| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
| ||||||||||||||||||||||||
Correspondence: Address correspondence to Louis D. Burgio, PhD, Center for Mental Health and Aging, Box 870315, University of Alabama, Tuscaloosa, AL 35487-0315. E-mail: lburgio{at}sw.ua.edu
 |
Abstract |
|---|
 TOP Abstract MethodsResultsDiscussionReferences |
|---|
Key Words: Agitation • Nursing home • Longitudinal methods • Direct observation
Researchers have made efforts to understand the extent to which resident characteristics are associated with agitation and how these symptoms change over time as a function of those characteristics (Cohen-Mansfield & Werner, 1998a; McCarty et al., 2000; Wagner, Teri, & Orr-Rainey, 1995). Known personal factors associated with agitation include but are not limited to degree of cognitive impairment, age, gender, and physical and psychoemotional functioning (Cohen-Mansfield, 2000; Gruber-Baldini, Boustani, Sloane, & Zimmerman, 2004; Snowden, Sato, & Roy-Byrne, 2003; Vance, DeLaine, Washington, & Kirby-Gatto, 2003). Unfortunately, empirical findings are often contradictory and do not allow conclusions to be drawn regarding the correlates of agitation. Older age may be associated with more (Haupt, Kurz, & Janner, 2000) or less (Levy et al., 1996) agitation. Women may exhibit more overall agitation, but gender has been associated with different components of agitation (Wagner et al., 1995). For example, data have indicated that men are more physically aggressive than women, but women exhibit more verbal agitation (Hall & O'Conner, 2004). Investigators have reported rather consistent cross-sectional findings with physical and cognitive functioning in that more physical and cognitive impairments are related to increased agitation (Hall & O'Conner, 2004; Haupt et al., 2000; Levy et al., 1996; Schultz et al., 2002).
In cross-sectional studies, cognitive impairment has received considerable attention in predicting agitation among individuals with dementia (Gruber-Baldini et al., 2004; Levy et al., 1996; Schultz et al., 2002; Vance, Burgio, et al., 2003). Beck and colleagues (1998), for example, found that cognitive impairment based on the Mini-Mental State Examination (MMSE) score was the most significant predictor of agitation and that more cognitive impairment was associated with a higher incidence of agitation. Similarly, Vance, Burgio, et al. (2003) reported that cognitive impairment was the most consistent predictor of agitation as measured by both staff report and direct observation. Evidence from these cross-sectional studies suggests that agitation increases as individuals' cognitive functioning worsens with dementia progression.
Relatively little is known about the natural course of agitation, particularly among nursing home residents with profound cognitive impairment. Most longitudinal studies have used proxy-reporting of agitation among community-dwelling older adults with dementia, leaving the trajectory of agitation among nursing home residents with profound cognitive impairment largely unknown (Cohen-Mansfield & Werner, 1998a, 1998b; McCarty et al., 2000).
Investigators have assessed agitation almost exclusively through caregiver and staff reports and chart reviews (Cohen-Mansfield, 1995; Gruber-Baldini et al., 2004), but only a limited number of studies have examined agitation using direct observation (McCann, Gilley, Hebert, Beckett, & Evans, 1997; Vance, DeLaine, et al., 2003). These studies, all cross-sectional in nature, have suggested that direct observation, though more tedious, provides more accurate and reproducible data on agitation than staff ratings.
The purpose of this project was to examine aspects of Algase and colleagues' (1996) model of need-driven dementia-compromised behavior. In this model, the authors hypothesized that background factors (e.g., age, gender, motor ability) combined with proximal factors (e.g., physical and social environment) determine the display of agitation. The model suggests that as the residents' cognitive and functional behaviors decline, residents become more vulnerable to "environmental stress" and more likely to display agitation because of their inability to meet their own needs.
We focused on the trajectory of agitation in relation to a subset of four important background factors (resident age, gender, activities of daily living [ADLs], and cognitive functioning) in nursing homes over a time span of up to 18 months. This study not only investigated resident characteristics that contribute to baseline behavior function (cross-sectional component), but also examined the pattern and rate of change in agitation and its relation to resident characteristics over time (longitudinal component). In addition, this study used information from both staff ratings and direct observation of agitation. Prior studies examining these resident background characteristics were either cross-sectional or, if longitudinal, focused on a short span of time. None of the longitudinal studies used direct observation to quantify agitation. In our analyses, we used longitudinal hierarchical linear modeling (HLM) to capture dynamic behavioral change in relation to resident characteristics.
|
Methods |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
Researchers recorded data during four assessment epochs. An epoch is defined as a sequential assessment of all entered residents across the four nursing homes. After an epoch was completed (i.e., approximately 6 months), the researchers sequentially assessed the same four nursing homes (Epoch 2). This was completed for four epochs over 2 years. In all, 123 residents met entry criteria and gave proxy consent. Because the focus of this article is longitudinal analysis, we selected 78 residents from the larger sample for whom we had data for at least three sequential observation epochs (i.e., each resident had 12 or 18 months of data). We followed 55 residents for four epochs and 23 residents for three epochs, thereby yielding an analysis sample of 78. We compared these two samples (55 and 23) on all personal and behavioral characteristics using t tests and a chi-square test. There were no significant baseline differences between the two groups on any personal characteristic, direct behavioral observation, or staff-reported behavioral disturbances.
Although most of the residents were recruited at Epoch 1, new residents were added every 6 months to replace residents who were discontinued due to death, transfer to another nursing home, or return to the community. Over the 2-year period, the study lost 49 residents and added 32 residents. Out of 23 residents who were followed at three times in our study, 11 started at Epoch 1 (2 residents remained through Epoch 4, but they were missed in the middle epoch) and 12 started at Epoch 2. Of the residents with data in only three epochs (n = 32), 23 were in the moderately impaired group and 12 were in the profoundly impaired group.
Table 1 presents baseline personal characteristics with their means and standard deviations, along with the means and ranges of the two measures of agitation over all four epochs. About 64% of the sample was female. Mean age was 82.2 years (SD = 8.4), mean cognitive status as measured by the MMSE was 8.0 (SD = 8.1), and mean functional status as measured by the Barthel Self-Care Rating Scale was 48.1 (SD = 13.4). At Epoch 1, staff-reported agitation as measured by the modified Nursing Home Behavior Problem Scale (NHBPS) was 15.2 (SD = 8.6) out of 56, and direct observation showed a mean of 18.0% (SD = 22.9) of observation with agitation out of 100% of observation. Medical record examination showed that 15% of the sample was diagnosed with Alzheimer's disease, 6% with vascular dementia, and 25% with mixed Alzheimer's disease/vascular dementia; 54% were marked "physician uncertain."
|
MMSE
The MMSE (Folstein, Folstein, & McHugh, 1975) is a simple and relatively quick assessment of cognitive mental status with a maximum score of 30. Trained research staff administered the MMSE.
Modified NHBPS
The full NHBPS (Ray, Taylor, Lichtenstein, & Meador, 1992) is a 29-item inventory of behavioral disturbances. The rater indicates the frequency of each behavior as observed over the past 3 days using a frequency-of-occurrence scale consisting of 5 points (0 = never, 4 = always). The sum of the 29 items indicates the overall score, with higher scores representing a greater frequency of behavior problems. For the larger study, research assistants administered this instrument to the resident's primary CNA during a face-to-face interview. Our analysis used only 14 items that directly corresponded to the operational definitions used in the computer-assisted behavioral observation system (CABOS; see Appendix). This allowed a possible range of 0 to 56. (We also used this modified scale in Vance, DeLaine, et al., 2003.) The interrater reliability reported by the authors ranged from.75 to.83. All behaviors are subsumed under Cohen-Mansfield's definition of agitation (Cohen-Mansfield & Billig, 1986).
CABOS
We used the CABOS to record the occurrence and duration of resident behaviors (Burgio et al., 1994). Kappa reliabilities for behavioral categories ranged from 0.78 to 1.00, with a mean of 0.95 across categories. The observation system sampled resident behaviors on a second-by-second basis throughout the day on the nursing units. Twelve 30-min observations were scheduled for each resident during a 3-week period. Each hourly interval was sampled between 8:00 a.m. and 8:00 p.m. The schedule of observations was generated through a stratified random time-sampling method with the condition that no more than one observation be conducted on an individual resident on any day. This condition was imposed in an attempt to distribute the observations evenly over the 3-week period. Consequently, the CABOS sampled 6 hr of each resident's time, during which agitation (e.g., verbal disruptions, restlessness, and physical aggression) could be recorded. We calculated the percentage of observation time during which agitation was observed, and we analyzed this percentage as the primary dependent variable in this study. In past studies, this method and schedule of observation have been sufficiently sensitive to detect changes in agitation in response to psychosocial interventions in the nursing home (e.g., Burgio et al., 2002).
Procedure
Before the observation sessions, the observer entered identifying information in the resident's computer file (e.g., participant number, date, and time of day). Only one resident was observed at a time. The observer then keyed the location, activity, sound, social environment, and restraint use of the target resident. The three behaviors corresponding to agitation were keyed when they occurred. Observers were instructed to keep the resident in view at all times during the observation period. Observers were present in the setting 2 to 3 weeks before observation to reduce observer reactivity. Haynes and Dantes (1987) indicated that lengthy preexposure has been found to decrease reactivity. Observers attempted to position themselves at the maximum distance from the target resident, and they were instructed to limit eye contact with the resident. If a resident was aware of the observer, the observer greeted the resident and then informed him or her that the observer would be doing some work on the other side of the room and that the resident should ignore the observer. It was the impression of the observers that residents quickly habituated to their presence.
A research assistant completed the paper-and-pen assessments by abstracting information from the residents' medical records or from face-to-face interviews with the residents' primary CNA. All paper-and-pen assessments were administered during the 3-week direct observation period. The specific timing of the assessments was determined by the availability of the CNAs; assessments were completed throughout the 3-week period. These procedures were repeated during each observation epoch.
Analyses
We compared the personal characteristics of the participants (N = 78) included in the analyses with those of residents who were excluded (n = 45) using t tests and a chi-square test. There were no statistically significant differences (p >.05) between the two groups at baseline with respect to the key variables agitation by staff report and direct observation, ADL, MMSE, age, and the proportion of gender. Furthermore, there was no statistically significant difference at baseline between residents who were followed through four time points (n = 55) and those who were measured at three time points (n = 23) regarding those key variables (p >.05). Thus, we combined the residents contributing data to Epochs 3 and 4 and based analyses on this sample of 78.
We combined the three behaviors subsumed under the definition of agitation to form an aggregated agitation variable. We used the HLM procedure (also known as multilevel analysis, random effects model, or mixed model) to estimate the trajectory of agitation over an 18-month period, and especially to examine the variability in baseline agitation and pattern of change. HLM has the capacity to analyze information about the rate and the pattern of change in targeted variables over multiple time points, taking into account inter- and intraindividual variability in change and cross-level interactions of time with predictors (Raudenbush & Bryk, 2002; Singer, 1998; Snijders & Bosker, 1999). We conducted HLM analyses separately for staff ratings and direct observation of agitation using the PROC MIXED procedure in SAS for Windows 9.1 (SAS Institute, 2002/2003). We examined main effects (i.e., factors that predict variability in baseline agitation) and cross-level interactions (i.e., interaction between predictors and time effect). We used restricted maximum likelihood estimation method with a specification of the unstructured covariance (Littell, Milliken, Stroup, & Wolfinger, 1996). In addition, we analyzed linear and curvilinear time effects. In post hoc analyses, we examined cognitive status (profoundly vs moderately impaired) modeling time as a regression variable to compare cognitive status at specific times. Based on a median split, we categorized residents as moderately (MMSE > 7) or profoundly impaired (MMSE 
7). In HLM procedures, we modeled continuous predictors (e.g., age, ADL) in the mean deviation form to reduce multicollinearity concern (Raudenbush & Bryk, 2002).
|
Results |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
Table 2 presents linear and quadratic effects of time, coefficients, and standard errors of personal characteristics on trajectories of agitation in staff reporting and direct observation. We detected no significant linear and quadratic effects of time in staff-reported NHBPS. That is, agitation changed little over the 18-month period according to staff report. In direct observation, however, both the linear effect (–3.65) and quadratic effect (1.21) were statistically significant (p <.05), indicating that the trajectory of agitation had a decreasing trend linearly, but the rate of decrease lessened over time.
|
7) was associated with more staff reported (3.60) and directly observed (11.83) agitation (p <.05). In both measures, personal characteristics did not predict longitudinal trajectory of agitation.
Influence of Mental Status on Agitation Trajectories
Correlational coefficients between modified NHBPS and CABOS over four time points were 0.33, 0.35, 0.35, and 0.49 (p <.01), indicating that both measures assessed comparable domains of agitation. Given the clinical importance of cognitive impairment in predicting agitation, we closely examined behavioral change as a function of cognitive impairment in staff reporting and direct observation. Using a median split, we found a different trajectory of agitation over the 18-month period between the profoundly impaired (MMSE 7) and moderately impaired (MMSE > 7) groups. At baseline, the mean MMSE score for the profoundly impaired group was 1.7 (SD = 2.5), and the mean MMSE score for the moderately impaired group was 15.4 (SD = 5.6). Due to a basement effect, the mean MMSE score for the profoundly impaired group had changed little at 18 months (2.0 ± 3.4), whereas the score for the moderately impaired group had declined to 9.5 ± 5.6. Agitation in residents who were moderately impaired remained stable over time; in the profoundly impaired group, the symptoms improved slightly up to 12 months and worsened at the 18-month point. We should note that we detected this differential trend only in direct observation. Staff reporting revealed approximately the same pattern for residents in the two groups of cognitive status (see Figures 1 and 2). To investigate possible explanations for the spike in agitation at 18 months, we examined whether there were differences in number of painful conditions and overall illness severity between the moderately and profoundly cognitively impaired groups. We found no differences.
|
|
7) = 25.66 – 7.95 x Time + 1.85 x Time2; and agitation for the moderately impaired group (MMSE > 7) = 9.82 – 0.73 x Time + 0.61 x Time2. Both linear and quadratic effects of time were statistically significant (p <.05) only in the profoundly impaired group. The equation for the moderately impaired group demonstrated little change in agitation over time. |
Discussion |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
The staff-report measure (modified NHBPS) asks the primary CNA to rate the frequency of 14 behaviors associated with agitation over the past 3 days. This retrospective period is much briefer than that of most other staff-report measures. For example, the Cohen-Mansfield Agitation Inventory (Cohen-Mansfield, 1986) requires that the staff recall a resident's behavior over the past 2 weeks. One would expect that agitation reported over a 3-day period would be more accurate. Nevertheless, staff reported that, on average, residents displayed the 14 agitation behaviors "sometimes." The HLM analysis indicated no linear or quadratic trends across 18 months (see Figure 1) and no significant variability in intercept. Thus, staff continued to report resident agitation as occurring "sometimes" over 18 months with very little variability.
Direct observation showed agitation occurring during 18% of total observations at the Epoch 1 data collection point. This percentage is well within the range reported in our prior research (Burgio, Allen-Burge et al., 2001; Burgio, Butler et al., 2001; Burgio et al., 2002). We found both significant linear and quadratic effects over the 18-month period, showing that residents' agitation improved through the 12-month observation (Epoch 3) but increased from Epoch 3 (12 months) to Epoch 4 (18 months; see Figure 2).
The only personal characteristic found to be related to change in agitation was cognitive status. Age, degree of ADL impairment, and gender were not significant. To further investigate the effects of cognitive status, we created two groups of residents based on a median split (MMSE 7; MMSE > 7). Due to the extremely low MMSE scores in the 7 group (M = 1.7), we labeled the group profoundly impaired. With an average MMSE score of 15.4, the >7 group was termed moderately impaired. Because inclusion in the study was based on degree of agitation and not MMSE score or dementia diagnosis, we included an outlier with an MMSE score of 26 in the moderately impaired group. We included this participant's data in the analysis because this person still met entry criteria.
Because of a basement effect, MMSE remained stable across the 18 months in the profoundly impaired group. However, in the moderately impaired group, the mean MMSE scores declined from 15.4 to 9.5 over 18 months. Again, we found no significant trends in either cognitive impairment group using the staff-report NHBPS (Figure 1). It is interesting to note that the direct observation system showed no significant trend in agitation over the 18 months in the moderately impaired group, even though there was a significant decline in MMSE scores (15.4 to 9.5). The scores hovered at 10% of observations during the 18-month study. Similar to the results of the total group, there was a significant linear and quadratic trend in agitation for the profoundly impaired group, with mean agitation decreasing from 26% to 21% of the intervals from Epoch 1 to Epoch 3 (12 months) and increasing to 29% at Epoch 4 (18 months).
As discussed previously, cross-sectional studies using staff report have shown a linear, inverse relationship between cognitive impairment and agitation using both staff report and direct observation (Beck et al., 1998; Vance, DeLaine et al., 2003). Our longitudinal findings of a linear and quadratic relationship between cognitive status and agitation partially support Algase and colleagues' (1996) model. Residents in our profoundly cognitively impaired group were most vulnerable to environmental stress. They displayed significantly more agitation than the moderately impaired group, whose agitation remained unchanged over 19 months with both staff report and direct observation measurement. Although the profoundly impaired group showed some improvement from Epoch 1 to 12 months (26% of intervals to 21%), their agitation jumped to 29% at 18 months. In an attempt to explain this marked increase in agitation, we examined a possible group difference in the number of medical diagnoses and medications (proxies for more severe illness). The groups did not differ on these variables.
Although we cannot explain this increase with empirical data, we do offer a hypothesis. Considering that Epoch 1 data indicated a mean MMSE of 1.7, it is possible that some of these residents were near death 18 months later. Various researchers have found that an increase in agitation occurs near death and may actually be a marker for near death (Allen, Burgio, Fisher, Hardin, & Shuster, 2005; Brajtman, 2003; Jones, King, Speck, Kurowsaka, & Tookman, 1998). Unfortunately, in this study we do not have data on when residents died. Still, we consider "terminal agitation" a plausible explanation.
We believe that there are important clinical implications for both our longitudinal trajectory data and our measurement systems used to collect these data. Our results show that profoundly cognitively impaired nursing home residents display more agitation than moderately impaired residents. Even though the data show that agitation may decrease somewhat over time, agitation may worsen at a later point. Current nursing home personnel can use this information when allocating resources, particularly staff-to-resident ratios, while contemplating their case mix.
Also, with the trend of keeping cognitively impaired residents out of the nursing home with the use of dementia units in assisted living, nursing homes will include a greater percentage of residents with profound cognitive impairment, and thus more agitation. This knowledge can help treatment planners and policy makers predict the resources needed to maintain nursing homes in the near future.
We and others have argued in past research that staff-reported estimates of agitation underestimate agitation (McCann et al., 1997; Vance, Burgio et al., 2003). The current data suggest that staff-report measures are relatively insensitive to change over time. Although the CABOS system is more sensitive, it can be expensive, it requires training to use, and it can be too time intensive for everyday use in the nursing home. We believe that there is a reasonable alternative. Developers designed the Pittsburgh Agitation Scale as a brief, user-friendly observation measure that provides reliable and valid data from staff on four categories of agitation: aberrant vocalization, motor agitation, aggressiveness, and resident care. Thus, it is a hybrid observation/staff-report measure. The authors validated the measure against the CABOS in the original article (Rosen et al., 1994). In using this measure, nursing staff observe the occurrence and severity of the four categories for each of their assigned residents during their shifts. The measure is user friendly because staff complete recording on the four categories only once at the end of their shift.
There are limitations to this study. First, although most definitions of agitation include wandering, we did not record wandering in this study. Wandering is defined as "aimless ambulation." Recording wandering requires that the observer sample an adequate amount of ambulation to judge whether it is aimless. The CABOS records behaviors in real time (i.e., as they occur). This requires an event (e.g., throwing objects, screaming) that notifies the observer to press the agitation key. As discussed previously, wandering does not offer a discrete event that the observer can identify in real time to trigger recording. There are no data to indicate that wandering, unlike screaming, is an integral component of agitation—particularly in nursing homes where resident egress is difficult. Thus, although we believe that the inclusion of wandering would have been preferable, the definition of agitation used in this study is still valid. (Since we developed the CABOS, other researchers have developed effective systems that record wandering in real time: Algase, Beattie, Bogue, & Yao, 2001; Algase, Beattie, & Therrien, 2001).
Second, because we did not use an MMSE cut score for participant entry, an outlier with agitation and an MMSE of 26 met entry criteria and was included in the moderately impaired group, in spite of the fact that he would not be considered moderately impaired by most researchers and clinicians. We could not justify dropping this resident's data. Thus, we should note that, even though the average MMSE score for this group was 15.4, this group included some individuals who would be considered mildly cognitively impaired.
Finally, our study focused on the personal characteristics included in Algase and colleagues' (1996) model. Many other variables such as environmental factors, emerging versus established behaviors, and type of dementia can affect agitation. Study of these and other variables is important to truly understand the nature of agitation. Additional research is needed to ascertain the contribution of these variables to the occurrence of agitation.
In conclusion, the results of this study show that profoundly cognitively impaired nursing home residents display higher rates of agitation than moderately impaired residents when studied using direct observation methods. Longitudinal data suggest that moderately impaired residents show stable agitation over an 18-month period, but profoundly cognitively impaired residents show an initial improvement over 12 months, followed by an increase in agitation at 18 months. These data have policy implications both for current nursing homes and nursing homes in the near future. The data suggest that direct observation is a more sensitive measure of agitation compared with staff report but one that may not be feasible for day-to-day use. User-friendly observation measures such as the Pittsburgh Agitation Scale (Rosen et al., 1994) are an alternative to both direct observation and staff-report measures.
|
Footnotes |
|---|
1 Center for Mental Health and Aging, The University of Alabama, Tuscaloosa.
2 Department of Psychology, The University of Alabama, Tuscaloosa.
3 School of Social Work, The University of Alabama, Tuscaloosa.
4 Applied Statistics Program, The University of Alabama, Tuscaloosa.
Decision Editor: Linda S. Noelker, PhD
Received for publication April 10, 2006. Accepted for publication April 2, 2007.
|
References |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
| ||||||||||||||||||||||||
| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
