time in assistance-related activities; i.e. it can be inferred that the caregiver-reported assistance duration is zero. However, it was felt that preserving the data as they were collected could yield useful information regarding the validity of caregivers assistance time estimates. Thus, two sets of analyses were conducted to investigate the relationship between caregiver-reported and observed assistance duration. First, caregivers who had reported a FIM score of 6 or 7 for a particular ADL were assigned a caregiver-reported assistance duration of zero and included in the analyses. These analyses will be discussed as representing the entire sample (all subjects) and are based on N = 21 unless otherwise noted. In a second set of analyses, these same caregivers were excluded, such that only the subsample of

5 40 EM Cotter et al. Table 1 Correspondence of FIM levels of function to observational system de nitions FIM level FIM de nition Observational system de nition (numerical score) Complete All tasks which compose the No assistive devices used independence (7) activity are performed safely, No concern for safety expressed within a reasonable time, and Caregiver does not supervise, without modi cation, assistive prompt, or touch patient devices, or help from another No setup of materials provided person Patient does not take more than three times as long as normal to perform ADL Modi ed Patient requires an assistive device, Caregiver does not supervise, independence (6) the activity takes more than prompt, or touch patient reasonable time to perform, or there No setup of materials provided are safety considerations Assistive device used, concern for safety expressed, and/or patient takes three times as long as normal to perform ADL Minimum Patient requires no more help than Patient spends at least 75% assistance (4) touching and expends 75% or more of ADL time performing the task of the effort Moderate Patient requires more help than Patient spends 50 75% of ADL assistance (3) touching or expends 50 75% of the time performing the task effort Maximum Patient expends 25 50% of the Patient spends 25-50% of ADL assistance (2) effort time performing the task Total assistance (1) Patient expends less than 25% of Patient spends less than 25% of the effort ADL time performing the task caregivers who had reported a FIM score of 5 or below were included. Because the number of caregivers who reported a FIM score of 5 or less varies with each ADL, the number of dyads examined in this second set of analyses varies accordingly. Results A total of minutes of observational data were collected from the 21 dyads in this study. Of the total observation time, 996 minutes were devoted to eating, to bathing, to dressing below the waist, to grooming, 180 to dressing above the waist, 144 to toileting, and 40.8 hours to transferring. Correspondence of caregiver-reported and observation-derived FIM scores Correlations between caregiver-reported and observation-derived FIM scores are shown in Table 2; means are in Table 3. All ADLs were observed at least once per dyad (i.e. N = 21) with the exception of grooming (N = 19) and transferring (N = 20). The correlations for each ADL are based on only those dyads for whom both caregiver-reported and observation-derived data exist. All correlations between caregiverreported and observation-derived FIM scores were statistically signi cant at p < Wilcoxon signed-ranks tests indicated that for six of the seven ADLs, the difference between the caregiver-reported and observation-derived FIM score was not statistically signi cant. The one exception was for transferring, Z(20) = 2.42, p =.0014.

6 FIM correspondence 41 Table 2 Correlations between caregiver-reported and observation-derived ADL data ADL Correlation between Correlation between Correlation between reported and reported and reported and observation-derived observed assistance times observed assistance times FIM scores (all dyads) (subsample) Bathing 0.904*** 0.506* a Dressing above the waist 0.909*** b Dressing below the waist 0.818*** 0.564** 0.493* a Eating 0.862*** c Grooming 0.620*** a a c Toileting 0.858*** 0.586** d Transferring 0.701*** b 0.560* b e N = 21 unless otherwise noted. All correlations are Spearman s rho. a N = 19. b N = 20. c N = 15. d N = 13. e N = 10. ***p < **p < *p < Table 3 Descriptive statistics for caregiver-reported and observation-derived FIM scores ADL Mean Standard deviation Median Interquartile range Bathing Reported Observation-derived Dressing above the waist Reported Observation-derived Dressing below the waist Reported Observation-derived Eating Reported Observation-derived Grooming a Reported Observation-derived Toileting Reported Observation-derived Transferring b Reported Observation-derived N = 21 unless otherwise noted. a N = 19. b N = 20.

7 42 EM Cotter et al. Table 4 Descriptive statistics for caregiver-reported and observed ADL assistance times ADL Reported assistance Reported assistance Observed assistance time (subsample) time (all dyads) time (all dyads) Mean Standard Mean Standard Mean Standard deviation deviation deviation Bathing 19.27*** a *** Dressing above the waist 4.30* b Dressing below the waist 6.68** a * Eating c Grooming 11.67* c a 2.10 Toileting 8.77*** d * Transferring 1.40 e b 0.87 N = 21 unless otherwise noted. All times are reported in minutes. a N = 19. b N = 20. c N = 15. d N = 13. e N = 10. ***Signi cantly different from observed assistance time at p < **Signi cantly different from observed assistance time at p < *Signi cantly different from observed assistance time at p < Correspondence of caregiver-reported and observed ADL assistance durations Because caregivers were asked to report their ADL assistance time in minutes, it was believed possible that caregivers may have in ated their time estimates for ADLs of brief duration (e.g. transferring). To correct for this possibility, 1 minute was subtracted from each reported assistance duration in an attempt to prevent upwards bias. Although subtracting 1 minute from all reported times is a conservative overcorrection, it was felt that this approach would eliminate bias. The results of both sets of correlational analyses are shown in Table 2, and the corresponding means of the caregiver-reported and observed assistance durations for all subjects and the subsample can be found in Table 4. In general, the correlational analyses indicated that the caregiver-reported assistance durations were signi cantly correlated with the observed assistance duration when caregivers who reported providing no assistance were inferred to have a reported assistance duration of zero (all subjects). Four of the seven ADLs (bathing, dressing below the waist, toileting and transferring) demonstrated these signi cant relationships. However, when correlations were computed only for those caregivers who had reported providing assistance with an ADL during the past week (subsample), only one of the seven ADLs (dressing below the waist) demonstrated a signi cant correlation between caregiver-reported and observed assistance duration. The relationship for toileting approached statistical signi cance (r = 0.491, p = 0.088). All caregiver-reported assistance durations were larger than the observed assistance durations, with these differences being statistically signi cant for bathing, dressing below the waist, and toileting in the entire sample. In the subsample, the differences between the caregiver-reported and observed assistance durations were signi cant for all ADLs except eating and transferring. On average, caregiver-reported assistance durations were two to three times the length of the observed assistance durations. For toileting and transferring, the magnitude of difference was even larger, with caregivers overestimating their ADL assistance time by a factor of as much as six. Discussion As stated previously, the purpose of this study was twofold: (a) to determine the correspon-

8 Clinical messages Although caregivers proxy reports of the nature of ADL assistance they provide do appear to be valid, the extent to which they re ect actual patient limitations and needs is unclear. Caregivers reports of the duration of ADL assistance appear to be systematically in ated and should be used with caution. It is possible to obtain information regarding ADL performance by directly observing ADLs conducted in the natural environment; reliance on observation of analogue tasks or IADLs is not mandatory. dence between caregiver reports of ADL dependence and observed functioning and (b) to determine the correspondence between caregiver estimates of ADL assistance time and observed ADL assistance time. In this study, dementia caregivers descriptions of the extent of their ADL assistance were similar to what was exhibited during ADL interactions conducted in the home. Correlations between caregiver-reported and observation-derived FIM scores were significant and positive for all seven ADLs, indicating that this correspondence was not dependent upon a speci c ADL or type of ADL (e.g. long ADLs involving several steps, such as dressing, versus brief ADLs involving relatively fewer steps, such as transferring). Examination of mean differences indicated that for all ADLs except transferring, there were no signi cant differences in the mean caregiver-reported or observationderived FIM scores. These ndings are con rmatory of other studies that have shown a high consistency between health professionals FIM ratings and self-reported FIM ratings. 12,13 The results also indicated that, for four of the seven ADLs, the caregiver-reported assistance duration was signi cantly correlated with observed assistance time when caregivers who reported providing no assistance were inferred to have a reported assistance duration of zero. When these caregivers were removed from analysis such that correlations were computed only for those caregivers who had reported providing FIM correspondence 43 assistance with an ADL during the past week, only one of the seven ADLs demonstrated signi cant correlations between caregiver-reported and observed assistance duration. However, examination of the respective means for both the overall sample and the subsample indicated that caregivers consistently overestimated their ADL assistance time, although the degree of overestimation (and therefore its relative signi cance) varied depending on the ADL and whether the caregivers who had reported providing no assistance were excluded from analysis. Overall, caregivers overestimated their ADL assistance time by a factor of two or three, although the magnitude of difference was larger for some ADLs. Thus, the results of the correlational analyses are misleading without additional knowledge of the caregiver-reported and observed assistance duration means. If only the correlational analyses were examined, it would appear that caregivers described the duration of their ADL assistance somewhat accurately. However, when the mean caregiver-reported and observed assistance durations are examined, it becomes clear that the caregivers in this sample markedly overestimated their ADL assistance time. The significant correlations indicate that their patterns of overestimation were more consistent for some ADLs than for others. Previous research has demonstrated that caregivers of Alzheimer s disease patients display signi cant burden related to their caregiving duties. 14,15 It is reasonable to expect burdened caregivers to overestimate the duration of their ADL assistance due to the frequency of and stress associated with providing assistance, as well as the stressful nature of caregiving in general. It should be noted that caregivers might provide several types of assistance (e.g. both verbal prompting and supervision) during a single ADL. When estimating the duration of their assistance, caregivers may consider the durations of the different types of assistance in an additive or even a multiplicative fashion; in other words, not considering that verbal prompting and supervision occur simultaneously, caregivers may in ate their assistance time estimates. The cognitive processes used by caregivers in making judgements of their ADL assistance time should be

9 44 EM Cotter et al. explored in more detail in future studies. Despite the strong relationship between caregiver-reported and observation-derived FIM scores, these scores cannot be considered to represent patients true abilities. In other words, these data only re ect what caregivers allow the patients to do, rather than what the patients are truly able to do. In this manner, caregivers biases and perceptions involving patient ability may be re ected not only in their ratings of ADL performance but also in their behaviour during ADL interactions. Statements regarding the functional abilities of dementia patients should thus be phrased and interpreted cautiously until more conclusive data regarding the relationship between potential and actual independence can be obtained. Although caregiver reports of ADL assistance time are used to evaluate the economic impact of Alzheimer s disease and thereby in uence public policy, these ADL assistance time estimates had not previously been compared with objective measures. The results of the current study suggest that caregiver time estimates may be systematically in ated. Although more information regarding the nature and cause of this discrepancy is still needed to determine the source of the disagreement, it appears that these time estimates should be used with caution. Given the sensitive nature of the data collection, it is necessary to address any ethical and practical concerns that may occur. The study received IRB approval prior to any data collection, and informed consent procedures were followed strictly. Both caregivers and patients were given the opportunity to refuse or stop participation at any point during the study. No attempt was made to deceive either the caregivers or the patients regarding the method of data collection, although several measures were taken to minimize reactivity, such as using a relatively unobtrusive video camera and having the investigator refrain from initiating interaction during observations. Caregivers were also asked, informally, whether the patient s behaviour differed considerably from normal during the observations; this request was made to ensure that the patient was not overly distressed by the observations and to determine as closely as possible that differences in reported and observed ADL data were not due to observer reactivity. The small sample size constitutes a noteworthy limitation of the study. Sample size was limited somewhat by the need to keep the observation schedule manageable and ensure that the dyads observations were always conducted by the same person. Although a large quantity of data were collected for each caregiver patient dyad in an effort to compensate for the small sample size, the generalizability of the results is hampered to some degree. Additionally, it should be noted that the sample drops to a very small size for some of the correlations, which may explain at least in part the poor correspondence between the reported and observed ADL assistance durations. However, the signi cant ndings were robust and indicate consistent patterns, providing some con dence in the utility of these ndings. In conclusion, one can draw three inferences from the results of the study. First, caregivers of dementia patients can accurately describe the amount of assistance they provide during ADL interactions. These caregiver-reported data were validated by direct observation of ADLs and also indicate that the FIM s seven-point rating scale and operational de nitions have good external validity and are appropriate for use with dementia patients. Second, comparison of caregiverreported and observed ADL assistance times indicated that caregivers overestimate the duration of their ADL assistance time, sometimes by a factor of three or more. Thus, caution should be used when deriving and using caregivers time estimates. Additional research is needed to document the nature and source of these discrepancies. Third, this study also demonstrated that videotaping ADL interactions in the home is feasible and, by incorporating dementia patients as its subjects, added to the existing knowledge regarding functional performance in this group. However, other issues remain unanswered, most importantly the relationship between the assistance provided and the assistance actually needed to maximize patient independence and, hopefully, help decrease caregiver distress. Acknowledgements This research was conducted while the rst author was a graduate student at the University of Alabama at Birmingham. Financial and

10 FIM correspondence 45 material support were provided by the Center for Aging and Civitan International Research Center at UAB, and the Applied Gerontology Program at the University of Alabama. For assistance with reliability coding, data analysis, and subject recruitment respectively, the authors wish to thank Kay Scilley, John Gerstle and David Vance and the UAB REACH Project. The authors also appreciate the editorial comments of two anonymous reviewers, Karlene Ball, Edward Taub and Gary Fisk. References 1 Eakin P. Problems with assessments of activities of daily living. Br J Occup Ther 1989; 52: Albert SM, Sano M, Bell K, Merchant C, Small S, Stern Y. Hourly care received by people with Alzheimer s disease: Results from an urban, community survey. Gerontologist 1998; 38: Skurla E, Rogers JC, Sunderland T. Direct assessment of activities of daily living in Alzheimer s disease: a controlled study. J Am Geriatr Soc 1988; 36: Folstein MF, Folstein SE, McHugh PR. Mini-Mental State. J Psychiatr Res 1975; 12: Granger CV, Hamilton BB, Keith RA, Zielezny M, Sherwin FS. Advances in functional assessment for medical rehabilitation. Top Geriatr Rehabil 1986; 1: Keith RA, Granger CV, Hamilton BB, Sherwin FS. The Functional Independence Measure: a new tool for rehabilitation. In: Eisenberg MG, Grzesiak RC eds. Advances in clinical rehabilitation, Volume 1. New York: Springer-Verlag, 1987: Segal ME, Gillard M, Schall RR. Telephone and inperson proxy agreement between stroke patients and caregivers for the Functional Independence Measure. Am J Phys Med Rehabil 1996; 75: Smith PM, Illig SB, Fielder RC, Hamilton BB, Ottenbacher KJ. Intermodal agreement of follow-up telephone functional assessment using the Functional Independence Measure in patients with stroke. Arch Phys Med Rehabil 1996; 77: Smith P, Hamilton BB, Granger CV. FIM decision tree: the Fone FIM. Buffalo, NY: Research Foundation of State University of New York, Burgio LD, Scilley K, Hardin JM et al. Studying disruptive vocalization and contextual factors in the nursing home using computer-assisted real-time observation. J Gerontol B Psychol Sci Soc Sci 1994; 49: P Cotter EM, Burgio LD, Hsu C, Hardin JM. Behavioral microanalysis of helping interactions between mothers and adult children with mental retardation: a comparison of young and older dyads. J Adult Dev 1998; 5: Fuhrer MJ, Rintala DH, Hart KA, Clearman R, Young ME. Relationship of life satisfaction to impairment, disability, and handicap among persons with spinal cord injury living in the community. Arch Phys Med Rehabil 1992; 73: Grey N, Kennedy P. The Functional Independence Measure: a comparative study of clinician and self ratings. Paraplegia 1993; 31: Zarit SH, Reever KE, Bach-Peterson J. Relatives of the impaired elderly: Correlates of feelings of burden. Gerontologist 1980; 20: Zarit SH, Zarit J. Families under stress: interventions for caregivers of senile dementia patients. Psychother Theory Res Prac 1982; 19: