Cognitive Status and Relocation Stress: A Test of the Vulnerability Hypothesis

Abstract

Purpose: This study investigated whether cognitively impaired nursing home residents are at particular risk of experiencing harmful effects during a mass, intra-institutional, interbuilding relocation. Design and Methods: A pretest–post-test experimental-comparison group design was used. Data on cognitive status, functional capacity, psychosocial health status, physical health status, and mortality were abstracted from the Minimum Data Set Plus and were analyzed using continuous and discrete survival analyses, controlling for covariates as well as baseline status of outcome variables. Results: None of the Relocation × Cognitive Status interaction effects were significant. Relocation main effects indicated that movers in general were more likely than nonmovers to decline in physical health status. Evidence also emerged for a positive long-term effect of moving on psychosocial health status. Implications: These findings suggest cognitively impaired residents are not at unusual risk of harmful effects as a consequence of mass, interbuilding transfer. Given the significant relocation main effects, though, caution must be taken in moving cognitively impaired residents, as it should be in moving any residents.

Decision Editor: Laurence G. Branch, PhD

Since the mid-1980s there has been an enormous proliferation of dedicated special care units (SCUs) for nursing home residents with dementia (Ohta and Ohta 1988). As a consequence, an increase in the occurrence of intra-institutional transfers among residents of long-term care facilities has undoubtedly occurred. The development of an SCU in a facility often entails transferring residents out of an area to construct the unit (Reingold and Werner 1994). Previously admitted residents who meet the admission criteria of a newly opened SCU may be relocated into the unit. Residents admitted to non-SCUs who exhibit cognitive decline may be transferred into an already existing SCU (Weiner and Reingold 1989). In addition, because about half of SCUs admit residents with the expectation that they will be discharged some time prior to their death (Holmes, Teresi, and Monaco 1992), residents also experience transfers out of SCUs (Krovach 1998).

In particular, two kinds of intra-institutional relocations have been noted to occur as a result of the establishment of an SCU in a facility: "Intrabuilding transfers," that is, moving residents from one room to another within the same building, and "Interbuilding transfers," that is, moving residents out of one building into another building within the same nursing home complex (Reingold and Werner 1994).

Curiously, during this period as intra-institutional transfers have increased in frequency, the number of articles published on relocation effects has noticeably declined (Mirotznik 1995). Most research on this phenomenon dates from prior to 1985. Important questions remain, however. One such question concerns whether cognitively impaired nursing home residents are at particular risk of negative relocation effects. This question takes on added significance given current estimates that 48% of nursing home residents have a diagnosed dementia (Leon and Moyer 1999) and given the increasing incidence of intra-institutional relocation of such residents as a consequence of the establishment of SCUs (Krovach 1998).

The present study investigated a mass, intra-institutional, interbuilding transfer of long-term care nursing home residents. Its aim was to determine whether cognitive status moderated relocation effects.

Review of the Literature

It is a truism in the relocation literature that cognitively impaired nursing home residents are at particular risk of experiencing harmful relocation effects (Blenkner 1967; Goplerud 1979; Kowalski 1981; Pastorello 1975; Yawney and Slover 1973). However, a critical review of the literature indicates a lack of adequate evidence to support this belief.

Typically, the moderating influence of cognitive status has been determined by contrasting subsets of movers in one of two ways: (a) by comparing relocated residents who are cognitively impaired with those who are not impaired for differences on an outcome variable (e.g., mortality and/or morbidity; Friedman et al. 1995; Kral, Grad, and Berenson 1968; Lander, Brazill, and Ladrigan 1997; Marlowe 1974; Nirenberg 1983) or (b) by comparing relocated residents exhibiting deterioration on an outcome variable (e.g., those who died) and those exhibiting no change (e.g., those who survived) for a difference in terms of cognitive status (Aldrich 1964; Guttman and Herbert 1976; Lieberman and Tobin 1983, pp. 150–153; Markus, Blenkner, Bloom, and Downs 1972; Miller and Lieberman 1965; Ogren and Linn 1971). In general, studies that have compared subsets of movers in these two ways have found evidence of a positive association between cognitive impairment and posttransfer deterioration. Specifically, five studies found greater mortality among the cognitively impaired (Aldrich 1964; Kral et al. 1968; Lieberman and Tobin 1983, p. 152; Marlowe 1974; Markus et al. 1972), whereas two studies did not (Guttman and Herbert 1976; Ogren and Linn 1971). And five studies documented greater morbidity among the cognitively impaired (Friedman et al. 1995; Kral et al. 1968; Lander et al. 1997; Miller and Lieberman 1965; Nirenberg 1983), whereas three did not (Lieberman and Tobin 1983, pp. 151–152; Marlowe 1974).

Interpreting these subgroup differences as proof of the greater susceptibility of the cognitively impaired to relocation stress is, however, problematic. The greater mortality and/or morbidity among relocated residents who are cognitively impaired may merely reflect the fact that such residents deteriorate at a higher rate in general, regardless of the occurrence of relocation (Coffman 1981, Coffman 1983; Csank and Zweig 1980; Goldfarb, Fisch, and Gerber 1966; Tobin and Lieberman 1976; van Dijk, Dippel, and Habbema 1991). In the terminology of analysis of variance (ANOVA), differences among subgroups of relocated residents may signify a main effect of cognitive status as opposed to an interaction between cognitive status and relocation stress. The only way of disentangling these possibilities is by contrasting cognitively impaired and unimpaired movers to comparable groups of nonmovers. Should movers who are cognitively impaired significantly deteriorate in comparison to their controls, while movers who are cognitively unimpaired either not differ in outcome or improve in comparison to their controls, one would be able to conclude that cognitive impairment heightened vulnerability to relocation stress.

Four studies in the literature investigated the impact of cognitive status on relocation outcome by contrasting a group of movers with an external control group of nonmovers. Of these, two failed to find evidence in support of the greater vulnerability of the cognitively impaired (Goldfarb, Shahinian, and Burr 1972; Lieberman and Tobin 1983, p. 149). Of the remaining two studies, one found moderately impaired patients (Pruchno and Resch 1988) and the other severely impaired patients (Csank and Zweig 1980) to be at greatest risk. A fifth study (Mirotznik 1995), which used relocated residents as they existed at any earlier period as their own controls, found that moderately cognitively impaired and unimpaired residents as a group experienced more effects, including a positive prerelocation and negative postrelocation response, than did residents who were severely impaired.

In summary, of the 17 separate studies investigating the cognitive status vulnerability hypothesis, 12 lacked a comparison group of nonmovers and the remaining 5 have yielded contradictory findings. As such, the evidence does not allow one to conclude with confidence that cognitively impaired residents are at particular risk of harmful relocation effects. Further, given that only two studies (Mirotznik 1995; Pruchno and Resch 1988) investigated the moderating effect of cognitive status in an intra-institutional relocation, the question of how the cognitively impaired react in this type of transfer warrants further research.

Theoretical Models

Three broad types of theoretical models have been suggested to explain how cognitive status influences relocation outcome. These explanations may be labeled stress-reaction models, coping models, and stressor-coping models. Stress-reaction models assume that the cognitively impaired and unimpaired experience relocation differently, that it involves a differential degree of stress for these two groups. Lawton and Simon 1968 environmental docility hypothesis, for example, suggests that the cognitively impaired are more dependent on their external environment and that, consequently, relocation involves more of a disruption for them (Pruchno and Resch 1988). (For another example, see Grad, in Csank & Zweig, 1980.)

Coping models emphasize the lessened abilities and resources of the impaired for dealing with relocation. Schulz and Brenner 1977 hypothesized that perceived control and predictability during relocation influence outcome. This model suggests that the cognitively impaired may not be able to buffer themselves from the stress of relocation by developing a sense of personal control or by rendering the change more predictable by availing themselves of preparation programs (Pruchno and Resch 1988).

Stressor-coping models elaborate upon the particular aspects of the relocation stressor that induce stress as well as incorporate the notion of the lack of availability of coping resources. Csank and Zweig 1980 suggested that the cognitively impaired are incapable of experiencing anxiety and deleterious effects in anticipation of relocation. Rather, stress results for the impaired from the actual relocation itself, specifically, from loss of a familiar environment that triggers an anxiety reaction. It is this anxiety reaction coupled with a lack of coping resources that results in heightened vulnerability to negative relocation effects. (For other examples, see Aldrich 1964; Markus et al. 1972.)

Two stressor-coping models distinguish the moderately from the severely impaired, but make different predictions. Lieberman and Tobin 1983 pathognomonic sign model suggests that the severely impaired may be so limited in their capacity to cope that they are likely to experience deleterious effects regardless of the degree of stress inherent in the particular relocation. The effect of relocation on those with adequate coping resources (i.e., the moderately impaired and the unimpaired) would largely depend on the exact nature of the relocation stressor. Pruchno and Resch 1988 modified environmental docility hypothesis suggests the highly impaired are unlikely to be affected by relocation because they lack the threshold capacity needed for dependency on environmental cues. The moderately impaired, in contrast, are very dependent on such cues and, as such, experience relocation as highly stressful. The unimpaired are also unlikely to be affected because they have the coping reserves for adjusting to change.

Hypotheses

Several authors (Borup 1981; Kasl 1972; Pastorello 1975; Tobin and Lieberman 1976; Yawney and Slover 1973) have described relocation as a process consisting of distinct stages: (a) a decision and preparation stage prior to relocation, also known as an anticipatory stage, (b) an impact stage within which the actual physical transfer occurs, and (c) a settling-in or long-term adjustment stage. Each of these stages may have its own dynamics and its own potential for stress (Tobin and Lieberman 1976). During the first stage, stress may result from anticipation of the imminent relocation. There is evidence to suggest that anticipation of life events, including relocation, can be quite stressful (Kasl 1972; Tobin and Lieberman 1976; Zweig and Csank 1976). During the second stage, stress may result from the physical move itself and actual loss of a familiar environment, whereas during the third stage it may result from the difficulty of adjusting to a new environment.

The theoretical models described above to explain the greater vulnerability of the cognitively impaired focus on events that occur during Stages 2 and 3 of the relocation process. They suggest that the cognitively impaired are likely to exhibit negative effects as a result of an actual change in their environment. Indeed, one of the models explicitly argues against the possibility of the cognitively impaired experiencing stress in anticipation of relocation (Csank and Zweig 1980). On the basis of these models, therefore, one would predict that the cognitively impaired, in comparison with the unimpaired, would be more likely to exhibit deleterious effects postrelocation as opposed to prerelocation.

Further, on the basis of the various models, it is possible to derive competing predictions about the pattern of these negative postrelocation effects. Several of the models suggest a positive linear relationship between cognitive impairment and negative effects. For instance, the environmental docility hypothesis (Lawton and Simon 1968) implies that the more impaired the individual, the more dependent he or she is on his or her external environment and, consequently, the greater the deleterious effects. Such a linear pattern also appears to be implied in the models of Aldrich 1964, Csank and Zweig 1980, Grad (Csank and Zweig 1980), Markus and colleauges (1972), and Schultz and Brenner (1977). The modified docility hypothesis, argued by Pruchno and Resch 1988, suggests a curvilinear pattern, with those highest and lowest in impairment exhibiting the fewest effects and those with moderate impairment exhibiting the greatest effects. Lastly, the pathognomonic sign model of Lieberman and Tobin 1983 suggests a nonlinear pattern, whereby individuals with high impairment would necessarily experience harmful effects, whereas those below that threshold level (i.e., the moderately impaired and the unimpaired) would not. Hence, the following specific questions were addressed in this study:

1. Are cognitively impaired residents less likely than cognitively unimpaired residents to experience an anticipatory stress reaction, that is, negative relocation effects manifested in preparation for or in anticipation of the actual physical transfer?

2. Are cognitively impaired residents more likely than cognitively unimpaired residents to exhibit negative relocation effects during the impact and/or settling-in stages of the relocation process?

3. Do severely or moderately cognitively impaired residents exhibit more harmful postrelocation effects? In other words, do posttransfer effects manifest a linear, curvilinear, or a nonlinear pattern with regard to cognitive status?

Methods

Setting

This study was conducted at two nursing homes that were participating agencies of Metropolitan Jewish Health System (MJHS): Shorefront Jewish Geriatric Center and MJG Nursing Home, both located in Brooklyn, New York. On July 1, 1994, all of the residents of Shorefront experienced a mass, intra-institutional, interbuilding relocation, moving into a new replacement facility constructed next to their former building. MJG Nursing Home, Shorefront's sister facility, did not experience an equivalent move.

Residents were informed of the upcoming move in January 1994. Also, at that time formal preparation procedures began. Therapeutic recreation staff, social workers, and nursing staff provided residents with information and emotional support, helped residents identify roommate preferences, helped select belongings to take to the new facility, and, when needed, introduced the residents to new staff and/or other residents. Families of residents were also notified about the upcoming relocation and asked to provide support to the residents. During the postmove period, staff continued to provide both emotional support and information and were encouraged to confer frequently with one another regarding residents' coping and general health status.

Two points are noteworthy about this relocation. First, given the preparation and support provided to residents, this move needs to be seen as a best case scenario. Should deleterious effects occur in this instance, it would suggest that in interbuilding moves without such preparation and support, there may be a greater likelihood of harmful effects. A second point concerns the fact that residents who underwent the interbuilding transfer were aware at least 6 months prior to relocation that they were going to be moved. As such, it is possible that during the months prior to the move, some of the residents experienced changes in health status in anticipation of relocation (Mirotznik 1995).

Study Design

The study used a variation of the pretest–posttest experimental-comparison group design. The experimental group consisted of 405 residents at the Shorefront facility admitted by July 1993 who underwent the interbuilding transfer. The comparison group consisted of 383 MJG Nursing Home residents admitted to that facility by July 1993 who did not undergo such a transfer. Although each nursing home had a larger patient census, a number of residents from the two facilities were excluded from the study. For example, given the focus on long-term nursing home residents, at Shorefront 47 and at MJG Nursing Home 83 subacute residents (i.e., those admitted for short-term rehabilitation and discharged back to the community relatively quickly) were excluded. In addition, to increase the comparability of the experimental and comparison groups, all Hospice and Maximum Care Unit patients at MJG Nursing Home were excluded.

Data were collected on all participants for a 24-month time frame, from July 1993 through June 1995, representing four periods. The months from July 1993 through December 1993 constituted the baseline period; January through June 1994 constituted the prerelocation period; July through December 1994 constituted the short-term postrelocation period; and January 1995 through June 1995 constituted the long-term postrelocation period.

Data

Data for this study were derived from the Minimum Data Set Plus (MDS+), a federally mandated, standardized clinical assessment form for monitoring the functional and medical status of nursing home residents (Morris et al. 1990). There has been some discussion in the literature about the utility of nursing home resident assessment instruments for the purpose of research. In particular, concerns have been voiced about the reliability and validity of the MDS+ (Hawes, Phillips, Mor, Fries, and Morris 1992; Teresi and Holmes 1992). In recent years, a number of studies have investigated the psychometric properties of the instrument, indicating that the MDS+ appears to measure certain dimensions better than others. Hawes and colleagues 1995 found that the MDS items for areas such as cognition, activities of daily living (ADLs), continence, disease diagnoses, and administered medication exhibited excellent interrater reliability. Psychosocial well-being exhibited good reliability, whereas, in comparison, the MDS items for mood exhibited poorer reliability. Several groups of investigators have documented that the MDS cognitive status items correspond closely with scores on the Mini-Mental State Examination and also with other measures of cognitive status (Hartmaier et al. 1995; Morris et al. 1994; Phillips, Chu, Morris, and Hawes 1993). Mor and colleagues 1995, conducting a confirmatory factor analysis, documented that MDS items representing the constructs of social engagement, mood problems, conflicted relationships, and behavior problems factored as predicted. In contrast, Frederiksen, Tariot and De Jonghe 1996 found that although the MDS measures of ADLs, cognitive impairment, and communication correlated highly with comparable rating scales scores, the MDS measures of problem behavior and mood did not. Similarly, using confirmatory factor analysis, Casten, Lawton, Parmelee and Kleban 1998 found evidence for the hypothesized factor structure of the MDS measures of cognition, ADLs, and time use (MDS section Sense of Initiative/Involvement), but not for social quality (MDS section Unsettled Relationships), depression, and problem behavior. The findings for the MDS continence items are mixed. Although Resnick, Brandeis, Baumann, and Morris 1996 as well as Hawes and associates 1995 found that these items had excellent interrater reliability, Crooks, Schnelle, Ouslander, and McNees 1995 documented a weak and nonsignifciant correlation between MDS continence ratings and physical checks for wetness performed by nursing home staff.

Evidence suggests that MDS items may be less reliable for the cognitively impaired (Casten et al. 1998), particularly those items based on staff observation and assessment (Phillips et al. 1993). MDS items based on medical records exhibit the highest interrater reliabilities that in practical terms differ little between the cognitively impaired and intact (Phillips et al. 1993). Evidence also indicates that the internal consistency reliability of the MDS items representing ADLs does not vary across levels of cognitive impairment (Phillips and Morris 1997).

Measures

A comprehensive MDS+ assessment is completed at admission, annually, and in response to a significant change in a resident's status. In addition, quarterly reviews are completed every 3 months on a large subset of the MDS+ items. To have the requisite data to measure health status in each period of the study, only those items included in the quarterly assessment were used. This meant that for each 6-month study period, each participant had available a minimum of two completed MDS+ assessments.

MDS+ items recorded quarterly were further screened in terms of their reported psychometric properties. Items suggested by the literature to have questionable reliability and/or validity, such as those pertaining to mood, problem behavior, unsettled relationship, and continence, were excluded. It is worth noting that these items also exhibited poor local internal consistency reliability (data not shown).

To operationalize cognitive status we used the MDS Cognitive Performance Scale (CPS; Morris et al. 1994). The CPS, which is based on five MDS items (comatose status, ability to make decisions, short-term memory, making self understood, and self-performance in eating), classifies residents into seven categories: intact (0), borderline intact (1), mild impairment (2), moderate impairment (3), moderately severe impairment (4), severe impairment (5), and very severe impairment (6). In one validation study, the CPS explained 74% and 75% of the variance, respectively, in MMSE scores and scores derived from a combination of the MMSE and the Test for Severe Impairment (TSI; Morris et al. 1994). In a second validation study, the CPS, also when tested against the MMSE as a criterion variable, showed 94% sensitivity, 94% specificity, and 96% diagnostic accuracy as measured by the area under the receiver operating characteristics curve (Hartmaier et al. 1995). For the purposes of our study, the seven CPS categories were collapsed into three broader groupings: intact (0–1), moderate impairment (2–3), and severe impairment (4–6).

To determine if the CPS measure was behaving as it should within our sample (i.e., exhibited local validity) we subjected it to a number of tests. Following Morris and colleagues 1994 lead, we crosstabulated CPS scores with the clinical diagnoses of dementia. Although neurological diseases are known to be underdiagnosed in nursing homes (Barnes and Raskind 1981), we hypothesized that the prevalence of diagnosed cases of dementia should still increase across the three CPS categories. As seen in Table 1 , this pattern did indeed emerge. Second, following Phillips and Morris 1997 we assessed the association of CPS and ADL scores (see below for ADL measure). The overall correlation was .50. Table 1 indicates that ADL dependence increased linearly across the three CPS categories. And finally, a considerable body of research indicates that cognitively impaired residents are at risk of deteriorating over time (van Dijk et al. 1991). Table 1 shows that the more impaired a resident was, the greater the chances of dying over the 2-year period of 1993–1995.

ADLs consisted of five items, concerning bed mobility, transfer ability, locomotion, dressing, and toilet use, each of which was scored from 0 (independence) to 4 (total dependence). Responses to the items were summed and divided by the number of answered items. The Cronbach's alpha for the ADL measure was .94.

Psychological well-being, derived from the Sense of Initiative/Involvement section of the MDS+, consisted of seven items (e.g., "at ease interacting with others," "at ease doing planned or structured activities"), each coded in our study so that 0 represented yes and 1 represented no. All seven items were summed for a composite score that ranged from 0 to 7. The Cronbach's alpha for this scale was .73.

Physical health status was operationalized on the basis of several MDS+ items abstracted from patients' medical records. Specifically, measures were constructed of the number of (a) diagnosed diseases, (b) conditions and signs/symptoms, (c) administered medications, (d) emergency room transfers, and (e) hospital admissions. As suggested above, we also had data on whether the resident expired (mortality) during the study period.

Data were also abstracted on the following covariates: age, gender, length of stay, Medicare or Medicaid payment source for nursing home stay, primary language (English or other), marital status (ever or never married), and race (White or non-White).

Analytic Approach

For mortality we used continuous survival analysis for categorical time-to-event data. For ADLs, psychological well-being, and the five physical health status measures we used discrete survival analysis (Allison 1982). Our analytic approach made use of all of the assessment data for each individual (i.e., up to 14 assessments during the study period). And it allowed the modeling of the effects of relocation status, cognitive status, and the interaction of Relocation Status × Cognitive Status, while controlling for baseline values of outcome variables as well as of extraneous covariates.

The time periods were defined for mortality as number of days from the start of the prerelocation period. For ADLs, psychological well-being, and the five physical health status measures, the time periods were the number of months during the study period. For mortality, failure was defined as the occurrence of death, and for the other outcome measures it was defined as a decline in a resident's health status from baseline (i.e., the MDS+ assessment immediately preceding the prerelocation period). Following Phillips and colleagues 1997, who used MDS data to document the effects of residence in an Alzheimer's disease special care unit, the month of decline was identified as the month at the midpoint between the assessment noting the decline and the previous assessment. Also similar to Phillips and colleagues, decline was operationalized as a 1 unit increase in a scale or count score. So a resident whose ADL score or number of administered medications changed since baseline from 3 to 4 exhibited such a decline. Residents whose baseline ADLs and psychological well-being scores indicated they could not decline further were excluded from the analysis of these two outcome variables. Residents who did not die or decline from baseline or who exited the two facilities prior to the end of the study period were considered censored at the time of their last assessment. Exit rate was found not to be associated with relocation status, cognitive status, or the interaction of these two variables.

A Cox regression was fit for mortality. In Step 1, relocation status (movers vs. nonmovers), cognitive status (represented by two dummy variables for moderate and severe impairment), and the two Relocation Status × Cognitive Status interaction terms were entered. To address the issue of resident selection bias, in Step 2 the covariates in terms of which the two nursing homes differed at baseline were entered. The proportional hazard assumption was evaluated by entering in Step 3 interaction terms representing the product of each independent variable with survival time. Significant Covariate × Survival Time interactions were to be retained in the final version of the equation, thereby adjusting for violations of the proportional hazard assumption (Kleinbaum 1996; Singer and Willett 1991). Significant interactions between theoretically central independent variables (relocation status, cognitive status, and their interactions) and survival time were to be further explored by segmenting time into three periods (prerelocation, short-term postrelocation, and long-term postrelocation) and by testing the relative risk of death within each period separately (Kleinbaum 1996).

Discrete survival analysis was conducted using logistic regression (Singer and Willett 1991). A separate regression was run for ADLs, psychological well-being, and the five physical health status measures. These regressions followed the same steps as the analysis for mortality, with two exceptions. Prior to entering the theoretically central independent variables, each outcome variable was regressed on 18 dummy variables, representing each month of the study. In addition, these equations adjusted for all baseline values of the outcome variables as well as initial differences between movers and nonmovers.

Results

Participant Characteristics

Of the total sample of 788 residents, 75% were women, 96% were White, with a mean age of 83 years and a 3-year length of stay (LOS). Seventy-seven percent spoke English as their primary language. For most, the source of payment for their nursing home stay was a combination of Medicaid and self-pay/private insurance (92% and 95%, respectively). Only 10% had their stays covered by Medicare. Fifteen percent had never married, 18% were still married, 60% widowed, and 7% separated or divorced. About 46% needed extensive assistance or were totally dependent in terms of ADLs. The most frequent diagnoses were arteriosclerotic heart disease (45%), cataracts (39%), dementias other than Alzheimer's disease (29%), anemia (28%), arthritis (28%), hypertension (28%), Alzheimer's disease (24%), and cerebrovascular accident (22%).

Movers and nonmovers were similar in terms of age, gender, psychological well-being, number of emergency room transfers, and number of administered medications (Table 2 ). Movers were somewhat less likely to have been married and to have their stay paid by Medicaid. They were more likely to speak English as a primary language, to be non-White, and to have their stay paid by Medicare. Movers were more cognitively intact, had a shorter mean LOS, and were less dependent in terms of ADLs. They also, however, had a greater number of diagnosed diseases, conditions and signs/symptoms, and hospital admissions than nonmovers.

The hazard ratios for the theoretically central independent variables of relocation status, cognitive status, and their interactions appear in Table 3 . All ratios have been adjusted for differences between movers and nonmovers in baseline values of length of stay, marital status, race, Medicaid coverage, Medicare coverage, primary language, ADLs, diagnosed diseases, conditions/signs/symptoms, and hospital admissions. In addition, the hazard ratios for the seven morbidity outcome measures were adjusted for the baseline values of these measures. Further, as indicated in Table 3 , certain ratios were adjusted for significant Covariate × Survival Time interactions. For clarity, the hazard ratios for the covariates are not presented, but the full models are available from the first author.

As presented in Table 3 , none of the hazard ratios for the Relocation Status × Cognitive Status interactions were significant, indicating that relocation effect was not moderated by level of cognitive impairment. Significant main effects emerged for relocation status and cognitive status. Movers in comparison with nonmovers exhibited greater rates of decline as indicated by number of diagnosed diseases, number of conditions, signs, symptoms, and hospital admissions. Further, severely cognitively impaired in comparison with cognitively intact residents, regardless of relocation status, declined more quickly in ADLs. None of the Relocation × Survival Time and Cognitive Status × Survival Time interactions for these particular outcome variables were significant. This suggests that relocation and cognitive status effects occurred across the pre- and two postrelocation periods equally. An exception to this pattern was found for psychosocial well-being. Whereas movers and nonmovers exhibited similar rates of decline in this outcome variable during the prerelocation and short-term postrelocation periods, during the long-term postrelocation period nonmovers declined at a higher rate, as indicated in Table 3 .

Discussion

The present study tested whether cognitively impaired nursing home residents were more vulnerable to harmful relocation effects during a prerelocation period, and a short- and long-term postrelocation period. These effects were measured in terms of eight outcome variables that covered mortality and physical and psychological morbidity. In spite of the large number of outcome variables investigated across three relocation periods, no evidence was uncovered to support the cognitive impairment vulnerability hypothesis. This finding is consistent with two of the studies in the literature that compared movers with a control group of nonmovers (Goldfarb et al. 1972; Lieberman and Tobin 1983).

It was noted above that MDS items, particularly those involving subjective assessment, are less reliable for cognitively impaired residents. This could have attenuated the Relocation Status × Cognitive Status interaction effects, especially for ADLs and psychological well-being. Yet, the outcome variables operationalized on more objective medical record data (number of administered medications, hospital admissions, etc.) also failed to exhibit significant interaction effects.

Our research uncovered four significant relocation status main effects. Three of those effects indicated that movers in general, regardless of cognitive status, exhibited a decline in physical health across all three relocation periods. The remaining effect suggested that movers exhibited an improvement in psychological health during the long-term postrelocation period. This represents 10 times the number of significant relocation status main effects we would have expected by chance alone (4/8=.5; .5/.05=10). Some studies in the literature have documented anticipatory relocation effects (Tobin and Lieberman 1976; Zweig and Csank 1976); many others have documented postrelocation effects (for a review, see Grant, Skinkle and Lipps 1992). The present study suggests that relocation effects, in terms of physical health at least, may begin prerelocation and continue throughout the first year postrelocation. Of course, the dynamics underlying and responsible for these deleterious effects may differ during pre- and postrelocation phases. Research investigating those dynamics would seem to be warranted. A practical implication of our finding is that administrators and staff of nursing homes about to undergo relocation must pay attention to all residents, not just to those who are cognitively impaired. Further, this attention must begin prerelocation and should extend to at least 12 months postrelocation. It is unclear why movers exhibited an improved psychological health status during the last 6 months of the postrelocation year. Future research might benefit by including additional psychological measures and determining their behavior in contrast to physical health measures during the various stages of the relocation process.

What implications do our results have for the transfer of cognitively impaired residents into SCUs? Prior to addressing this question, two qualifications must be stated. First, given our design, caution should be used in generalizing our findings to the transfer of one individual at a time into SCUs. It is possible that the dynamics of mass versus single-person transfers are different enough to result in different outcomes for severely cognitively impaired residents, and/or for residents in general. Research determining the effects on the severely cognitively impaired versus the moderately impaired and the unimpaired of single-person transfers between buildings and between rooms within a building would greatly contribute to our knowledge. Second, cognitively impaired residents moved into SCUs may differ from those who remain on nonspecialized units. Although the literature is inconsistent, studies have found differences in cognitive functioning, ADLs, problem behaviors, and so forth (Holmes et al. 1990; Riter and Fries 1992). To the degree that cognitively impaired residents selected for SCUs constitute a unique subset, they may in fact respond to transfer in unique ways.

With these caveats in mind, our findings would suggest that cognitively impaired residents are not at unusual risk of experiencing harmful effects as a consequence of being transferred into SCUs. Although this certainly leads to a more optimistic assessment of the possible unintended consequences of SCUs than would be suggested by previous relocation studies, there remains good reason to be concerned. To say that the cognitively impaired are not at special risk, does not mean that they are at no risk. Recall that residents in general, including those who were moderately and severely cognitively impaired, exhibited deleterious relocation effects. The possibility would seem to continue to exist that transfers into SCUs may indeed cause harm and that the benefits derived from living in an SCU might be somewhat offset by the stresses associated with being transferred into the unit. All of this would indicate that caution should be taken in moving cognitively impaired residents into SCUs, as it should be in moving any residents. Our findings would also suggest the need to study more directly the impact of transferring cognitively impaired residents into SCUs.

This study was supported by Alzheimer's Association Grant RG3-96-003. We thank the administration of Metropolitan Jewish Health System for its cooperation and Martin Piccochi for data management.

References