Scaling ADLs Within the MDS

Journal ofgerontology: MEDICAL SCIENCES 1999. Vol. 54A. No. I I. M546-M553 Copyright 1999 by The Gerontological Societv ofamerica Scaling ADLs Within the MDS John N. Morris,1 BrantE. Pries,' and ShirleyA. Morris' 'Hebrew Rehabilitation CenterforAged, Boston, Massachusetts. 2Institute of Gerontology, Schoolof Public Health,University of MichiganandAnnArborVAMedicalCenter,AnnArbor. Background. Dependency in activities ofdaily living (ADLs) is a reality within nursing homes, and we describe ADL measurement strategies based on items in the Minimum Data Set (MDS) and the creation and distributional properties of three ADL self-performance scales and their relationship to other measures. Methods. Information drawn from four data sets for a multistep analysis was guided by four study objectives: (I) to identify the subcomponents of ADLs that are present in the MDS battery; (2) to demonstrate how these items could be aggregated within hierarchical and additive ADL summary scales; (3) to describe the baseline and longitudinal distributional properties of these scales in a large, seven-state MDS database; and (4) to evaluate how these scales relate to two external criteria. Results. Prevalence and factor structure findings for seven MDS ADL self-performance variables suggest that these items can be placed into early, middle, and late loss ADL components. Two types of summary ADL self-performance measures Were created: additive and hierarchical. Distributional properties of these scales are described, as is their relationship to two external ADL criteria that have been reported in prior studies: first as an independent variable predicting staff time involved in resident care; second as a dependent variable in a study of the efficacy of two programs to improve resident functioning. Conclusions. The new ADL summary scales, based on readily available MDS data, should prove useful to clinicians, program auditors, and researchers who use the MDS functional self-performance items to determine a resident's ADL status. ACTIVITIES of daily living (ADLs) are among the most J-\.basic building blocks of life and are central to our ability to comprehend the experience of residents in nursing homes (1,2). Although ADL dependency is rare in the community, it is pervasive in the nursing home; and once in a nursing home, the specter of continued decline sets the tone that drives much of nursing and rehabilitative care. In U.S. nursing homes, the measurement of ADLs is based on the items in the mandated Minimum Data Set (MDS) assessment instrument-an integrated information battery with many dimensions and associated summary measures (1,3-12). The individual ADL items in the MDS have been shown to cross-walk with elements in the Functional Independence Measure (AM) rehabilitation tracking system (13), to be powerful predictors of resource utilization in the Resource Utilization Groups (RUG-III) system recently adopted as part of the federal Medicare Prospective Payment System (14), and to have high interassessor reliability values (6,9). For those who desire to track a single ADL area, e.g., locomotion or eating, the current scoring of the individual MDS ADL items is quite satisfactory. There are five progressive levels of dependency, and given the demonstrated reliabilities for these items, even a single category change in status can have real meaning. At the same time, for the vast majority of applications, there is a need to go beyond the resident's status in a single ADL area. Clinical and programmatic initiatives are almost always focused on a broader conceptualization of the self-performance status of the resident. What is needed is a system for summarizing the individual ADLs. Is the resident improving or declining in some holistic sense? Is a program effective or ineffective in improving resident status or slowing down decline? These are crucial questions, and in this article we demonstrate how the MDS ADL items can be combined to create ADL summary measures, how these measures distribute in the nursing home setting, how they change over time, and how they relate to two external criteria. Under the MDS system, trained clinical professionals assess resident performance over a 7-day period. Each ADL has an explicit performance-based definition, and cues are provided in the User's Manual to help staff determine who should be questioned, what to observe, and what records to review (15). The MDS ADL performance areas are broad in scope. Included are items that assess the last remnants of the individual's continued involvement in personal activities-eating and body movement while in bed-the "late-loss" ADLs. Also included are items that look at higher levels of functioning, representing tasks that begin to decline prior to entering a nursing home-e.g., dressing and personal hygiene-the "early-loss" ADLs (16). ADL classification schema were initially popularized by Katz (17,18); over time, many have commented on the aspects of personal care that should be included in an ADL battery and how these items should be scaled (16). Among the widely known scales, in addition to the Katz ADL, are the Barthel and FIM (13,19,20); and investigators have commented continually on how best to summarize these items (21-23). In this article information is presented on how the MDS ADL items can be combined to form the two types of summary ADL measures that are typically found in the literature: (i) a single, functionally meaningful, hierarchical ADL self-performance rating scale (the MDS ADL Self-Performance Hierarchy); and (ii) two versions of additive ADL scales based on the same item M546

SCALING ADLs IN THE MDS M547 pool (the MDS ADL-Short Form and the MDS ADL-Long Form). One can identify advocates for each type of measurement strategy. Hierarchical systems permit precise specification of discrete impairment levels, whereas additive systems tend to be sensitive to minute shifts in resident status; and both types of systems can be used to detect shifts in status at a programmatic level. METHODS Samples Data from four samples are reported in this article. The first sample consists of 187 residents from 21 facilities for whom there are dual, independent, reliability assessments of each ADL item (6). The second consists of 175,920 MDS assessments from all nursing homes in a seven-state area. This sample of residents in 1994 was used to identify the MDS ADL items that fall into early, middle, and late-loss categories, and to model how these items can be brought together within summary ADL scales. The two final samples include specialized data that help us understand how well the new ADL summary scales could replace the ADL measures used in the earlier studies (14,24). In one sample, we evaluate how well the MDS ADL summary scales measure ADL change in a rehabilitation intervention trial (24). This sample includes 389 residents from six nursing facilities that participated in a trial of exercise and nursing-based rehabilitation intervention. For this sample, four facilities were randomly designated as experimental sites; two were controls. Using the last of the four samples, we evaluate how well the ADL measures predict the staff-time resources---case mix-in nursing homes. We contrast the new measures with each other and an additive ADL index (14) that is used in the Resource Utilization Groups (RUG-III) system recently adopted as part of the federal Medicare Prospective Payment system. This sample includes 2204 residents from seven states. ADL Self-Performance Items in the MDS Figure 1 lists the ADL items used in this study. It is important to note the scope of the item definitions, as there can be dramatic differences across instruments (16). For example, note that the dressing item requires that the person be in street clothes, and staff were required to ask questions pertaining to a wide spectrum of subtasks for this and all ADLs. Finally, note that a resident who was independent in one aspect of bed mobility could be dependent in that ADL if he/she received extensive assistance in another aspect of the activity. Figure 2 presents the definition for the single, unified ADL coding system. There are six response alternatives, from independent to dependent, and one category for recording that the activity did not occur (e.g., there was no transfer, the patient was bed-bound). For the purpose of analysis, for each ADL, a response of the "activity did not occur" was converted to the code representing "total dependence." This coding scheme identifies whether the activity was or was not performed over the measurement period, the types of assistance provided, the frequency of task occurrence, and the mix of self-performance and physical support. This scoring scheme also recognizes that ADL patterns are not necessarily fully consistent over time, and the MDS coding schema attempts to record a balanced view of this diversity. For example, codes 0, 1, 2, and 3 (independent, supervision, limited assistance, and extensive assistance) are defined to permit one or two exceptions for the provision of heavier care. Permitting such exceptions has been shown to increase the average interassessor weighted kappa reliabilities by some 10 to 15% (25,26). Analytic Strategy The analytic activities have four objectives: (i) to identify the ADL subcomponents present in the MDS; (ii) to demonstrate how to aggregate these items in hierarchical and additive scales; (iii) to describe the scale distributions at baseline and over time; and (iv) to evaluate how these scales relate to two external criteria. Four steps were involved. In step 1, two factor analyses with oblique rotations were reviewed to provide an initial, preliminary indication of how the ADL items may be arrangedwith each factor representing the coalescing of items with high shared correlational and covariance structures. One solution was based on a cross-sectional, operational form for each of the ADL items; the second was based on a change score over 90 days for these same items. In step 2, we studied further how these items could coalesce into distinct ADL components. DRESSING: How resident puts on, fastens, and takes off all items of street clothing, including donning/removing prosthesis. PERSONAL HYGIENE: How resident maintains personal hygiene, including combing hair, brushing teeth, shaving, applying makeup, washing/drying face, hands, and perineum (EXCLUDE baths and showers). TOILET USE: How resident uses the toilet room (or commode, bedpan, urinal); transfer on/off toilet, cleanses, changes pad, manages ostomy, or catheter, adjusts clothing. LOCOMOTION ON UNIT: How resident moves between locations in his/her room and adjacent corridor on same floor. If in wheelchair, self-sufficiency once in chair. TRANSFER: How resident moves between surfaces-to/from: bed, chair, wheelchair, standing position (EXCLUDE to/from bath/toilet). BED MOBILITY: How resident moves to and from lying position, turns side to side, and positions body while in bed. EATING: How resident eats and drinks (regardless of skill). Includes intake of nourishment by other means (e.g., tube feeding, total parenteral nutrition). Figure I. ADL self-performance items.

M548 MORRIS ETAL. Using cross-sectional data, we determined which ADLs residents first moved from the independent to a nonindependent status, and which ADLs residents were last able to retain an independent status. The factor solutions were expected to separate items based on a hierarchy of loss, with early loss ADL items coming together as distinct from middle and late loss ADLs. We also believed that the solutions using cross-sectional and changeo. INDEPENDENT-No help or oversight-or Help/oversight provided only 1 or 2 times during last 7 days 1. SUPERVISION-Oversight, encouragement, or cueing provided 3 or more times during last 7 days-or-supervision (3 or more times) plus physical assistance provided only 1 or 2 times during last 7 days 2. LIMITED ASSISTANCE-Resident highly involved in activity; received physical help in guided maneuvering of limbs or other nonweight bearing assistance 3 or more times-or-more help provided only 1 or 2 times during last 7 days 3. EXTENSIVE ASSISTANCE-While resident performed part of activity, over last 7-day period, help of following type(s) provided 3 or more times: -Weight-bearing support -Full staff performance during part (but not all) of last 7 days 4. TOTAL DEPENDENCE-Full staff performance of activity during entire 7 days 8. ACTIVITY DID NOT OCCUR during entire 7 days Figure 2. ADL Self-performance coding schema (code for resident's performance over all shifts during last 7 days-not including setup--which is coded separately). score operational terms might vary slightly-that the same dynamic might not be seen (16). The factor solutions gave one view of the progression from early to late loss, and the analysis in step 2 provided a slightly different view. Together, these two sets of analyses helped us finalize the assignment of ADLs into early, middle, and late loss components. In step 3, we described the hierarchical and additive ADL summary scales. For the additive scales, KR 20 alpha reliabilities provided an indication of the consistency of the item relationships. In addition to describing these scales, we examined how these scales performed when compared to the results of previously reported ADL scales used in explaining resource utilization in nursing homes and in identifying changes in functional performance in an experimental exercise and rehabilitation intervention. Finally, in step 4, we presented information on how these ADL measures changed over 3- and 12-month periods. RESULTS Reliability and Distributional Properties ofadl Items Table 1 presents information on the interassessor reliability and population distributions for the ADL items. The weighted kappas are all above the.75 threshold, considered to be evidence of excellent reliability (27). Across the ADL items, three different distributional patterns are seen: (i) for dressing, personal hygiene, and toilet use, 20% or less of the residents are in each of the three least dependent categories; about 20 to 25% receive extensive assistance, and about 40% are totally dependent; (ii) for locomotion and transfer, the distribution is bimodal in character; 20 to 30% are independent, and 33 to 41% are totally dependent; and (iii) for bed mobility and eating, the distribution is also bimodal, but there are more independent residents (36% and 44%, respectively) than dependent residents (about 20%). Factor Analysis ofadl Items The seven ADL items were factored and subjected to an Oblimin rotation. For the cross-sectional measures, and restricting attention only to items with a factor loading of at least.40, three factors emerged: Early Loss: dressing (-.86), personal hygiene (-.94), and toilet use (-.74) Table 1. MDS ADL Item Reliabilities and Distributions* Distribution in Seven-State Sample (N =175,920) Weighted Kappa % Independent % Supervision % Limited Assistance % Extensive Assistance % Total Dependence MDS ADL Items (N= 187) (0) (I) (2) (3) (4) Dressing.90 11.8 7.5 17.6 24.2 38.8 Personal hygiene.87 11.9 8.4 15.5 22.3 41.9 Toilet use.93 20.0 6.4 13.0 19.6 41.0 Locomotion on unit.92 32.0 10.6 13.0 11.3 33.0 Transfer.91 24.7 7.4 16.6 20.3 31.1 Bed mobility.91 44.4 6.4 13.4 14.4 21.3 Eating.94 36.8 24.2 10.4 8.4 20.1 *Responses of "activity did not occur" (code 8) were recoded to "total dependence" (4).

SCALING ADLs IN THE MDS M549 Middle Loss: transfer (.69), locomotion (.70), bed mobility (.91) Late Loss: eating (.87) A slightly different three-factor solution was seen for the analysis based on 90-day item-change measures. There were more items in the Middle Loss factor, only one element in the Early Loss factor, and the Late Loss factor included both bed mobility and eating. Early Loss: personal hygiene (.89) Middle Loss: dressing (.73), transfer (.81), locomotion (.70), and toilet use (.83) Late Loss: bed mobility (.76) and eating (.78) Hierarchical Profile ofadl Loss To understand further the nature of the item configuration that would best represent early, middle, and late loss ADLs, we examined which ADLs remained independent longest as residents became less and less likely to maintain any residual areas of functional independence. In Table 2, the column headings represent the count of the number from 1 to 6 of the seven ADL areas in which the residents maintained independence. The rows represent each of the individual ADL items. The values in the cells represent the percentage of persons who remained independent for the ADL under the condition that there were only the indicated number of total areas in which the resident was still independent. Finally, the bolded items in the table represent each ADL for which there was a consistent indication that independence was lost as the number of total areas of independent functioning decreased. In Table 2, the early loss ADLs appear to be dressing and personal hygiene, the middle loss ADLs include three progressively deteriorating functions (toilet use, transfer, and locomotion), and the late loss ADLs are bed mobility and eating. From these analyses of ADL loss, we conclude that there are two early loss ADLs, three middle loss ADLs, and two late loss ADLs. Early Loss: dressing and personal hygiene Middle Loss: toilet use, transfer, and locomotion Late Loss: bed mobility and eating The middle loss ADLs can be further disaggregated into two clinically relevant categories: (i) toileting use and (ii) movement (transfer and locomotion). Table 2. Relationship Between the Count of the Number of Independent ADL Areas and Independence in Specific ADL Areas Independent in: ADLArea IADL 2ADLs 3ADLs 4ADLs 5ADLs 6ADLs Hygiene 0.0% 2.1% 4.8% 7.2% 11.4% 51.2% Dressing 0.1 0.3 1.4 3.8 11.4 70.4 Toilet use 0.5 3.3 15.2 48.1 94.3 96.6 Transfer 0.5 8.8 48.2 85.4 95.5 98.0 Locomotion 8.0 34.7 73.3 82.6 94.2 91.6 Bed mobility 34.7 72.1 94.7 97.8 99.6 99.9 Eating 56.2 78.7 62.3 75.1 93.6 92.4 Note: Bold data reflect where there is an indication that independence is lost as the number of total areas of independence decreases. Creation ofsummary ADL Scales Three summary ADL scales have been created. The first two are sums of selected ADL items. The MDS ADL-Long Form includes all seven of the ADL items, and the resulting scale has a range from 0 to 28. For the seven-state MDS sample, relevant scale statistics are as follows: The alpha (KR 20) internal consistency is quite high:.94. The overall scale distribution is relatively flat; only 7 k have a score of zero, 12.8% have a score of 28, the skewness value for extreme values is low (-.11), and the kurtosis value for peakedness is equally low (-1.26). The overall scale mean is 15.24, the median is 16.0, and the standard deviation is 9.25. The second summed scale, the MDS ADL-Short Form, includes only four ADL items, with one each from: early loss (personal hygiene); middle loss toileting, middle loss movement (locomotion), and late loss (eating). This scale has a range of 0 to 16. Its alpha (KR 20) measure of item internal consistency equals.90; the scale mean is 8.73, the median is 9. and the standard deviation is 5.36. The third scale is more complex. It is based on a synthesis of the most consistent specification of early, middle, and late loss ADL items as seen in the factor analyses and the hierarchy in Table 2. Figure 3 presents the scoring rules for the MDS ADL Self-Performance Hierarch}'. This seven-category scale employs the four ADL items used in the MDS ADL-Short Form. Within this seven-category ADL hierarchical measurement system, two categories represent relatively independent residents; one category represents residents with a limited impairment; two categories represent residents who receive extensive (weight-bearing) help; and two categories reflect more severe. total dependence. For the seven-state MDS sample, the two more independent categories encompass 16% of residents. whereas about 34% of residents fall into the two most depcn dent categories. Relation ofsummaryadl Scales to Two External Measures Table 3 examines the performance of the three ADL scales by predicting two external measures that have been reported elsewhere. First we predict the average daily minutes of care provided by nursing assistants. In this comparison, the MDS ADL-Long Form and the MDS ADL--Short Form explain approximately the same proportion of variance as does the RUG ADL; whereas the MDS ADL Self-Performance Hierarchy has a somewhat lower value. In the second half of Table 3, we evaluate how the three new ADL measures would perform if they were to take the place of the lo-item (40 point) ADL summary measure used as the outcome measure in a trial of the effectiveness of two types of rehabilitation services. In Table 3, the values represent the standardized differences between the mean rates of ADL change for the three experimental groups and the mean ADL change rate for the entire sample. Each of the three ADL scales provides results of the same magnitude, and with all three the results are statistically significant at a two-tailed alpha level of.0i or lower. Distributional Properties ofsummary ADL Scales To assist in understanding the magnitude of the change in ADL that can be expected in nursing homes, Table 4 presents

M550 MORRIS ETAL. Scoring rules-note the fouritems usedto score this scale are the same as the fouritems used to Percent of Residents scorethe MDS ADL-Short FormScale: personal Category Category in the Cross-State hygiene, toileting, locomotion, eating ScoreValue Label MDS Sample in Category AII4ADLs 0 Independent 8.6 MDS ADL-Short Form range> 0 AND Supervision 7.4 All fouradls < 2 All fouradls < 3 AND 2 Limited 13.2 One or moreof the four ADLs=2 Both eatingand locomotion < 3 AND 3 Extensive 1 24.9 Eitheror both of personal hygiene and toilet> 2 Eithereatingor locomotion =3 AND 4 Extensive 2 12.1 Neitherof these 2 ADLs =4 One or both of eating and locomotion =4 5 Dependent 17.7 All four ADLs =4 6 Total dependence 16.1 Figure 3. Scoring rules for the MDS ADL Self-Performance Hierarchy. Table3. Relationshipof ADL SummaryScalesto External Measures External Measure Being Compared MDS ADL-Long Form MDS ADL-Short Form MDS ADL Hierarchy RUG ADL ADL Sum of 10 Items Varianceexplanation of nursing time* 25.4% 23.2% Standardized differences for experimental groupst Controls Exercise Nursing rehabilitation.20 -.01 -.17.22.01 -.20 20.9% 24.8%.20.04 -.18.24 -.03 -.17 *Percentage of variance explained by ADL measures for average daily minutes of care resident received from nursing assistants. tstandard deviation difference between mean ADL change rate for the entire sample and the rates for the three experimental groups. information on longitudinal change rates over 3- and 12-month periods. Over 3 months there is an average decline of about 4% of one standard deviation unit; over 12 months the rate is about 13% of one standard deviation unit. As there is a large difference in the numberofpossible scale points across the three summary measures of ADL performance, it is not unexpected that there would be large interscale differences in the proportion of residents who either decline or improve over time. For example, using either the Short or Long Form scales, the proportions of persons who decline or improve are almost double the rates found for the ADL Hierarchy scale. Thus, although overall mean change is about the same, the Short and Long Form scales are more likely to detect singlepoint shifts in ADLs over time. Across all three scales, the rates for decline are higher than those for improvement, and the disparity increases over time. At the same time there are residents who improve, and their proportions also increase over time. It is also important to recognize that change in ADLs, even over 12 months, is seldom movement to either a status of total dependency or to a status of total independence. For example, for the ADL-Long Form, of those who declined over a 12 month period, the average change was 4.7 points, and 25.7% of those who declined changed by only 1 point. Ofthose who im-

SCALING ADLs IN THE MDS M551 proved, the average improvement was 3.3 points, and 35.2% of those who improved changed by only 1 point. The 12-month rates ofchange were relatively constant across all baseline scale scores for all three ADL summary sca1esafter first excluding persons who had topped out and could therefore not further improve, or bottomed out and therefore could not further decline. Using the ADL-Long Form as an example, about 40-45% of persons with a baseline score of0--3 declined over 12 months, about 50% of persons with a baseline score of 4-19 declined, and about 40-45% of persons with a baseline score of 20-27 declined. A similar pattern exists for improvement. Finally, for the ADL Hierarchy, when change was assessed over a 3-month period (which is the mandated time interval between MDS assessments), of the 17% of residents who either improved or declined, most changed by only a single point. Five percent had a I-point improvement and 7% had a l-point decline. DISCUSSION Using ADL functional performance items from the MDS, three new ADL scales are described-two additive, the other hierarchical. The items in these scales and the scales themselves are highly reliable, and these scales are capable of displaying resident self-performance levels along a continuum of selfinvolvement in personal activities ofdaily living. When tested in predictive and evaluative applications, these new measures have been shown to be capable of replacing other established measures. Thus, we conclude that these new scales have many applications and should prove useful to clinicians and researchers using the MDS. In clinical applications, the ADL-Long Form will be more successful than the ADL Hierarchy in identifying residents who may undergo more minor, incremental changes. For example, if we were to use the hierarchical measure to detect change over a 3-month follow-up, our data suggest that 10.5% would decline and 6.8% would improve. On the other hand, the expected proportion of residents who would be assessed as changing based on the ADL-Long Form would be more than twice as high 23.2% would decline and 15.1 % would improve. Thus, if the goal is to maximally identify residents whose ADL status is beginning to change, the ADL-Long Form is the measure of choice. However, there is a counter position. Both measures seem to be approximately equally sensitive to issues of system change, detecting whether there are differential rates of change in ADL status across large cohorts of residents. This counter-intuitive finding arises because of the relationship between the mean change level and standard deviation value for each measure. It is the two of these factors in tandem that sets the power of the effect estimate in an experiment, and in this case there is really little reason to choose between the two types of measures. The expected mean difference over time for the ADL-Long Form may be higher than the difference for the ADL Hierarchy, but this comes with a price. The standard deviation will also be much higher for the ADL-Long Form, and the significance based on the two resulting effect estimates will be quite similar. In addition, even from a clinical perspective, there may be situations in which one would prefer the ADL Hierarchy measure. With a hierarchical system, all shifts have immediate, substantive meaning--each of the code levels represents a distinct performance pattern. Such is not the case for a summary ADL measure, and there will be many applications in which clinicians are seeking less ambiguous guidance on what is happening to the resident. It is the ADL Hierarchy that can best respond to this need. Thus, although additive summary scores may be useful for describing overall function, they can obscure as much information as they reveal. Sometimes a change in only one key variable may be crucial in terms of understanding or tracking patient status, other times it may not; and it is the hierarchical measure that will best differentiate between these alternatives. The component parts of these scales conform to a general ADL hierarchy and we have identified the ADL items that are associated with early loss, middle loss, and late loss. Although there is some interindividual variation, the first ADL areas for which help is required are dressing and personal hygiene; conversely, the last areas of loss, where the resident is most likely to remain at least somewhat engaged in personal activities of daily living, are moving in bed and eating. The findings on change in the ADL scales over time point both to the challenge of preventing decline and the need to recognize that not everyone will decline and not all decline will be into a state of total dependency. Although many nursing home residents experience shifts in ADL status over a 3- to 12 month period, not all change involves decline. Over 3 months, 15% had improved, and by 12 months, 20% had improved. For nursing home residents, some decline results from transitory causes, and it is crucial for staff and families to understand that recovery can occur once the underlying acute problem is addressed (15). The challenge to minimize decline and Table4. ExpectedADL ChangeOver 3 and 12 Months Percent Change Time 1 in Standard to Time 2 Percent Percent Baseline Mean Follow-up Mean Mean Change Deviation Units Correlation Improved Declined 3-Month change ADL-Long Form 15.12 15.13 0.41 4.4.95 15.1 23.2 ADL-Short Form 8.66 8.89 0.23 4.3.95 12.3 18.8 ADL Hierarchy 3.39 3.46 0.07 3.9.93 6.8 10.5 12 -Month change ADL-Long Form 14.73 16.01 1.28 13.8.88 20.1 41.6 ADL-Short Form 8.44 9.15 0.72 13.4.88 17.5 35.7 ADL Hierarchy 3.32 3.55 0.23 12.6.88 10.6 23.2

M552 MORRIS ETAL. maximize improvement is clear. With an appropriate model, staff can affect the rates of change experienced by nursing home residents (24). For residents in nursing homes, a general maturational decline seems to be at work; rates of change are relatively constant along the entire ADL continuum. Shott of a precipitating event, there is a process at work much like that which we have previously reported for elders in the community (28). Irrespective of where one starts at baseline on ADLs, rates of decline and improvement are relatively constant. The MDS system and the new ADL scales described in this paper are germane to researchers and clinicians, not only in the United States, but in a wide diversity of other countries as well. The MDS is used in a large number of European countries, Canada, Japan, Korea, and Taiwan. The same scoring system is in use in nursing homes, home care, and acute in-patient mental health settings. As such, it is appropriate to ask how MDS-based scales such as the ADL measures relate to the World Health Organization (WHO) classification schema-international Classification of Impairment, Disability, and Handicaps [ICIDH (29,30)]. The World Health classfication schema of impairment, disability, and handicap has existed for almost two decades. The WHO is now testing ICIDH-2, which moves towards an impairment activities and participation schema from the perspective of the MDS, and particularly the ADL measures described in this article. They emphasize both more complex causal pathways and a concern for performance-based activities. The MDS-based measures fit nicely into the WHO conceptual model. To use the ICIDH terminology within the activity area, the MDS measures are based on a description of resident performance over specified periods of time. The basic tasks ofeveryday life that are relevant to the end of life are succinctly captured in these ADL measures. We assess actual performance, not the potential of performance; and we qualify our measures based on the level of resident involvement, the difficulty entailed, and the need for assistance. Also within the ICIDH-2 model, we do not necessarily imply that the functional performance of the resident is due to a single disease or disorder. Such causes do exist and are often predominant, but there are also contextual or environmental factors that can impact on the activities the resident performs and the level of supports provided by others. Finally, these internationally relevant ADL measures go a long way toward providing a consistent international language that describes function and will help us to understand the disease and motivational and environmental factors that impact on persons in nursing homes and other health care settings. The definitions of the items and the response alternatives for each item are systematic and functionally based. With these measures, we are able to establish a common language across countries. It is our hope that this wiii improve communication, facilitate comparisons from one country to another, and stimulate evaluations to determine how to best provide care to persons in nursing homes and other health care environments. ACKt\O\\LEDGME,nS Dr. Morris was supported, in part, by Grant AG 11719 from the National Institutes of Health, National Institute on Aging. Dr. Fries and Dr. Morris were supported, in part, by HS09455, an Agency for Health Care Policy and Research, Q-Span grant; and also acknowledge the support of interrai (an international group of clinicians and researchers who collaborate to promote research on the Resident Assessment Instrument and quality outcomes for eiders). Dr. Morris holds the Alfred A. and Gilda Slifka Chair in Social Gerontological Research. Address correspondence to John N. Morris, PhD, Hebrew Rehabilitation Center for Aged, HRCA Research and Training Institute, 1200 Centre Street, Boston, MA 02 I3 I-1097. E-mail: jnm@thor.hrca.harvard.edu REFERENCES I. Phillips CD, Morris.TN, Hawes C, et ai. Association of the Resident Assessment Instrument (RAJ) with changes in function, cognition, and psychosocial status..! Am Geriatr Soc. 1997;45:986-993. 2. lkegami N, Monis.TN, Fries BE. Low-care cases in long-term care setting: variation among nations. Age Aging. 1997;26:67-71. 3. Hirdes JP, Rook M. Quality applications of the Resident Assessment InstrumentIMinimum Data Set (RAIIMDS): a new era for quality improvement. Can'! Quality Health Care. 1998;14:3-4. 4 Phillips CD, Manis IN. The potential for using administrative and clinical data to analyze outcomes for the cognitively impaired: an assessment of the Minimum Data set for nursing homes. Alzheimer Dis Assoc Disord. 1998;11:162-167. 5. Brandeis GH, Baumann MM, Hossain M, et ai. The prevalence of potentially remediable urinary incontinence in frail older people: a study using the Minimum Data Set..! Am Geriatr Soc. 1997;45:179-184. 6. Monis.TN, Nonemaker S, Murphy K, et al. A commitment to change: revision of HCFA's RAl. JAm Geriatr Soc. 1997;45:1011-1016. 7. Sgadari A, Morris IN, Fries BE, et al. Efforts to establish the reliability of the RAl. AReAging. I997;26(Supp 2):27-30. 8. Resnick NM, Brandeis GH, Baumann MM, et ai. Evaluating a national assessment strategy for urinary incontinence in nursing home residents: reliability of the Minimum Data Set and validity of the resident assessment protocol. Neurol Urodynamics. 1996;15:583-598. 9. Hawes C, Morris IN, Phillips CD, et al. Reliability estimates for the Minimum Data Set for nursing home residents and care screening (MDS). Gerontologist. 1995;35: I72-178. 10. Mol' V, Branco K, FleishmanJ, et al. The structure of social engagement among nursing home residents..! Gerontol: Psycho!Sci. I995;50:PI-P8. I I. Hartmaier SL, Sloane PD, Guess HA, et al. Validation of the Minimum Data Cognitive Performance Scale: agreement with the Mini-Mental State Examination. J Gerontol:A BioI Sci Med Sci. 1994;50:MI28-MI33. 12. Monis.TN, Fries BE, Mehr DR, et al. MDS Cognitive Performance Scale. J Gerontol: Med Sci. 1994;49:M174--MI82. 13. Williams BC, Li Y, Fries BE, et al. Predicting patient scores between the Functional Independence Measure and the Minimum Data Set: development and performance of a FlM-MDS "Crosswalk." Arch Phvs Med Rehabil. 1997;78:48-54. 1-1-. Williams BC, Fries BE, Foley WI, et ai. Activities of daily living and costs in nursing homes. Health Care Finane Rev. 1994;I5: I 17-135. 15. Monis.TN, Nonemaker S, Murphy K, Long-Term Care Facility Resident Assessment Instrument (RAI) User's Manual. Briggs Health Care Products; 1995. 16. Monis IN, Morris SA. ADL assessment measures for use with frail elders. In: Teresi JA, Lawton Mp, Holmes D, Ory M, eds. Measurement in Elderly Chronic Care Populations. New York: Springer Publishing Company; 1997:130-153. 17. The Staff of Benjamin Rose Hospital. Multidisciplinary study of illness in aged persons, 1. Methods and preliminary results..! Chronic Dis. 1958;332-345. 18. Katz S, Ford AB, Moskowitz RW, et ai. Studies of illness in the aged: the index of ADL: a standardized measure of biological and psychosocial function. lama. 1963;185:914--919. 19. Hamilton BB, Laughlin JA, Fiedler RC, et ai. Interrater reliability of the 7 level Functional Independence Measure (FIM). Scand 1 Rehab Med. 1994;26:115-119. 20. Mahoney FI, Barthel DW Functional evaluation: the Barthel index. Md State Med J. 1965;14:61-65. 21. Lazaridis EN, Rudberg MA, Furner SE, et al. Do activities of daily living have a hierarchical structure? An analysis using the longitudinal study of aging..! Gerontal: Med Sci. 1994;49:M47-M51. 22. Heinemann AW, Linacre JM, Wright BD, et al. Relationship between im-

SCALING ADLs IN THE M DS M553 pairment and physical disability as measured by the Functional Independence Measure. Arch Phys Med Rehabil. 1993;74:566-573. 23. Spector WD. Takada HA. Combining activities of daily living with instrume ntal activ ities of dai ly livi ng to mea sure functiona l disability. J Gerontal: Soc Sci. 1997;53B:546-557. 24. Moni s IN, Fiatarone M, Kiely DK, et al. Nursing rehabilitation and exercise strategies in the nursing home. J Gerontol Med Sci. 1999;54A: M494--M500. 25. Moni s IN. Fries BE, Steel K, et al, Comprehe nsive clinical assessment in commun ity setting: applicability of the MDS- HC. JAm Geria tr Soc. 1997;45:1017-1024. 26. Morris IN. Hawes C. Fried B. et al, Designing the National Reside nt Assessment Instrument for nurs ing homes. Geron tologist. 1990;30: 293-307. 27. Rei ss JL. Statistical Me/hods for Rates and Proportions. 2nd cd. New York: JWiley: 1981. 28. Kiely DK. Monis IN, Monis SA. ct al. The effect of special medical conditions on functional decline. JAm Geria tr Soc. 1997;45: 1459-1463. 29. World Health Organization. lnternational Classification of Impairment. Disabilityand Handicaps. Geneva: World Health Organization: 1980. 30. World Health Organization. /CDlH -2. Geneva: World Health Organization; June. 1977. Received June 23. / 998 Accepted March 5. /999 The Oklahoma University Health Sciences Center DONALD W. REYNOLDS DEPARTMENT OF GERIATRIC MEDICINE Th e Department is searching for clinician educators who are board eligible or board certified in geriatrics to enhance educational programs within the College ofmedicine. The Donald W. Reynolds Department ofgeriatric Medicine at University of Oklahoma is committed to the development of significant geriatrics exposure for medical students in every year of medical school training. Individuals interested in joining this evolving educational endeavor should contact Marie A. Bernard, M.D., Professor and Chairman, 921 N.E. 13 th (1IG), Oklahoma City, OK, 73104, #405-297-5957, Fax # 405-270-5195. The University of Oklahoma is an equal opportunity institution.