Use of Physical Restraints and Psychotropic Medications in ... - NCBI

Use of Physical Restraints and Psychotropic Medications in Alzheimer Special Care Units in Nursing Homes

A B S T R A C T Objectives. This study analyzed the use of mechanical restraints and psychotropic medication in Alzheimer special care units (SCUs) in nursing homes. Methods. We analyzed 1993 data for more than 71000 nursing home residents in 4 states, including more than 1100 residents in 48 SCUs. The dependent variable in multinomial logistic regression was use of physical restraints or psychotropic medication. Models contained covariates representing facility and resident characteristics, and multivariate matching strategies were used to protect against selection bias. Results. Residents in SCUs did not differ from similar residents in traditional units in their likelihood of being physically restrained. Residents in SCUs were more likely to receive psychotropic medication. Conclusions. With regard to the measures used in this research, the findings indicate that residents in the SCUs in the 4 study states did not receive quality of care superior to that provided to similar residents in traditional units. In fact, the results related to drug use raise the question of whether some may have received poorer care. (Am J Public Health. 2000;90:92–96)

92

American Journal of Public Health

Charles D. Phillips, PhD, MPH, Kathleen M. Spry, PhD, Philip D. Sloane, MD, MPH, and Catherine Hawes, PhD Because nursing home residents with dementia present nursing facility staff with a number of special challenges, a new care modality, Alzheimer special care units (SCUs), has been developed. 1 In the last decade, these units have developed into an important segment of the nursing home industry. Alzheimer SCUs constitute two thirds of all SCUs in nursing homes, and recent estimates indicate that over 12% of US nursing facilities operate SCUs.2 However, a number of questions about the performance of SCUs remain unanswered.1,3,4 One question is how care in such units varies from that in traditional units. An important specific issue is whether SCU residents are less likely to be subjected to potentially dangerous interventions such as the use of physical restraints or powerful psychotropic agents. Mechanical restraints carry with them all of the potential medical problems one associates with immobility, as well as the possibility of negative changes in affect and of injury or death due to use of restraints.5–8 The list of potential negative sequelae for psychotropic medication is equally lengthy and disturbing: delirium, tachycardia, urinary retention, tardive dyskinesia, parkinsonism, falls, decreased mobility, negative affect, and increased cognitive impairment.9 The available evidence comparing the use of chemical and physical restraints in SCUs and other nursing home units is limited and somewhat contradictory. One study found residence in an SCU to be associated with a lower likelihood of physical restraint and no difference in the likelihood of pharmacologic restraints.10 Another study found lower levels of restraint use in SCUs but higher levels of antipsychotic drug use.11 However, both studies were carried out before full implementation of the Omnibus Reconciliation Act of 1987 (OBRA-87) requirements that emphasized reduced use

of physical restraints and antipsychotic medication in nursing homes. The available research indicates that these provisions resulted in a significant reduction in the use of restraints.12 Use of antipsychotic medication also seems to have decreased, while evidence on the use of other psychotropic drugs is mixed.13–19 Our research extends previous work by analyzing more recent data from a much larger number of SCUs. Also, unlike the units included in previous studies, the units used in this study were not chosen because of their reputed high quality of care, nor were they part of the initial wave of SCU development that may have more heavily represented higher-quality facilities.10,11 Finally, these data were gathered more than 2 years after the implementation of the OBRA-87 reforms, so the use of restraints and medication in all study facilities should reflect the post-OBRA, industrywide changes in care strategies.

Methods Data The resident-level data for this research were obtained from the Health Care Financing Administration’s (HCFA’s) Nursing

Charles D. Phillips, Kathleen M. Spry, and Catherine Hawes are with the Myers Research Institute, Menorah Park Center for Senior Living, Beachwood, Ohio. Philip D. Sloane is with the Department of Family Medicine, University of North Carolina at Chapel Hill. Requests for reprints should be sent to Charles D. Phillips, PhD, Myers Research Institute, Menorah Park Center for Senior Living, 27100 Cedar Rd, Beachwood, OH 44122 (e-mail: cphillips@myersri. com). This article was accepted August 20, 1999.

January 2000, Vol. 90, No. 1

Alzheimer Special Care Units

Home Case-Mix and Quality Demonstration. All of the assessment data were collected with an expanded version of the Minimum Data Set for Nursing Home Resident Assessment and Care Screening (MDS). 20 This instrument was used by all of the facilities in the 4 states involved in the demonstration: Kansas, Maine, Mississippi, and South Dakota. The resident-level data available for the study included all assessments in 1993 in all certified nursing homes in the 4 states. For the results presented here, the f irst full assessment performed during 1993 was used. Data were available for a total of 71748 nursing home residents. Forty-three percent of the nursing home residents in the data set resided in facilities in Kansas, 17% in Maine, 29% in Mississippi, and 11% in South Dakota. A total of 841 facilities that among them operated 48 SCUs were included in the study. Facilities that participated in the study allowed project staff to conduct data collection during site visits to their SCU. The SCUs involved in the site visits made up 85.5% of the SCUs in the 4 states. Information on the facilities in the 4 states was obtained from the HCFA’s Online Survey and Certification (OSCAR) system.2,21,22 Data specifically concerning the SCUs themselves were gathered during site visits by project staff.

Hypotheses We believe that “special care” for those with Alzheimer disease includes the promise of avoiding, to the degree possible, the use of potentially dangerous interventions such as physical restraints and psychotropic medication. Thus, we hypothesized that compared with residents in traditional units, SCU residents would be (1) less likely to be physically restrained, (2) less likely to receive psychotropic medication, and (3) less likely to receive psychotropic medication while they were also physically restrained.

Measurement All of the resident-level data derive from items included in the MDS. This instrument has been shown to provide high-quality data in studies where it is used by trained research nurses.23 Recent research has also indicated that MDS data in statewide databases contain high-quality, valid data in a number of domains.24 Previous analyses indicated that the data used in this research exhibit internal consistency equal to that found in MDS data collected by specially trained research nurses.25 Dependent variable. The dependent variable identified the following 4 groups of residents: January 2000, Vol. 90, No. 1

1. Residents who were not physically restrained and were not receiving psychotropic medication 2. Residents who were physically restrained but were not receiving psychotropic medication 3. Residents who were not physically restrained but were receiving psychotropic medication 4. Residents who were physically restrained and were receiving psychotropic medication Residents were defined as physically restrained if at any time in the 7 days before the assessment they were restrained with limb restraints, a trunk restraint, or a chair that prevented rising. Residents were considered to be receiving psychotropic medication if they received antipsychotics or sedatives/ hypnotics at any time during the 7 days before the assessment. Independent variables. Our independent variable of primary interest (i.e., “treatment” variable) had the following 3 categories: residents of an SCU (SCU), residents of a facility with an SCU but not of the SCU itself (SCU-F), and residents of a facility with no SCU (SCU-N). This variable included the category SCU-F to ensure that any general differences between facilities with and without SCUs would not be confounded with any differences related to residence in an SCU itself.4,11 On the basis of previous research, we included a number of covariates in the models.5–7,10,11,26,27 These included a resident’s age, sex, source of payment, length of stay, performance of activities of daily living (ADL), cognitive performance, history of falls, body control problems, and behavior problems. Other covariates reflected a facility’s ownership, location (i.e., state), and size.26 The covariates included 2 summary scales. The first, based on 7 MDS items, captured the residents’ cognitive function (α = .89). Previous research has shown that a similar MDS-based scale provides a reliable and valid measure of nursing home residents’ cognitive performance.28,29 The ADL summary scale (α = .90) was based on the number of ADLs (i.e., dressing, grooming, toileting, locomotion, transfer, and eating) in which residents needed physical assistance.

Analysis Strategy Multivariate models were estimated by multinomial logistic regression. Because our dependent variable had 4 categories, this procedure involved the estimation of 3 equations, each with parameter estimates for every independent variable. These equations esti-

mated the effects of the independent variables on the following: 1. The likelihood of being subjected to neither psychotropic medication nor physical restraints vs the likelihood of being physically restrained 2. The likelihood of receiving neither psychotropic medication nor physical restraints vs the likelihood of receiving psychotropic medication 3. The likelihood of receiving both psychotropic medication and physical restraints vs the likelihood of receiving neither Because residents of SCUs are a special population, difficulties may arise in comparing them with residents receiving care in other settings.4 To adjust for differences in the SCU and non-SCU populations under study, we included in our models all of the covariates noted earlier. As further protection against selection bias, we also classified residents as moderately or severely cognitively impaired, solely on the basis of their score on the cognitive scale, and estimated separate models for these 2 groups. In addition, a more complex strategy of matching by propensity scores was used.30 Residence in an SCU or a traditional unit was used as the dependent variable in a logistic regression that included all of the covariates noted above. This model then generated a propensity score, which is an estimated probability of being in an SCU for each resident in the sample. We then formed a comparison group by using only those SCU-F and SCU-N residents whom the model identified as most similar to residents of SCUs. Analyses were then performed that included only SCU residents and the comparison group composed of these “false positives.” This procedure has been used successfully in previous research with these data.31 Other research has indicated that differences in quality do exist among SCUs.32 To ensure that our results reflected patterns of care in the better SCUs, as well as in the “average” SCU, we also estimated our models with the SCU group composed only of residents in higher-quality SCUs. These SCUs were ranked on the Special Care Unit Environmental Quality Scale as being in the top one third of the study SCUs, based both on indicators of environmental quality and on measures focusing on staff–resident interactions. The scale (α = .77) has been validated against the Professional Environmental Assessment Protocol (r = .52) and in analyses of SCU quality.31,32 Finally, we determined whether rates of restraint and medication use were significantly different for SCU residents and SCU-F residents by estimating additional multinomial logistic regression models that American Journal of Public Health

93

Phillips et al.

used the same covariates and a slightly different form of the treatment variable. Where appropriate, these results are reported in the text as the probability that a significant difference existed between SCU and SCU-F residents. Although these data are for populations, tests of statistical significance were used because the data were subject to random measurement and processing errors. All models were estimated with SUDAAN, a statistical package that uses a robust variance estimator for clustered data.33 Variables in the multivariate models were considered statistically significant predictors on the basis of the variance ratio test for the variable as a whole. When a variable as a whole was significant, the relative odds ratios for each category of the variable were examined. Because of the number of models estimated, only those odds ratios with a probability of less than .01 were considered statistically significant.

Results The data used in these analyses came from only 4 states, so there may be some concern that the results are linked to a population very unlike that found across the nation. However, residents in the study facilities were quite similar to those residents observed in the 1996 Medical Expenditures Panel Study. Our 4-state population had slightly fewer residents younger than 65 years (7% vs 9%) but the same proportion of residents older than 84 years (49% vs 48%). Seventythree percent of the 4-state residents were women, compared with 72% of the national sample. The study population also contained only slightly fewer residents who were severely impaired in their decision-making ability (23% vs 25%). These results imply that members of the study population, though only from 4 states, were typical of nursing home residents in the nation as a whole.34 The same can be said of the facilities in these 4 states that had SCUs. According to 1993 data from the HCFA’s OSCAR system, the 48 facilities in the 4 states with SCUs were very similar to the 1543 facilities in the nation with SCUs. They varied little from the national population regarding ownership: 59% of the study facilities were for-profit enterprises compared with 62% of the 1543 facilities nationwide. The 4-state facilities averaged a 57% Medicaid census, while the national facilities averaged a 60% Medicaid census. Both groups of facilities reported approximately 4 residents per full-time nursing aide, but the 4-state facilities had slightly better nurse staffing, with a lower resident-tolicensed-nurse ratio (32 vs 35). The only 94


TABLE 1—Resident Characteristics by Type of Setting Type of Setting Resident Characteristics

SCU, % (n = 1175)

Female Age, y 84 Cognitive impairment None–very mild Mild–moderate Severe Behavior problems Wandering Verbally abusive Physically abusive Socially inappropriate Length of stay 90 days or less More than 90 days

SCU-N, % (n = 5572)

SCU-F, % (n = 68 242)

70

73

73

4 15 44 38

6 13 35 47

7 12 33 47

10 42 46

33 46 21

31 46 23

51 31 30 36

10 13 9 13

8 11 8 14

29 71

33 67

29 71

Note. SCU indicates residents living in a special care unit (SCU); SCU-F indicates residents living in a facility with an SCU but not in the SCU itself; SCU-N indicates residents in a facility with no SCU.

TABLE 2—Percentage of Residents Restraineda and Receiving Psychotropic Medication,b by Care Setting Type of Setting, % Treatment Received

SCU

Traditional Unit in Facility With SCU

Facility With No SCU

Neither restrained nor receiving psychotropic medication Physical restraints only Psychotropic medication only Restrained and receiving psychotropic medication Total residents physically restrained Total residents receiving psychotropic medication

41

56

50

7 39 13

9 29 5

11 30 8

20

14

19

52

34

38

Note. SCU = special care unit. a Does not include bed rails. b Antipsychotics and sedative-hypnotics.

marked difference between the study facilities and those of the nation as a whole was in facility size; in general, study facilities were smaller, and the same holds true for those facilities with SCUs.31 Table 1 indicates differences between the 3 subpopulations of residents: SCU, SCU-N, and SCU-F. Residents not in an SCU were most likely to be 85 years or older. SCU residents were more likely to be between 75 and 84 years than they were to be 85 or older. SCU residents were much more likely to be very cognitively impaired

and to exhibit problem behaviors. Members of the 3 groups were equally likely to be women and did not differ dramatically in their length of stay in the facility. Table 2 presents the distribution of residents across the categories of the dependent variable and the primary independent variable. These bivariate results suggest that significant differences may have existed in the treatment of residents in the 3 settings. SCU residents were slightly less likely than other residents to be restrained, much more likely to receive psychotropic medication, and more January 2000, Vol. 90, No. 1

Alzheimer Special Care Units

TABLE 3—Relationship Between Setting and Use of Restraint and Medication: Adjusted Odds Ratios and 95% Confidence Intervalsa Adjusted Odds Ratios (95% Confidence Intervals) for Categories of Dependent Variable Resident Settings and Groups Full population Residence in an SCU Residence in a facility with an SCU All SCU residents and a multivariate matched comparison group Residence in an SCU Residence in a facility with an SCU Mildly or moderately cognitively impaired residents only Residence in an SCU Residence in a facility with an SCU Severely cognitively impaired residents only Residence in an SCU Residence in a facility with an SCU All residents in non-SCU facilities and in higher quality SCU facilities Residence in an SCU Residence in a facility with an SCU

Physical Restraints Only

Psychotropic Medications Only

Both Restraints and Medications

0.75 (0.51, 1.10) 0.83 (0.66, 1.04)

1.55*** (1.26, 1.90) 0.95 (0.83, 1.10)

1.17 (0.79, 1.75) 0.64** (0.49, 0.85)

0.67* (0.47, 0.97) 0.90 (0.62, 1.31)

1.48*** (1.19, 1.82) 0.76* (0.59, 1.00)

0.93 (0.62, 1.39) 0.50*** (0.33, 0.75)

0.81 (0.43, 1.50) 0.67** (0.50, 0.90)

1.38* (1.04, 1.82) 0.95 (0.80, 1.14)

1.34 (0.71, 2.53) 0.63** (0.45, 0.89)

0.67* (0.48, 0.95) 0.96 (0.74, 1.25)

1.50** (1.12, 2.00) 0.88 (0.68, 1.13)

0.95 (0.64, 1.42) 0.66** (0.48, 0.90)

0.57 (0.27, 1.17) 0.83 (0.66, 1.04)

1.53** (1.11, 2.12) 0.95 (0.83, 1.09)

1.16 (0.60, 2.23) 0.65** (0.50, 0.85)

Note. SCU = special care unit. a Results derived from multinomial logistic regression including covariates (i.e., age, sex, payor, length of stay, activities of daily living [ADL] function, cognitive function, history of falls, body control problems, behavior problems, facility ownership, facility location, and facility size). Table entries are parameters for different care settings only. Parameters for covariates are not reported to simplify the presentation. Full tables are available from the corresponding author. *P < .05; **P < .01; ***P < .001.

likely to be both restrained and receiving psychotropic medication. However, considering the differences displayed in Table 1, only multivariate results can indicate whether these differences were attributable to differences in the 3 populations or to different care processes in the 3 settings. The results of the multivariate analysis appear in Table 3, which displays the relative odds ratios and their 95% confidence intervals for the 2 categories of the “treatment” variable in each of the 3 equations estimated in the multinomial logistic regression. The reference category for the treatment variable was composed of the SCU-N group. The reference group for the 4-category dependent variable included residents receiving neither psychotropics nor physical restraints. Parameter estimates for the covariates in each model are not displayed for the sake of presentational clarity and simplicity. Table 3 presents 5 sets of results, each for different configurations of the study population. The first set of results was obtained when the model was estimated with the entire study population. They indicate that SCU residents were restrained at rates that were not significantly different from the rates observed in the other 2 settings. However, SCU residents were significantly more likely to receive psychotropic medication than were any other residents. These results also indicate that SCU-F residents were significantly less likely to receive both physical restraints January 2000, Vol. 90, No. 1

and psychotropic medication than were SCU-N residents. In summary, results for the entire population imply that SCUs did not differ from other settings in their use of physical restraints and were more likely to give residents psychotropic medication. The second set of results presents estimates obtained when the analysis was performed only with SCU residents and residents determined to be very similar to SCU residents on the basis of our multivariate matching strategy. These results also indicate that SCU residents were more likely to receive psychotropic medication and that SCU-F residents were less likely both to be physically restrained and to receive psychotropic medication. The lower likelihood of physical restraint among SCU residents than among SCU-N residents does not reach the level of statistical significance (i.e., P < .01) set as our standard in this study. In addition, restraint rates for SCU residents were not significantly different from restraint rates for SCU-F residents (P = .20) . The third and fourth sets of results were obtained when models were estimated separately for individuals with different levels of cognitive impairment. In neither of these models does one see evidence that residence in an SCU reduced the likelihood of being physically restrained; however, the models do indicate that SCU residents with severe cognitive impairment were more likely to receive psychotropic medication.

The final set of results in Table 3 was obtained with a model that used all the residents in non-SCU facilities but only those residents in facilities with SCUs that scored higher on our SCU quality measure. These results were consistent with those obtained in earlier analyses: SCU residents were significantly more likely to receive psychotropic medication. None of the 3 initial hypotheses were consistently sustained. SCU residents were not physically restrained at rates significantly lower than the restraint rates for similar residents in traditional units. SCU residents were more likely, rather than less likely, to receive psychotropic medication, and they were treated with both physical restraints and psychotropic medication at rates no different from the rates observed in facilities without SCUs.

Discussion Each year, more facilities offer “special dementia care.” The fundamental question about these special units is whether they do, in fact, offer anything special. Previous research has indicated that residents in these units receive no more direct care than similar residents in other units in the same facility.11 Other research has indicated that receiving care in an SCU seems to have no differential effect on rates of decline in residents’ function.31 One of the areas in which previous American Journal of Public Health

95

Phillips et al.

research has found significant differences is in the use of physical restraints and psychotropic medication. Previous research has consistently found lower use of restraints in SCUs, but results have been mixed on the use of psychotropic medication in SCUs.10,11 Our results imply that the industry as a whole may have caught up with SCUs in physical restraint policy. The most plausible reason that we once saw, but no longer see, any difference in the use of restraints in SCUs and non-SCUs is that, owing to OBRA-87, the non-SCUs have reduced the use of restraints to the lower levels seen earlier in SCUs. In essence, changes in the industry seem to have made one “special” aspect of SCU care more commonplace. The results for psychotropic medication are somewhat more troubling, with higher rates of drug use in the SCUs. These results held even when a variety of strategies were used to ensure the comparability of residents regarding ADLs, cognition, and behaviors. It may be that the concentration of residents with cognitive and behavioral problems in SCUs pushes staff over some threshold in their receptivity to psychotropic drug use. Staff may be willing to try a range of responses to problem behaviors when only a few residents present with such problems, but when some larger proportion of residents on a unit present with behavior problems, staff may be more likely to resort to a “standardized” response involving medication. Such standardization is, of course, exactly what advocates of SCUs hoped would not occur. These results, obviously, reflect care patterns in the 4 study states during a specific time period. Thus, they do not represent a complete picture of how all SCUs operate. Also, to the degree that we have failed in our efforts to adjust for resident and facility differences across our 3 settings, the results are suspect. However, when taken in conjunction with earlier research, these findings lead to a very mixed picture of SCUs’ performance. There are documented differences in quality among SCUs,32 but the available evidence, to which this research contributes, demonstrates no clear superiority of SCUs over traditional units in either their outcomes or their processes of care.

Contributors C. D. Phillips, C. Hawes, and P. D. Sloane were involved in the conception and design of the study, as well as the interpretation of the results. K. M. Spry was involved in the analysis and interpretation of the results. C. D. Phillips was responsible for the initial draft of the paper. K. M. Spry, P. D. Sloane, and C. Hawes provided comments, corrections, and revised text for inclusion in the final draft.

96


Acknowledgments This research was supported by grant U01-AG10313 from the National Institute on Aging. The data used in this research were collected as part of the Health Care Financing Administration’s Nursing Home Case-Mix and Quality Demonstration.

References 1. Office of Technology Assessment. Special Care Units for People With Alzheimer’s and Other Dementias: Consumer Education, Research, Regulatory, and Reimbursement Issues. Washington, DC: Office of Technology Assessment; 1992. Publication OTA-H-543. 2. Freiman M, Brown E. Special Care Units in Nursing Homes—Selected Characteristics— 1996. Rockville, Md: Agency for Health Care Policy and Research; 1999. MEPS Research Findings No. 6. 3. Ohta RJ, Ohta BM. Special units for Alzheimer’s disease patients: a critical look. Gerontologist. 1988;28:803–808. 4. Sloane PD, Lindeman DA, Phillips CD, Moritz DJ, Koch G. Evaluating Alzheimer’s special care units: reviewing the evidence and identifying potential sources of study bias. Gerontologist. 1995;35:103–111. 5. Evans LK, Strumpf N. Tying down the elderly: a review of the literature on physical restraints. J Am Geriatr Soc.1989;37:65–74. 6. Tinetti ME, Wen-Liang L, Ginter SF. Mechanical restraint use and fall-related injuries among residents of skilled nursing facilities. Ann Intern Med. 1992;116:369–374. 7. Werner P, Cohen-Mansfield J, Brown J, et al. Physical restraint and agitation in nursing home residents. J Am Geriatr Soc. 1989;37:1122–1126. 8. Miles SH, Irvine P. Deaths caused by physical restraints. Gerontologist. 1992;32:762–766. 9. Kane RL, Ouslander JG, Abrass IB. Essentials of Clinical Gerontology. 3rd ed. New York, NY: McGraw-Hill; 1994. 10. Sloane PD, Mathew LJ, Scarborough M, et al. Physical and pharmacological restraint of nursing home residents with dementia. JAMA. 1991;265:1278–1282. 11. Mehr DR, Fries BE. Resource use on Alzheimer’s special care units. Gerontologist. 1995;35: 179–184. 12. Castle NG, Mor V. Physical restraints in nursing homes: a review of the literature since the Nursing Home Reform Act of 1987. Med Care Res Rev. 1998;55:139–170. 13. Rovner B, Edelman BA, Cox MP, et al. The impact of antipsychotic drug regulations on psychotropic prescribing practices in nursing homes. Am J Psychiatry. 1992;149:1390–1392. 14. Shorr RI, Fought RL, Way WA. Changes in antipsychotic drug use in nursing homes during implementation of the OBRA-87 regulations. JAMA. 1994;271:358–362. 15. Semla TP, Palla K, Poddig B, Brauner DJ. Effect of the Omnibus Reconciliation Act 1987 on antipsychotic prescribing in nursing home residents. J Am Geriatr Soc. 1994;42:648–652. 16. Lantz MS, Giambanco V, Buchalter EN. A tenyear review of the effect of OBRA-87 on psychotropic prescribing practices in an academic nursing home. Psychiatr Serv. 1996;47:951–955. 17. Hawes C, Mor V, Phillips CD, et al. The OBRA87 nursing home regulations and implemen-

18.

19.

20.

21. 22.

23.

24.

25.

26.

27.

28.

29.

30.

31.

32.

33.

34.

tation of the resident assessment instrument: effects on process quality. J Am Geriatr Soc. 1997;45:977–985. Borson S, Doane K. The impact of OBRA-87 on psychotropic drug prescribing in skilled nursing facilities. Psychiatr Serv. 1997;48:1289–1296. Snowden M, Roy-Byrne P. Mental illness and nursing home reform: OBRA-87 ten years later. Psychiatr Serv. 1998;49:229–233. Hawes C, Morris JN, Phillips CD, Fries BE, Murphy K, Mor V. Development of the nursing home Resident Assessment Instrument in the USA. Age Ageing. 1997;26(suppl 2):19–25. Zinn JS, Mor V. Nursing home special care units. Gerontologist. 1994;34:371–377. Banaszak-Holl J, Zinn JS, Mor V. The impact of market and organizational characteristics on nursing facility service innovation: a resource dependency perspective. Health Serv Res. 1996;31:97–117. Hawes C, Morris JN, Phillips CD, Mor V, Fries BE, Nonemaker S. Reliability estimates for the Minimum Data Set for Nursing Facility Resident Assessment and Care Screening (MDS). Gerontologist. 1995;35:172–178. Frederiksen K, Tariot P, De Jonghe E. Minimum Data Set Plus (MDS+) scores compared with scores from five rating scales. J Am Geriatr Soc. 1996;44:305–309. Phillips CD, Morris JN. 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. 1997; 11(suppl 6):162–167. Phillips CD, Hawes C, Mor V, Fries BE, Morris JN, Nennstiel ME. Facility and area variation affecting the use of physical restraints in nursing homes. Med Care. 1996;34:1149–1162. Harrington C, Tompkins C, Curtis M, Grant L. Psychotropic drug use in long-term care facilities: a review of the literature. Gerontologist. 1992;32:822–833. Morris JN, Fries BE, Mehr D, et al. MDS Cognitive Performance Scale. J Gerontol. 1994;49: M174–M182. Hartmaier SL, Sloane PD, Guess HA, Koch GG, Mitchell CM, Phillips CD. Validation of the Minimum Data Set Cognitive Performance Scale: agreement with the Mini-Mental State Examination. J Gerontol A Biol Sci Med Sci. 1995;50:M128–M133. Rosenbaum PR, Rubin DB. The central role of the propensity score in observational research for causal effects. Biometrika. 1983;70:41–55. Phillips CD, Sloane PD, Hawes C, et al. Effects of residence in Alzheimer disease special care units on functional outcomes. JAMA. 1997;278: 1340–1344. Sloane PD, Mitchell CM, Preisser JS, Phillips CD, Commander C, Burker E. Environmental correlates of resident agitation in Alzheimer’s disease special care units. J Am Geriatr Soc. 1998;46:862–869. Shah BV, Barnwell BG, Bieler GS. SUDAAN User’s Manual: Release 6.4. 2nd ed. Research Triangle Park, NC: Research Triangle Institute; 1996. Krauss NA, Altman BM. Characteristics of Nursing Home Residents—1996. Rockville, Md: Agency for Health Care Policy and Research; 1998. MEPS Research Findings No. 5.

January 2000, Vol. 90, No. 1