Values and Preferences of Individuals With Dementia: Perceptions of Family Caregivers Over Time

Abstract

Purpose of the Study: Cross-sectional evidence indicates that family caregivers reporting on the importance of daily care values and preferences of individuals with mild-to-moderate dementia consistently report less importance than individuals with dementia (IWDs) self-report. Discrepancy is primarily associated with caregivers’ beliefs about care. However, research has yet to address how caregivers’ perceptions of IWDs’ values and preferences change over time and how change is related to caregivers’ well-being. Design and Methods: The sample includes 198 dyads of an individual with mild-to-moderate dementia and a spouse or child providing assistance. Linear growth curve modeling was applied to investigate how caregivers’ perceptions of importance of IWDs’ values and preferences in daily care change over 4 years and the factors associated with change. Results: Caregivers’ perceived importance of IWDs’ values and preferences significantly decreased over time, with significant between-person differences in level and slope. Between-person differences in level were associated with caregivers’ beliefs about care, IWDs’ self-reported importance of values and preferences at baseline, and kin relationship. Changes in caregivers’ reports of importance covaried with caregivers’ own quality of life over time. Implications: Findings suggest that caregivers increasingly de-emphasize the importance of values held by their IWDs. Although creating this emotional distance may be adaptive, caregivers may become increasingly unable to make decisions that effectively represent IWDs’ own preferences.

As individuals with dementia (IWDs) experience decline in memory and decision-making capacity, caregivers inevitably slide into a role of progressive surrogacy, making more and more decisions of greater significance on behalf of IWDs (Elliott, Gessert, & Peden-McAlpine, 2009). However, accumulating evidence suggests a lack of concordance between surrogates and elders regarding care preferences, despite whether (a) the individual has mild-to-moderate dementia or no cognitive impairment, (b) the reports are about daily or end-of-life preferences, (c) the interviews are semi-structured, vignettes, or batteries of questions, or (d) the proxy reporter is a family member or physician (Carpenter, Lee, Ruckdeschel, Haitsma, & Feldman, 2006; Ditto et al., 2001; McCullough, Wilson, Teasdale, Kolpakchi, & Skelly, 1993; Reamy, Kim, Zarit, & Whitlatch, 2011; Uhlmann, Pearlman, & Cain, 1988). Surrogates are not as accurate as elders, creating a conundrum of who should be considered the “gold standard” reporter of IWDs’ preferences in care (Karel, Moye, Bank, & Azar, 2007). In some cases, surrogates are correct fewer times than chance alone predicts (Fagerlin, Ditto, Hawkins, Schneider, & Smucker, 2002; Uhlmann et al., 1988). Such discrepancies are likely related to factors that affect dyadic communication (i.e., individual beliefs in care) or lack of a shared vision of the challenges posed by dementia care.

Cross-sectional evidence suggests that family caregivers consistently report perceiving IWDs’ care values as less important for IWDs than IWDs self-report, and that such discrepancy is primarily associated with caregivers’ beliefs about IWDs’ involvement in decision making (Reamy et al., 2011). This finding, along with evidence suggesting that caregivers reappraise their situation with experience and the increase of behavior problems over time (Zarit, Todd, & Zarit, 1986), indicates the importance of understanding caregivers’ perceptions over time. Furthermore, research on end-of-life preferences suggests a possible adjustment of preferences as one approaches death (Winter & Parker, 2007). A similar change may occur for caregivers, altering their perceptions of IWDs’ daily values as illness worsens. However, research has yet to address if caregivers’ perceptions of IWDs’ daily values and preferences change or remain stable over time.

Caregivers’ perceptions of IWDs’ preferences about care may be affected by many factors. Demographic characteristics (i.e., kin relationship) or caregivers’ beliefs about care may be related to perceptions over time. Furthermore, theoretical work suggests that negative affect can alter one’s interpretation of life events as more negative (Beck, Rush, Shaw, & Emery, 1979). Caregivers’ self-perceived quality of life may alter, or be related to, caregivers’ perceptions of their relatives’ values and preferences in care. For example, if individuals are depressed, they may employ a persistent negative lens, judging other people and events more negatively as well (Beck et al., 1979). Ditto and colleagues (2003) demonstrate a relation between depression and change in reports of preferences for end-of-life treatment. Thus, it seems possible that changes in other indicators of well-being over time, such as quality of life, may affect how caregivers appraise IWDs’ values and preferences in daily care.

Ultimately, caregivers’ perceptions of IWDs’ values and preferences are likely to influence the daily decisions that are made for IWDs. Caregivers’ appraisals of the importance of specific values, such as autonomy, may dictate how caregivers subsequently treat their relatives. Thus, there is a need to understand how caregivers’ perceptions of IWDs’ values and preferences change over time and how change may be related to caregivers’ self-reported characteristics, particularly their own quality of life.

This study addresses the following questions: (1) Are caregivers’ perceptions of importance that IWDs place on their values and preferences changing over time? (2) Is change related to caregivers’ ratings of their own quality of life over time when accounting for baseline dyadic characteristics? Given prior evidence that caregivers report perceiving their IWDs’ as placing less importance on their values and preferences in daily care than IWDs self-report, we hypothesized that (a) caregivers’ reports of total importance of IWDs’ values and preferences will decrease over time and (b) there will be significant between-person differences in level and rates of change. Second, we hypothesized that changes in perceptions of importance will covary with caregivers’ self-reports of quality of life over time.

Method

Participants

Dyads (N = 198) consisting of an individual with mild-to-moderate dementia (IWD) and his or her spouse or child caregiver were recruited from research and service organizations in the San Francisco Bay area and Cleveland, Ohio. Eligible caregivers had to be the primary family member assisting an IWD with a confirmed diagnosis of dementia (n = 175) or symptoms of progressive memory problems (n = 23) who was living at home rather than in a congregate setting. Most IWDs had scores between 13 and 27 on the Mini-Mental Status Examination (MMSE; Folstein, Folstein, & McHugh, 1975); however, inclusion criteria also took into account other indicators of functioning and education. Individuals with high MMSE scores (27 or more; n = 31) were included if they had confirmed diagnoses of dementia. Additionally, individuals with low MMSE scores (<13; n = 4) were included if they had low formal education, their overall functioning was typical of mild-to-moderate dementia, and they could complete the interview. For more information about recruitment, see Whitlatch, Piiparinen, and Feinberg (2009).

Caregivers were predominantly women (78.0%, n = 154), aged 30 to over 90 (M = 65.21 years, SD = 13.97; at the request of the Institutional Review Board, ages above 90 were recoded to 90, to minimize identification of individuals based on age). Caregivers were primarily spouses of the IWD (60.1%, n = 119). About one third of the caregiver sample (n = 57) was African American. About half of the IWDs were women (48.0%, n = 95) and 27.3% (n = 54) were African American. IWDs were between 39 and over 90 years of age (M = 75.45 years, SD = 9.09) and had MMSE scores between 9 and 30 (M = 22.14, SD = 4.64; Table 1).

Table 1.

Characteristics of Caregivers and IWDs

	M	SD	Range
Caregivers
Baseline age	65.21	13.97	30–90+
Baseline education	4.30	1.19	2–6
Baseline income	5.29	2.04	1–8
Women (yes = 1)	0.78	0.42	0–1
African American (yes = 1)	0.29	0.45	0–1
Spouse (yes = 1)	0.60	0.49	0–1
Baseline employment (yes = 1)	0.33	0.47	0–1
Baseline coresidence (yes = 1)	0.60	0.49	0–1
Baseline length of caregiving (months)	30.80	37.40	0–265
Baseline best interests	2.80	0.20	2.29–3
Baseline decision involvement of IWD	1.90	0.74	0.00–3
Baseline dyadic strain	1.42	0.42	0–2.36
Baseline discussion of IWDs’ wishes	2.37	1.26	1–4
Quality of Life over time	2.96	0.44	1.85–3.95
IWDs
Baseline age	75.45	9.09	39–90+
Women (yes = 1)	0.48	0.50	0–1
African American (yes = 1)	0.27	0.45	0–1
MMSE score	22.14	4.64	9–30
Baseline importance of values and preferences	94.96	9.58	59.77–111.00

Note: Dyad N = 198; IWDs = individuals with dementia; MMSE = Mini-Mental Status Examination.

Procedure

Each member of the dyad was interviewed separately at home on two to five occasions, each approximately 1 year apart between 1998 and 2006. Visual aids for response choices were used for all participants to minimize burden on memory. Although content of the interviews differed slightly across the two merged samples, complete data were available on caregivers’ reports of importance of IWDs’ values and preferences in care and for the other measures used in these analyses. Cases dropped out over time due to death of the caregiver or IWD, placement in an institutional setting, or general attrition (Table 2).

Table 2.

Participant Attrition by Wave of Data Collection, Spaced Approximately 1 Year Apart

	T1	T2	T3	T4	T5
Total	N = 198	N = 73	N = 102	N = 64	N = 33
Sample 1
Interviewed	n = 97	—	n = 39	n = 23	n = 11
Lost to bereavement	n = 0	—	n = 24	n = 9	n = 6
Lost to placement	n = 0	—	n = 10	n = 3	n = 1
Lost to general attrition	n = 0	—	n = 24	n = 4	n = 5
Sample 2
Interviewed	n = 101	n = 73	n = 63	n = 41	n = 22
Lost to bereavement	n = 0	n = 4	n = 6	n = 6	n = 5
Lost to placement	n = 0	n = 11	n = 3	n = 9	n = 1
Lost to general attrition	n = 0	n = 13	n = 1	n = 7	n = 13

Note: Sample 1 and Sample 2 represent the two recruitment sites (San Francisco Bay area and Cleveland, Ohio).

Measures

Dependent Measure.—The Values and Preferences Scale (VPS; Whitlatch, Feinberg, & Tucke, 2005) consists of 37 items that assess IWDs’ overall importance of values and preferences in care (i.e., not specific care tasks) from both the caregiver’s perspective and the IWD’s perspective. Items were rated on a 3-point scale ranging from 1 (not at all important) to 3 (very important) by caregivers at each measurement point and by IWDs at baseline. Given our interest in caregivers’ perceptions, only the caregivers’ perceptions were examined as the dependent variable. While previous work has explored the use of this scale as a compilation of factors (Reamy et al., 2011; Whitlatch et al., 2005), this study used an overall level of perceived importance by multiplying mean scores by the number of items (M = 86.75, SD = 12.93; range = 48–110; α = .87–.94 across time). The measure instructs the caregiver to, “Please tell me if your [RELATIVE] thinks it is very important, somewhat important, or not at all important. How important is it for your [RELATIVE] to…” Sample items include: “have personal privacy”, “come and go as one pleases”, “avoid being a physical burden on one’s family”, “be with family or friends”, or “feel safe in one’s own home”. Higher scores indicate greater total perceived importance of the IWD’s values and preferences in daily care.

Covariates.—We considered factors that could affect caregivers’ perspectives within a caregiving relationship, including the following (see Table 1 for means and ranges):

Demographics. Caregivers reported their age, gender, race, education, kin relationship, employment status, coresidence, length of caregiving, and annual income at baseline. Race was coded 1 for African American and 0 for others. Education ranged from 1 (less than high school) to 6 (postgraduate degree). Kin relationship between an IWD and caregiver was coded 1 for spouse and 0 for child. Length of caregiving was the number of months caregivers had provided care. Annual income was rated from 1 (less than $8,000) to 8 ($60,000 or more). As in previous work (Reamy et al., 2011), gender, age, employment status, and coresidence were excluded from analyses, as they were all highly confounded with kin relationship (e.g., male caregivers were predominantly spouses). Income was also excluded due to non-random missing data.

IWD characteristics

IWDs’ baseline cognitive functioning. To assess global cognitive functioning, IWDs completed the MMSE at baseline (Folstein et al., 1975).

IWDs’ baseline values and preferences. Parallel to caregivers, IWDs completed the VPS measure (Whitlatch et al., 2005), rating their values and preferences from 1 (not at all important) to 3 (very important) as they plan for their care now and in the future at baseline. At subsequent times of measurement, a subset of IWDs completed a slightly different version in which their ratings were dichotomized to not important/important; therefore, due to measurement incongruence, only baseline reports of IWDs were used as a covariate. Higher scores indicate greater total importance of values and preferences in daily care (α = .86).

Caregiver’s baseline beliefs about care

Best interests. Seven items administered at baseline measured the importance for caregivers to consider IWDs’ best interests in care (Whitlatch & Feinberg, 2003). Responses were coded on a scale of 1 (not at all important) to 3 (very important). A mean item response was calculated (α = .49). Higher scores indicate greater importance for considering the IWD’s best interests in care decisions at baseline.

Decision-making involvement of the IWD. Caregivers completed the 16-item Decision-Making Involvement Scale, which assessed baseline involvement of the IWD in decision making around daily activities (Menne, Tucke, Whitlatch, & Feinberg, 2008; Menne & Whitlatch, 2007). Example items include decisions about visiting with friends, when to go to bed, and what to do in spare time. Responses were coded 0 (not involved at all) to 4 (very involved). Mean item scores were calculated (α = .92). Higher scores indicate greater perceived involvement of the IWD in making decisions in care at baseline.

Baseline dyadic strain. Caregivers responded at baseline to an 11-item Dyadic Relationship Scale (Bass, Tausig, & Noelker, 1989; Sebern & Whitlatch, 2007) regarding their relationship with their IWD. Items were scored 0 (strongly disagree) to 3 (strongly agree). A mean item score was calculated with higher scores indicating greater perceived strain at baseline (α = .70).

Baseline discussion of IWDs’ wishes Caregivers indicated the extent of discussion they had with IWDs about wishes concerning daily care with a single item at baseline. Responses ranged from 1 (have never discussed) to 4 (have talked about it a lot).

Quality of life over time. Caregivers completed the 14-item Quality of Life-AD Scale (Logsdon, Gibbons, McCurry, & Teri, 1999) at each measurement point. Example items include “How do you feel about your physical health?”, “How do you feel about your energy level?”, and “How would you describe your current relationship with your friends?” Responses were coded 1 (poor) to 4 (excellent). Mean item scores were calculated (α = .86–.90 across time). Higher scores indicate greater self-perceived quality of life of the caregiver.

Data Analysis

To analyze the change in caregiver perceptions over time of IWDs’ values and preferences, we utilized growth curve analysis techniques to fit a series of multilevel models for change to the raw longitudinal data for all participants. In contrast to approaches such as repeated-measures analysis of variance, growth curve analysis is helpful for analyzing change using longitudinal data, as it makes use of all available data from an individual, corrects for unreliability of measurement, and, most importantly, emphasizes individual trajectories rather than average values at each occasion (Raudenbush & Bryk, 2002; Singer & Willett, 2003).

As a first step, we set the baseline point in these data to be 0 months, to indicate change in time over months of enrollment in the study. We then plotted caregivers’ perceptions of IWDs’ values and preferences over time (Figure 1a). Utilizing SAS PROC MIXED, we calculated an unconditional means, no-growth model with no predictors of caregivers’ perceptions of values and preferences over time. This model allowed us to determine how much variance in the caregivers’ reports of IWDs’ values and preferences could be attributed to between-person level and within-person level variance:

Figure 1.

(a) A plot of caregivers’ report of perceived overall importance of individuals’ with dementia (IWDs’) values and preferences in daily care by months since study enrollment for all participants (N = 198), where month 0 is the initial point of assessment. This plot illustrates the pattern of caregiver report across time. (b) The predicted multilevel model of linear change of caregivers’ perceived importance of IWDs’ values and preferences in daily care across months of enrollment where month 0 is the initial point of assessment, demonstrating a slightly significant decreasing trend of 0.23 units per month.

(Model 0)

where Importance_ti is caregiver i’s expected score of total importance attributed to the IWD’s values and preferences in daily care at month t (where month 0 is baseline assessment), β_0i is caregiver i’s intercept parameter of the predicted perceived importance at baseline, e _ti are unobserved caregiver specific residual errors, and γ₀₀ is the expected score of the caregiver’s perceived importance when both e _ti and u _0i (the standard deviation of the estimated intercepts) are zero. From these estimates, we calculated the intraclass correlation coefficient (ICC) for between-persons (ICC = .53). This ICC indicates that the variance in caregivers’ perception of importance consisted of 53% of between-person variability and 47% of within-person variability.

To address Hypothesis 1, that there would be (a) change over time in caregivers’ reports of IWDs’ values and preferences and (b) significant between-person variation over time in level and slope, we estimated a linear growth curve model of change entering time into the equation at Level 1 (within-person level):

(Model 1)

where β_1i is the slope parameter for caregiver i’s change in the predicted perceived importance for a one unit change in month, and at Level 2 (between-person level) γ₀₀ and γ₁₀ are the group means for the intercept and slope and u _0i and u _1i are individual deviations from those group means. We also tested a quadratic time term; however, it was not significant (p = .58) and, therefore, was excluded.

In order to investigate the relationship between caregivers’ quality of life over time with their change in perceived importance of IWDs’ values and preferences over time (Hypothesis 2), we split quality of life into the within-person and between-person variance components to account for mean levels and variation of level in caregivers’ quality of life over time separately. We then entered the within-person variance (change in quality of life over time) into the Level 1 model:

(Model 2)

where β_2i is the slope parameter for caregiver i’s change in the predicted perceived importance for a one unit change in quality of life. We did not allow random variation for the β_2i slope, as the random effect was not significant.

Next, we added the mean level of caregivers’ quality of life, demographic variables (caregivers’ race, education, and kin relationship), caregivers’ beliefs about care at baseline (decision-making involvement of IWD and best interests), caregivers’ ratings on dyadic strain, caregivers’ reports of prior discussion with IWDs, length of caregiving at baseline, IWDs’ baseline reports of importance for their values and preferences, MMSE scores, and the interaction of each variable with time to account for between-person variability in level (β_0i) and slope (β_1i).

(Model 3)

where W_qj are between-person predictors of the effects of β_0i and β_1i, and γ_pqj are the corresponding coefficients. We present the trimmed, parsimonious model that only includes significant relationships.

Results

The first hypothesis predicted (a) significant change over time in caregivers’ reports of importance of IWDs’ values and preferences in daily care and (b) significant variation over time in level and slope. The linear multilevel model with only time (months) as a predictor indicates significant change over time in caregivers’ perceived importance of IWDs’ values and preferences and significant variation in slope and level (Table 3, Model 1). Time had a significant negative effect on caregivers’ perceived importance; for each 1-month increase in time there is a 0.23-point decrease in caregivers’ reports of importance of IWDs’ values and preferences, which translates into a drop of 2.76 points over 1 year. Figure 1b illustrates the predicted multilevel model of linear change. In addition, there is a significant amount of between-person variance in the intercept and a small but significant amount of between-person variance in the slope, despite the model’s efforts to estimate it close to zero. This finding indicates that individual growth patterns have a similar slope even though their initial starting values are different. Overall, in regard to Hypothesis 1, we see a significant downward slope in report of average importance of IWDs’ values and preferences by caregivers over time and significant variation in caregivers’ reports of IWDs’ values and preferences in both level of report and change over time.

Table 3.

Factors Associated With Caregiver Reports of IWDs’ Values and Preferences in Daily Care Over Time

	Model 1		Model 2		Model 3
	B	SE	B	SE	B	SE
Fixed effects
Intercept (initial status), γ00	89.47***	0.89	89.29***	0.89	81.11***	5.19
Time (change over months), γ10	−0.23***	0.04	−0.22***	0.04	−0.22***	0.04
QoL_resid (within-person), γ20	–	–	2.70	2.05	4.04*	1.98
Predictors of the intercept
QoL_mean (between-person)	–	–	–	–	3.38†	1.79
Spouse	–	–	–	–	−3.40*	1.54
Baseline decision-making involvement of IWDs	–	–	–	–	7.54***	1.04
Baseline best interests	–	–	–	–	10.00**	3.73
Baseline IWD self report of Importance of Values and Preferences	–	–	–	–	0.22**	0.08
Random effects
Variance intercept,	93.66***	16.57	92.18***	16.47	41.38***	11.75
Variance in change over time,	0.05**	0.02	0.06**	0.02	0.06**	0.02
Covariance intercept, change over time,	1.18**	0.45	1.16**	0.45	1.26***	0.38
Residual variance,	76.70***	7.68	74.84***	7.60	73.10***	7.38
−2LL		3922.7		3873.4		3700.1
AIC		3930.7		3881.4		3708.1

Notes: N = 198 caregivers. Model based on up to five occasions of measurement nested within 198 participants. AIC = Akaike Information Criterion; IWDs = individuals with dementia; −2LL = −2 log likelihood, relative model fit statistics.

*p < .05. **p < .01. ***p < .001. ^† p = .05.

In addressing Hypothesis 2, that caregivers’ quality of life over time would significantly covary with caregivers’ reports of perceived importance of IWDs’ values and preferences, we see that time remains a significant predictor of caregivers’ perceived importance of IWDs’ values and preferences in daily care when the model includes the within-person variance (split from the between-person variance) of quality of life over time (Table 3, Model 2). We then added demographic variables, baseline-level independent variables, and quality of life as a between-person effect as well as their interactions with time, and trimmed the model to be parsimonious. In Model 3, we see a significant independent effect of time and change in quality of life over time, after accounting for significant predictors of the intercept. For each one-unit increase in variability around a caregiver’s mean quality of life, there is a 4.04 unit increase in caregiver reports of perceived importance of IWDs’ values and preferences. No other demographic or baseline characteristics significantly interacted with time, and such interactions were removed from the model (Table 3). We do see, however, additional effects on the intercept. Higher levels of baseline decision-making involvement are related to higher levels of initial report of perceived importance. Spousal caregivers report lower levels of initial importance of values and preferences of the IWD than adult child caregivers. Higher baseline reports of caregivers taking into account IWDs’ best interests are related to higher initial reports of perceived importance. Finally, higher self-reported importance of values and preferences by the IWD is related to higher caregiver reports of importance at baseline. No other variables were significant (MMSE, race, length of caregiving, extent of previous conversation, or dyadic strain). Thus, the tests of Hypothesis 2 show that changes in caregivers’ reports on IWDs’ values and preferences are covarying with caregivers’ self-reports of their own quality of life over time even after accounting for the significant effects of caregivers’ perceptions of IWDs’ involvement in care, caregivers’ beliefs about responding to the IWDs’ best interests, IWDs’ self-reports of importance, and type of kin relationship on the level of importance.

Discussion

The results of this study demonstrate that caregivers’ perceptions of total importance of IWDs’ values and preferences in daily care change with time. Regarding the first hypothesis, caregivers’ perceptions of importance of IWDs’ values and preferences decreased over 4 years. Regarding the second hypothesis, caregivers’ perceptions of importance of the IWDs’ values and preferences in care covaried with caregivers’ own quality of life over time.

These findings carry several important implications for research and practice. First, time remains a significant explanatory variable in accounting for change in caregiver reports of IWDs’ values and preferences independent of the effects of caregivers’ self-reported quality of life and other variables examined. This result may indicate that the passing of time itself, or variables that time acts as a proxy for (i.e., increasing illness severity, systematic changes in a caregiver’s role), may influence caregivers’ appraisals of importance. Caregivers may increasingly believe that their relatives’ identities are changing and, as a result, see their relatives’ values and preferences as less important. Alternately, change in perceptions could reflect an adaptive process for caregivers; as IWDs are nearing the end of life, caregivers experience a form of anticipatory grief by attributing less importance to daily care values and preferences. Such a shift in perspective may be a coping mechanism or may allow caregivers to make more expedient decisions in care, rather than decisions IWDs would make for themselves. Although baseline MMSE was not a predictor of change, decline in cognition over time, and other processes not measured in the study, such as increased behavior problems or health complications, may account for the effect of time. However, overall, the average amount of change of 2.76 points per year in caregivers’ ratings of importance is relatively modest, given a 74-point maximum range in the scale. Further evidence is required to understand the true clinical significance of this drop.

Second, we found an independent effect of within-person changes in caregivers’ quality of life over time on caregivers’ perceptions of importance of IWDs’ values and preferences over time. To illustrate this effect further, we modeled at the individual level the covariation between caregivers’ quality of life and caregivers’ reports of importance of IWDs’ values and preferences over time (Figure 2). As shown in the examples in Figure 2, when caregivers’ quality of life decreases they perceive their relatives as placing less importance on their values and preferences. This finding supports theoretical work in depression and negative affect, whereby one’s negative disposition extends beyond self-image to one’s view of others (Beck et al., 1979). As caregivers become more negative about their own lives, they also make negative attributions to their IWDs’ lives. Caregivers with a lower quality of life who perceive their relatives’ values and preferences as less important may be more likely to discount their relatives’ needs or show greater discrepancy in understanding their relatives’ preferences (Reamy et al., 2011; Whitlatch & Feinberg, 2003). This finding implies that interventions that improve caregivers’ reported quality of life (Belle et al., 2006) may also improve caregivers’ recognition of IWDs’ values and preferences in care.

Figure 2.

Images (a) and (b) represent two individual-level examples of the covariation between caregivers’ self-reported person-mean centered quality of life over time and caregivers’ reports of importance of IWDs’ values and preferences in daily care by months since study enrollment for all participants (N = 198), where month 0 is the initial point of assessment.

Third, although no other predictors significantly interacted with time, baseline scores on a few predictors were independently associated with initial level of caregivers’ reports of importance of IWDs’ values and preferences. Consistent with our previous research (Reamy et al., 2011), caregiver beliefs of how involved the IWD is in decision making around his or her care (Baseline Decision Involvement) has an effect on the reported initial importance of IWDs’ values and preferences. Additional factors that were also predictive of the initial level of importance of values and preferences included kin relationship (spouse or child caregiver), IWDs’ self-reports of importance, and caregivers’ perceptions of taking into account IWDs’ best interests in care. Interestingly, spouses reported lower importance of values and preferences than did adult offspring. It may be that adult offspring make fewer assumptions of values and preferences compared with spouses and ask their parents about their preferences. Surprisingly, however, the extent of discussion caregivers reported having about IWDs’ wishes was not a significant predictor.

Fourth, we see an overall effect of caregivers’ perceptions in care and in other life domains affecting how they perceive the importance of IWDs’ values and preferences in care. Evidence suggests that individuals with mild-to-moderate dementia, despite their ability to participate in decision making (Feinberg & Whitlatch, 2001), are often not included in their own care decisions. The findings from this study suggest that one explanation for this occurrence may be caregivers’ own quality of life clouding their judgment or receptivity. There are many possible barriers that prevent surrogates from being as accurate in articulating elders’ preferences in care as older adults. The decisions are difficult for both surrogates and elders, older adults may be reluctant to discuss impending death or advance care planning, some elders may have never thought about their preferences and always deferred to others in decision making, or elders may only share general goals for care that do not translate to specific preferences (Black et al., 2009; Givens, Kiely, Carey, & Mitchell, 2009). Within any of these possible barriers, caregivers’ own negative perceptions may further augment misunderstanding, leading to decreased openness to IWDs’ preferences.

Based on these results, we cannot rule out the possibility that IWDs’ perceptions of importance of their own values and preferences in daily care are also declining. Caregivers’ perceptions could be accurate, or may accurately infer what IWDs find important in their daily life over time. At baseline, IWDs’ self-reported importance of their values and preferences was significantly related to caregivers’ reports of importance. However, because some IWDs responded on a different scale over time, congruent data are unavailable to address this hypothesis directly. Yet, the interaction of IWDs’ baseline reports did not significantly interact with time, and evidence from other studies suggest relative stability in IWDs’ self-reports of their values and preferences even as cognition declines (Feinberg & Whitlatch, 2001; Piiparinen, Tucke, & Whitlatch, 2008) and other preferences in care (Ditto et al., 2003).

In conclusion, these findings suggest that consistent with dyadic research (Berg & Upchurch, 2007), there is a need to examine characteristics of both members of the dyad when considering honoring the values and preferences of IWDs in daily care. In particular, caregivers’ well-being may affect caregivers’ perceptions of IWDs’ values and preferences. Second, these findings highlight the complexity of understanding caregivers’ perceptions. There is still significant variance to be explained in both the intercept (level) and the slope (change over time). There may be additional characteristics of IWDs and changing processes within IWDs that trigger changes in perceptions of caregivers. Thus, we may need more detailed observation of IWDs and caregivers over time to determine the major influences on caregivers’ views of IWDs’ values. By understanding this process, we can consider strategies that will help caregivers make surrogate decisions that best reflect what IWDs would want.

The results discussed here are not without limitations. First, while the sample was strengthened by its racial diversity, we used a sample of convenience; findings may not be generalizable to all IWDs and caregivers. Furthermore, due to non-random missing data those cases that were followed over time represent a unique subset of caregivers still providing care at home, not those who placed their relative in institutional care or whose relative died during the observation period. Third, findings may not extend to individuals with more severe dementia who may be less able to indicate clearly their preferences about care (MMSE M = 18.97, SD = 7.46, 3 years out and M = 19.50, SD = 7.33, 4 years out); there may be a different effect on caregivers’ perceptions (Elliott et al., 2009). Fourth, with the exception of caregiver reported quality of life, we only examined baseline level characteristics associated with caregivers’ perceptions of IWDs’ values and preferences over time. Other factors not considered in these analyses (e.g., physical disability) or factors of the care recipient that also change over time might be more salient indicators of changing perceptions. Moreover, Best Interests, a scale that captures caregivers’ strong intentions to act in the best interests of their relatives may be prone to effects of social desirability and does not capture much variance among reporters. This scale may need further measurement development to produce a more reliable estimate. Fifth, despite the longitudinal nature of the study, conclusions about directionality of quality of life and caregiver perceptions of IWDs’ values and preferences remain unclear; we can only conclude that there is a time covarying relationship. Furthermore, modeling the data with a different time metric, such as time since diagnosis rather than time since enrollment, may also reveal a different perspective of these processes not captured here. Seventh, we assumed measurement consistency over time; a closer look at the factor structure of these data over time is necessary to conclude that the scales maintain the same meaning. Finally, this study focused only on caregivers’ perspective of IWDs’ values and preferences over time, accounting for IWDs’ baseline perspectives. More could be learned by obtaining congruent measurements of IWDs’ self-report of values and preferences and caregivers’ reports over time and/or examining discrepancies between IWDs’ and caregivers’ perceptions over time.

Despite such limitations, the results presented here demonstrate the significant influences of caregivers’ perceptions in care and in other life domains on their interpretation of the importance of IWDs’ values and preferences in daily care at baseline and over time. This study is one of few to have followed caregivers of IWDs over time, and these data ultimately offer a unique perspective on how caregivers’ perceptions change up to 4 years out. A next research step is to examine these changing perspectives in the context of the dyad, to see how IWDs’ self-reported preferences change over time and whether changing perceptions translate to care provision itself. A practical step is continuing to develop dyadic-based interventions that address both caregivers’ perceptions and care recipients’ needs when caring for IWDs to ultimately improve person-centered, quality care for IWDs.