| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
| ||||||||||||||||||||||||
Correspondence: Address correspondence to Jules Rosen, MD, 3811 O'Hara Street, Pittsburgh, PA 15213. E-mail: rosenji{at}upmc.edu
 |
Abstract |
|---|
 TOP Abstract Caregiver Characteristics and...Research Setting and MethodsModel Testing and EstimationResultsDiscussionReferences |
|---|
Key Words: Adaptation-level theory • Job satisfaction • Perception gap • Quality of life
To measure quality of life, expert surveyors interview residents; their ratings on quality-of-life dimensions enable researchers to draw conclusions about their quality of life. Although nursing home residents themselves are the preferred source of information, it may not be always feasible to obtain information from residents (Kane, 2001). Medical conditions such as depression, dementia, and aphasia may render the residents unwilling or unable to provide quality-of-life ratings. In these cases, researchers must obtain information from proxy informants such as family members or caregiver staff. As expected, each type of proxy-based inference moves further away from the self-report (Kane; Pickard & Knight, 2005). This, as we review next, results in a perception gap in quality-of-life ratings among residents and caregivers.
Empirical studies have specifically examined the perception gap in quality-of-life ratings. Kane and colleagues (2000) compared quality-of-life appraisals by residents and staff and found significant differences in matched-pairs item responses. Epstein, Hall, Tognetti, Son, and Conant (1989) found that caregiver proxies tended to provide lower ratings than residents did on dimensions such as emotional health and satisfaction. Similarly, Berlowitz, Du, Kazis, and Lewis (1995) found differences in quality-of-life scores between nursing home residents and caregiver proxies. Consistent with past studies, proxy evaluations were lower than resident evaluations. Berlowitz and colleagues concluded that "We cannot be certain as to why differences exist between patient and proxy perceptions of health-related quality of life. ... Additional studies will be required to clarify these relationships" (p. 801). In this article we seek to answer this call by examining factors associated with the perception gap.
To be sure, there are some studies that have examined correlates of the perception gap based on proxy caregivers, though the results have been unclear and inconsistent. In their literature review of more than 40 studies examining the perception gap, Sprangers and Aaronson (1992) reported that fewer than five studies that had examined caregiver attributes as correlates of the perception gap, and the findings were decidedly mixed. For instance, whereas Muhlenkamp and Joyner (1985) found that nurses with more training provided more accurate ratings of patients' affective states, Husted and Johnson (1985) found that older or more experienced nurses underestimated patients' level of hostility. Magaziner, Simonsick, Kashner, and Hebel (1988) examined mean differences in proxy and patient scores and found larger differences for women, proxies younger than 65 years of age, and proxies who were not relatives. Specifically, younger proxies significantly underrated patients' functional independence.
All of these studies used a bivariate approach to examine the extent to which caregiver demographics predict the perception gap. The only study we found that took a multivariate approach was by Rothman, Hedrick, Bulcroft, Hickam, and Rubenstein (1991). Instead of examining predictors of the perception gap, Rothman and coauthors ran two separate regression analyses: one predicted patient-generated scores, and the other predicted proxy-generated scores on psychosocial dimensions of patient health. In terms of predicting patient-generated scores, they found both the age and education of the proxy to be nonsignificant, though self-perceived health status was a significant predictor. For proxy-generated scores, age was a significant predictor along with psychological stress, subjective burden, and objective burden. However, because no perception gap was computed, we cannot ascertain what factors may have predicted the gap. Moreover, it should be noted that many of these studies do not distinguish between different types of proxies (e.g., family member vs nursing home employee). As we explain later, this becomes important because different focal, background, and residual factors are associated with different types of proxies.
Knowing factors that are systematically associated with the perception gap can enable administrators to reduce the gap. Note that a gap can be conceptualized on the basis of magnitude as well as direction. The magnitude can be large or small, and the direction can be positive (caregivers have higher ratings than the patient) or negative (caregivers have lower ratings than the patient). Thus, it can be possible to have a large negative gap (caregivers give very low ratings compared with patients) or a large positive gap (caregivers give very high ratings compared with patients). A smaller gap can ensure that the caregivers more accurately identify residents' needs and wishes. This information would be invaluable in designing quality-improvement processes that focus on quality of life and not just regulation-mandated clinical outcomes (Oleson, Heading, Shadick, & Bistodeau, 1994; Willging, 2003). Further, understanding the correlates of the perception gap is also important for designing quality-of-life measurement systems. Preferably, these data should be collected directly from residents. However, in nursing homes, the use of proxies is likely given the higher proportion of cognitively impaired residents. In such situations, knowledge of predictors of the perception gap is important, as these predictors can be used as covariates to statistically adjust proxy ratings for analysis and reporting purposes.
|
Caregiver Characteristics and the Perception Gap |
|---|
|
TOPAbstractCaregiver Characteristics and...Research Setting and MethodsModel Testing and EstimationResultsDiscussionReferences |
|---|
According to Helson (1964), the degree of adaptation is a function of three sets of factors: (a) focal factors, (b) background factors, and (c) residual factors. Focal factors are those that directly influence the gap, or the degree of adaptation. Although Helson is silent about what would constitute a focal factor, research in many service settings shows that the level of satisfaction experienced by the service provider is a critical antecedent of many judgments and actions of the service provider, many of them directly impacting customer satisfaction and service quality. For instance, in his study of postsurgical care in nine hospitals, Gittell (2002) found that a strong relationship among various providers positively influenced patient satisfaction. Similarly, Stock and Hoyer (2005) showed the impact of salespersons' psychological orientation on customer satisfaction. These studies exemplify a growing realization that employee satisfaction affects customer satisfaction. To the extent that we can argue that caregivers with higher satisfaction with their work will be more involved with resident care, we can also argue that caregivers with higher job satisfaction are more likely to be aware of the residents' clinical conditions and the general factors surrounding their well-being. Thus, when making quality-of-life assessments, these staff members will factor in all the relevant factors and have more accurate perceptions of the residents' quality of life. This logic suggests that the higher the caregiver's job satisfaction, the smaller will be the magnitude of the perception gap. Our formal statement is this: We hypothesize that caregivers' satisfaction with their job will be related to the perception gap such that the higher the caregivers' job satisfaction, the smaller the perception gap.
Background factors are related to the setting in which the focal processes operate. For instance, characteristics of the nursing home facility—its structure and the processes of the units where the residents are housed and where the caregivers work—are deemed to be background factors. More broadly, regulatory and social forces may also affect or be part of the background factors. For instance, the pervasive use of clinical outcomes by regulators may affect the structure and the processes used in the nursing home, and thereby affect the perception gap. Regarding background factors, we posit that they will directly influence the perception gap, though we do not make any directional prediction. Notably, different structures and processes of a caregiving unit can potentially draw the caregivers' attention to different aspects of the residents. For instance, given the same level of adaptation, a unit with more acute residents may have a larger perception gap than a unit with nonacute residents. Given that our interest is in focal factors and because of the nature of our data, we do not test directional hypotheses about the impact of background factors. However, we do hypothesize that background factors will have an impact on the perception gap.
Finally, according to Helson, residual factors are characteristics of the individuals who must adapt to the stimulus, in this case characteristics of the caregivers. Helson's theory includes both demographic factors as well as personality variables (i.e., traits) as residual factors. Importantly, the impact of residual factors on adaptation is mediated by means of focal factors, because focal factors provide the psychological process by which residual factors impact the level of adaptation. This process-based prediction is fully consistent with other prominent models that posit the self as the link between characteristics of an individual (e.g., age, gender, race, and marital status) and goal-directed activities and adaptation (Graziano, Jensen-Campbell, and Finch, 1997). These factors affect the most proximal psychological phenomenon—job satisfaction—rather than more distal phenomena such as caregivers' assessments of residents' quality of life. To the extent that they influence job satisfaction, residual factors are hypothesized to impact the perception gap. In other words, we hypothesize that residual factors directly influence the perception gap such that their influence is mediated by means of job satisfaction, that is, focal factors.
The hypotheses we test lead to the general model shown in Figure 1 and are theorized based on Helson's adaptation-level model. First, we hypothesize that focal factors (i.e., job satisfaction) will affect the perception gap. Second, we hypothesize that focal factors will mediate the effect of residual factors (i.e., caregiver characteristics) on the perception gap. Finally, we hypothesize that the background factors (i.e., facility characteristics) will be associated with the perception gap. In testing these hypotheses, we explicitly note that variables signifying focal, background, and residual factors are limited to those that were empirically measured in our study. We do not claim that we have an exhaustive or the "correct" list of variables. Only through additional empirical research can we establish a meaningful typology of variables that are focal, background, or residual in nursing home contexts. The examined focal factors include five dimensions of caregivers' job satisfaction; the background factors represent differences in the units where care is provided; and demographic characteristics of caregivers represent the residual factors. These are all explained in detail in the next section. Our exploratory stance about the impact of background and residual factors on the perception gap is fully consistent with Helson (1964, p. 59), who argues that the inclusion and categorization of background and residual factors "is largely a matter of convenience and depends upon the ‘sense’ of the experimental situation. In identifying sources of variance in behavior, presumably focal and background stimuli are experimentally controlled, leaving all other sources of variance to be classified as residual."
|
|
Research Setting and Methods |
|---|
|
TOPAbstractCaregiver Characteristics and...Research Setting and MethodsModel Testing and EstimationResultsDiscussionReferences |
|---|
Concurrently, clinical staff provided their perceptions of the quality-of-life dimensions for the residents on their unit by using the same instrument as the residents; they also filled out a Job Satisfaction Scale. We had staff and resident surveys concurrently administered at each facility in June–July 2002, November–December 2002, and June–July 2003. Each wave of data collection lasted 2 weeks. This study was reviewed and approved by the University of Pittsburgh Institutional Review Board.
Resident Quality of Life
The resident quality-of life instrument assesses 11 dimensions, which are shown in Table 1. The validity and reliability of this instrument has been previously described (Kane et al., 2003). To facilitate comparison with staff surveys, we used a summary item for each dimension. Both residents and staff used a 4-point Likert-type scale with points labeled as excellent, good, fair, and poor.
|
There were 223, 227, and 218 residents eligible to be approached at Waves 1, 2, and 3, respectively. Of those eligible, 55 (25%), 75 (34%), and 49 (22%) were unable to complete the interview as a result of cognitive impairment, and 18 (11%), 17 (11%), and 30 (18%) refused (either themselves or their families) at Waves 1, 2, and 3 respectively. The overall participation rate among residents averaged 62% across the three waves and was similar for both facilities. The average ratings on the 11 dimensions for both facilities are shown in Table 1 and provide the quality-of-life indicators on which the perception gap is calculated.
When we examine the correlation among the 11 dimensions aggregated across both facilities, we find they are highly intercorrelated; correlations range from.71 to.89 (all ps <.001). Estimating separate models for each dimension would assume that each dimension is independent of the others. This can seriously bias the estimates. As we explain later, we address these issues in the empirical analysis through the use of a hierarchical linear model.
Staff Surveys
We surveyed nursing home personnel who had regular contact with residents during their work. Thus, we included administrators, nurse supervisors, registered and licensed nurses, certified nursing assistants, therapists (physical, occupational, recreational), therapy aides, and housekeeping or food service staff. We excluded from the sample those employees who were on leave or did not work full time during the 2-week data-collection window. There were a total of 325, 318, and 331 eligible employees at Waves 1, 2, and 3, respectively. To ensure anonymity, we numerically coded the surveys and removed identifiable data prior to returning the completed documents. The proportion of eligible employees who returned completed surveys was 76%, 75%, and 72% for Waves 1, 2, and 3, respectively.
The survey primarily measured staff members' job satisfaction and perceptions of resident quality of life. The Job Satisfaction Scale (Boyt, Lusch, & Naylor, 2001) is a 15-item instrument that assesses five dimensions of work satisfaction: (a) work content, (b) compensation, (c) promotion opportunities, (d) coworkers, and (e) superiors. This scale has been described in Castle and colleagues (2004), and the five dimensions have moderate to high internal consistency ranging from.61 to.82. We computed the score on each dimension as the average of the items comprising that dimension (see Table 2).
|
Employees took about 30 minutes to complete the survey and were compensated $10 for their time. Response rate was similar at both facilities for all three waves and ranged from 60% to 70%. A description of the caregivers who completed surveys is shown in Table 3.
|
|
|
Model Testing and Estimation |
|---|
|
TOPAbstractCaregiver Characteristics and...Research Setting and MethodsModel Testing and EstimationResultsDiscussionReferences |
|---|
A general approach to analysis for such data is the hierarchical linear model, or HLM. Technical details about this methodology are described in several sources (e.g., Raudenbush & Bryk, 2002; Singer & Willett, 2003). The HLM subsumes ordinary least squares (also known as single-level fixed-effects models), random-coefficients regression, and repeated-measures analysis of variance as special cases. In our multilevel model, we treat the units for which the caregiver works as random effects and the individual-level variables (demographic and job-satisfaction measures of each caregiver) as fixed effects. Thus, the background factor or units (random effects) are those that are assumed to be normally distributed with zero mean and some unknown variance. Declaring unit as a random effect sets up a common correlation among all caregivers within the same unit. It also implies that the units we observe are a sample of the set of unobserved units, and thus we can generalize to other units as well (Singer & Willett).
|
Results |
|---|
|
TOPAbstractCaregiver Characteristics and...Research Setting and MethodsModel Testing and EstimationResultsDiscussionReferences |
|---|
To examine the impact of factors related to the perception gap, we estimated a series of models by using PROC MIXED in the SAS software. The fixed-effects parameters are reported in Tables 5 and 6. Table 5 reports results pertaining to the three sets of factors. Note that the focal factors—job-satisfaction dimensions—have a statistically significant impact on the perception gap (Model B in Table 5). Background factors (i.e., the unit) also have a direct impact on the perception gap. However, residual factors (i.e., caregiver characteristics) are not significant when included just by themselves (Model A in Table 5). Model A in Table 5 ascertains, even without the job-satisfaction dimensions, that the statistical significance or lack thereof of the background and residual factors remains unchanged. Thus, a model positing direct effects of background and residual factors can safely be rejected. In other words, it is evident from Table 5 that demographic characteristics of caregivers (i.e., the residual factors) do not directly impact the perception gap. It is important to note that, in Model B, three out of five focal factors of job dimensions are statistically significant, and a fourth one approaches statistical significance (p <.10) in predicting the perception gap.
|
|
|
Discussion |
|---|
|
TOPAbstractCaregiver Characteristics and...Research Setting and MethodsModel Testing and EstimationResultsDiscussionReferences |
|---|
Conceptually, these results clarify null findings from previous studies. For instance, Kane and colleagues (1997) tried to predict the importance of control and choice in nursing homes on the basis of staff characteristics such as (shift, gender, age, ethnicity, education, length of employment, city of nursing home, and ownership of nursing home). Similar to Model A shown in Table 5, they found null results. Our classification of these variables as residual factors suggests that, although they do not directly influence the perception gap, caregiver characteristics influence other important constructs such as job satisfaction. Further research is needed to identify additional intervening constructs that may directly affect the perception gap.
Practical Implications
In a long-term-care setting, ratings of a cognitively intact resident are the gold standard for measuring quality of life. However, the gold standard is frequently not available because of the high level of residents' frailty and cognitive impairment in long-term-care settings. Nevertheless, as adaptation-level theory predicts and empirical studies show, the perception gap is an empirical reality that administrators must address. Presumably, the goal of assessing quality of life among nursing home residents is to improve it (or to evaluate the extent to which different interventions improve it). Increasing the agreement between residents and caregivers should increase the accuracy of such measurements by reducing caregiver bias associated with job satisfaction, which is a focal factor in Helson's adaptation-level theory.
This makes job satisfaction a key priority to be addressed by nursing homes. To outline the practical implications of the gap, we plotted the perception gap as a function of three levels of job satisfaction (low, medium and high for three job-satisfaction dimensions—work, pay, and promotion—that were statistically significant at the 5% level) in Table 5. Results are depicted in Figure 2 and show the following: As job satisfaction increases, the magnitude of the perception gap tends toward zero. Recall that the perception gap, which largely has a negative direction, is calculated as employee minus resident rating. In other words, employee ratings are lower than resident ratings. The positive coefficients for job satisfaction in Table 5 imply that every unit change in satisfaction decreases the magnitude of this negative perception gap (which is negatively coded). Thus, caregivers with higher job satisfaction have relatively more accurate perceptions of resident quality of life. By implication, in addition to potential human resource benefits such as decreased turnover, increasing job satisfaction can better align residents' and caregivers' perceptions. This could potentially increase both job satisfaction among caregivers and the quality of care delivered. However, more empirical research is needed to address this issue.
|
|
Also important will be research that addresses the measurement issues presented here. We used preexisting surveys to create our measure of the perception gap. Surveys that follow Pickard and Knight (2005) to create proxy-evaluation studies are needed. It is particularly important to ensure that the proxies understand the viewpoint they take when providing their ratings. As the literature on understanding proxy ratings of quality of life evolves, the development of such instruments will be critical. In this regard, studies examining the quality-of-life gap in related populations such as patients with dementia can be informative (Pickard & Knight). It should also be noted that, in our study, resident ratings of quality of life were limited to those residents who could complete the survey, whereas employee ratings included both patients with and without dementia. Although we do not know for sure if patients with dementia actually had lower quality of life, it is quite possible that employees perceive these patients to have lower quality of life. This could be one factor driving the negative perception gap. Related to this, two facts are worth noting. First, most previous studies examining the quality-of-life perception gap have found a negative gap. Second, in our study the variability in the gap was very large, with gap scores ranging from –.2.5 to 1.7. This means that despite a negative overall mean, there is tremendous variability in the perception gap. Both of these issues deserve further exploration.
These results should prompt future researchers to reexamine generic models such as those specified in Chou, Boldy, and Lee (2003, Figure 1). They treat the trait characteristics (e.g., gender, race) and satisfaction states as interchangeable and directly affecting quality-of-life-related outcomes. Researchers may benefit by first classifying them as focal, background, or residual factors and then deciding if each set of factors has a direct or indirect influence on quality-of-life perceptions.
A closer scrutiny of the residual factors is warranted. Only some of these differences can be explained intuitively. For example, it is intuitive that caregivers who have longer tenure at their jobs would be more dissatisfied with opportunities for promotion in a workplace that does not encourage career ladders or advancement. In contrast, the diversity in the magnitude and direction of the impact of marital status, dependents, or experience itself deserves additional research. Not only do we confirm the importance of job satisfaction at nursing homes, but we show that a single shotgun approach to all caregivers is likely to be unsuccessful. Clearly, different dimensions of job satisfaction have different meanings and associations for different caregivers.
Using adaptation-level theory, we provide a framework to hypothesize the existence of the perception gap and think about its antecedents. The daunting task of conceptually elucidating and empirically testing other factors that are antecedents of the perception gap remains. As Helson himself remarked, only a programmatic body of empirical research can forge the list of candidate focal factors. We hope our work is the first step in that direction.
|
Footnotes |
|---|
1 Katz Graduate School of Business, University of Pittsburgh, PA.
2 Department of Psychiatry, University of Pittsburgh, PA.
3 Graduate School of Public Health, University of Pittsburgh, PA.
4 Geriatric Research, Education and Clinical Center, VA Pittsburgh Health Care System, Pittsburgh, PA.
5 RAND Corporation, Pittsburgh, PA.
6 Department of Internal Medicine, University of Pittsburgh, PA.
7 School of Business, University of Mississippi, Oxford.
Decision Editor: Linda S. Noelker, PhD
Received for publication July 20, 2006. Accepted for publication November 27, 2006.
|
References |
|---|
|
TOPAbstractCaregiver Characteristics and...Research Setting and MethodsModel Testing and EstimationResultsDiscussionReferences |
|---|
| ||||||||||||||||||||||||
| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
