Optimizing Quality of Life with Cognitive Impairment: A Study of End-of-Life Care in Assisted Living

Abstract

Objectives:

Assisted living (AL) is an expanding site of end-of-life (EOL) care in the US. Understanding determinants of quality of life (QoL) for AL residents near EOL is vital to optimize care for this growing population, most of whom have some degree of cognitive impairment (CI). This analysis aimed to identify factors associated with QoL in a diverse sample of AL residents with CI approaching EOL.

Design:

Observational cross-section design.

Setting and participants:

Data are from a 5-year study funded by the National Institute on Aging examining EOL care of residents in 7 diverse AL communities in metropolitan Atlanta (n=67).

Methods:

CI was assessed with the Montreal Cognitive Assessment (MoCA) (scores ≤ 26) and QoL was determined with the self-reported QoL in Alzheimer’s disease (QoL-AD) survey adapted for use in AL. Psychological distress was assessed using the PHQ-4 and fatigue was assessed using the 13-item FACIT-Fatigue scale questionnaire. Initial descriptive analyses were followed by backward stepwise regression analyses to select a best-fitting model of QoL.

Results:

The final model predicted 27% of the variance in QoL. CI was not significantly correlated with QoL and was not retained in the final model. Pain and functional limitation also did not meet inclusion criteria (p ≤ 0.10) and were sequentially removed, producing a final model of QoL in terms of psychological distress (β=−.28, p=.032), fatigue (β=−.26, p=.048), and race (β=.21, p=.063).

Conclusions and Implications:

The lack of a significant correlation between degree of CI and self-reported QoL suggests that AL residents have the potential to experience high QoL, despite CI. Interventions to reduce psychological distress and manage fatigue could be implemented during EOL care to attempt to improve QoL for AL residents with CI. The correlation between race and QoL warrants investigation into possible racial disparities in AL and EOL care.

Brief Summary:

This study identifies modifiable factors associated with quality of life of assisted living residents with cognitive impairment at end of life and highlights areas for intervention and future study.

INTRODUCTION

Providing quality end-of-life (EOL) care to an aging population is one of the greatest challenges facing the American healthcare system. According to Census projections, 20% of the United States population will be older than 65 years by 2050.⁽¹⁾ An estimated 27 million of those Americans will require long-term care, much of which will be provided in assisted living (AL) communities.^{(1, 2)} Compared to nursing homes, AL has historically been guided by a social rather than medical care model, and does not provide 24-hour skilled nursing care. Nevertheless, AL residents have rising levels of medical comorbidity and frailty, and are increasingly aging and dying in place in AL.^{(2, 3)} Approximately 40% of AL residents have dementia, and as many as 90% have some degree of cognitive impairment (CI).^(2-4) Therefore, understanding experiences of AL residents with CI is important to guide best practices for EOL care.

EOL care has traditionally been evaluated by objective clinical outcomes. However, it is increasingly recognized that patients’ perceived quality of life (QoL) also is important.^(5-7) Numerous studies show inconsistencies between self-reported and caregiver-reported QoL, underscoring the importance of considering patient perspectives.^(8-11) Several studies show that QoL does not correlate with a person’s degree of CI.^(12-15) This finding suggests potential for positive EOL experiences despite CI, and further research is needed to understand how AL can facilitate optimal QoL for residents with CI at EOL. Factors beyond cognition shown to be associated with QoL in populations with CI include psychological distress and functional ability.^{(8, 9, 12-17)} Fatigue⁽¹⁸⁾ and pain⁽¹⁹⁾ are prevalent at EOL and also likely influence QoL in the context of CI and AL.

While prior research identifies factors that may influence QoL among people with CI, most studies have been conducted outside the United States and not in AL settings. This study investigates QoL among residents approaching EOL in a diverse sample of AL communities, and aims to expand knowledge of QoL with CI to the growing American population approaching EOL in AL. We hypothesize that QoL is not associated with CI but is associated with functional limitation, psychological distress, pain, and fatigue.

METHODS

Study Sample

This analysis included 67 residents with CI recruited from 7 AL communities in the metropolitan Atlanta area between November 2015 and September 2018 and included between 2 and 23 residents per community. We selected communities through maximum variation sampling to reflect national variability in factors such as size, location, ownership (corporate versus private), availability of specialized dementia care, and socioeconomic status of residents and staff.⁽²⁰⁾ Within each community, a demographically diverse sample of residents was selected through purposive maximum variation sampling. Those selected needed to meet at least one of the following criteria: be at least 85 years old, have multiple chronic medical conditions, be diagnosed with a life-limiting illness, and/or be enrolled in hospice. Consistent with earlier research, we found that hospice is underutilized in AL,^{(2, 21)} only 7 residents (10%) were on hospice at the time of data collection.

Enrolled participants completed a series of in-depth interviews and health assessments. These included validated measures of CI, QoL, functional limitation, psychological distress, pain, and fatigue. We obtained residents’ medical diagnoses by reviewing AL health records as well as Emory Healthcare electronic medical records when available.

We obtained IRB approval (Emory University #IRB00075456) for direct interaction with residents and review of medical records, including electronic medical records when available. Informed consent was obtained from each participant with capacity to consent prior to enrollment. Researchers contacted a legally authorized representative for consent for residents who expressed interest in participating, but lacked capacity to consent.

Measures

Outcome Variable

QoL was assessed with a version of the Quality of Life in Alzheimer’s Disease questionnaire (QoL-AD)⁽⁶⁾ adapted for use in AL (Cronbach’s α =.88).^{(7, 8)} It is a 15-item 4-point scale (1= poor to 4=excellent), with higher scores indicating higher QoL. Items address perceptions regarding: physical health; energy; mood; living situation; memory; relationships with AL staff, friends, family, and self; ability to keep busy, do things for fun, take care of one’s self, have choices in life, and live with others; and life overall. The original version of the QoL-AD has been validated among individuals with mild, moderate, and severe CI.⁽¹⁵⁾ The adapted version of the QoL-AD has been validated in AL residents with mild to moderate CI.^{(7, 8)}

Hypothesized Predictor Variables

CI was assessed with the Montreal Cognitive Assessment (MoCA),⁽²²⁾ which is scored out of 30 points, with higher scores reflecting higher cognitive function, and CI defined as ≤ 26 points.

Functional limitation was assessed based on the Older Americans Resources and Services (OARS) Multidimensional Assessment adapted for AL. It evaluates 7 activities of daily living (bathing, eating, dressing, grooming, walking, transferring, and getting to the bathroom on time) and 6 instrumental activities of daily living (using the telephone, getting to places outside AL, shopping, taking medication, managing money, and getting to meals) (Cronbach’s α =.82). Responses were scored on a 3-point scale (0= requires no assistance to 2 = requires total assistance), with higher scores indicating more functional limitation. Care providers corroborated OARS responses.

Fatigue was assessed with the Functional Assessment of Chronic Illness Therapy Fatigue Scale (FACIT), a 13-item questionnaire about fatigue during daily activities in the past week (Cronbach’s α =.74).⁽¹⁸⁾ FACIT scores were reverse-coded so that higher scores indicate more fatigue.

Psychological distress was assessed with the Patient Health Questinaire-4 (PHQ-4), a brief questionnaire on depression and anxiety symptoms (Cronbach’s α =.82).⁽²³⁾ Higher PHQ-4 scores indicate more psychological distress.

Pain was assessed with the Iowa Pain Thermometer, for which residents rate their physical pain in the past week on a 12-point visual scale.⁽²⁴⁾

Additional Relevant Sociodemographic and Health Characteristics

Sociodemographic variables collected included age (in years), gender, race, and education. Education was measured as a ten-level proxy variable for continuous years of education (0 = less than a high school degree to 9 = professional degree beyond a master’s degree).

As a standardized metric of number of chronic medical conditions, residents’ medical records were reviewed for the ten most common chronic conditions identified in AL. These include: hypertension, diagnosed dementia, heart disease, depression, arthritis, osteoporosis, diabetes, chronic obstructive pulmonary disease, cancer, and stroke.⁽²⁵⁾

Statistical Analysis

Study variables were characterized using descriptive statistics, and all regression assumptions were assessed and met. We computed Pearson correlations and used the variance inflation factor (VIF) to rule out multicollinearity. The mean VIF was 1.35, with a range from1.08 to 1.66. To assess how much of the variance in QoL might be accounted for by membership in a specific community, we calculated the intraclass correlation coefficient (ICC). Given an ICC equal to 0, we determined that we did not need to adjust for clustering during multivariable analysis.

Multivariable linear regression analysis was then conducted to construct a best-fitting model for QoL using MoCA scores (a key independent variable of interest that was not significant in bivariate analysis) and the variables found to be associated with QoL in bivariate analysis. Given the small sample size, backward stepwise multivariable regression was used to explore hypothesized associations and select the most parsimonious model. We used SPSS, version 24 (IBM Corp., Armonk, NY) for all statistical analyses.

RESULTS

Resident Characteristics and Variables of Interest

Of the 86 residents enrolled in this study, we included the 67 residents with CI (MoCA ≤ 26) who completed assessments in this analysis. Reasons for failure to complete the assessments included: moving out of the community (n=3), refusal (n=3), death (n=4), and advanced cognitive impairment (n=9). Residents’ sociodemographic and health characteristics are presented in Table 1.

Table 1.

Descriptive statistics for the study variables (n=67).

Variables
Sociodemographic Characteristics	n (%)	Min-Max
Female	43 (64%)
Race
White	37 (55%)
Non-white*	30 (45%)
Mean Age (SD)	85.7 (8.4)	58-103
Mean level of education†,‡ (SD)	5.0 (2.6)	0-9
Level of Cognitive Impairment§
Mild cognitive impairment	27 (40%)
Moderate cognitive impairment	26 (39%)
Severe cognitive impairment	14 (21%)
Health Characteristics	Mean (SD)	Range
Cognitive impairment§	15.5 (6.3)	2.0-26.0
Functional limitation∥, **	7.5 (4.9)	0-20.6
Psychological distress††	2.5 (3.0)	0-12.0
Pain‡‡, ‡	3.3 (2.8)	0-9
Fatigue§§	21.1 (10.0)	4.3-44.0
Chronic medical conditions∥ ∥, ‡	3.3 (1.7)	0-8
Outcome Variable
Quality of life***	43.5 (7.2)	28.0-58.6

^*Non-white residents identified their race as “black” (n=28) or “other” (n=2).

^†Education was a proxy for continuous years of education was quantified on a scale from 0 to 9 and; 0=less than high school diploma; 1=GED; 2=High school diploma (not GED); 3=Post-high school technical training certificate; 4=Some college or 2-year associate degree; 5=Three or more years of college but no Bachelor’s degree; 6=Bachelor’s degree; 7=One or more years of graduate training but no graduate degree; 8=Master’s degree; 9=Professional degree beyond a Master’s degree.

^‡One participant had a missing value (n=66).

^§Cognitive impairment (CI) was assessed with the Montreal Cognitive Assessment (MoCA), scored from 0 to 30. MoCA scores 18-26 are defined as mild CI; 10-17 are moderate CI; <10 are severe CI.

^∥Functional limitation was assessed with the OARS questionnaire; higher scores indicate greater functional limitation, with a maximum of 26 points.

^**The question regarding the ability to do laundry was dropped because all ALs in this study provided full laundry service to residents.

^††Psychological distress was assessed with the PHQ-4; higher scores indicate greater psychological distress, with of a maximum of 12 points.

^‡‡Pain was assessed with the Iowa Pain Thermometer; higher scores indicate more pain, with a maximum of 12 points.

^§§Fatigue was assessed with the FACIT; higher scores indicating more fatigue, with a maximum of 52 points.

^{∥ ∥}Chronic medical conditions were quantified on a scale from 0 to 10, considering the 10 most common chronic conditions in AL.

^***Quality of life was assessed with the QoL-AD; higher scores indicate higher quality of life, with a maximum of 60 points.

Preliminary Bivariate Analysis

Results of bivariate analysis are presented in Table 2. As hypothesized, there was no significant association between QoL and CI (r=.065, p=.60), but QoL was negatively correlated with psychological distress (r=−.43, p<.001), fatigue (r=−.40, p=.001), and functional limitation to significant degrees (r=−.33, p=.05). It tended to negatively correlate with pain (r=−.21, p=.091) and positively correlate with white race (r=+.22, p=.077). To further explore the relationship between race and QoL, a t-test was conducted to compare the mean QoL score for white residents with CI (44.9, n=37) versus non-white residents with CI (41.8, n=30). The mean difference of −3.14 did not meet 2-tailed statistical significance (p=.076; Cohen’s d = 0.45). Although the correlation and mean difference were not statistically significant, they were deemed notable enough to include race in the multivariable regression analysis.

Table 2.

Bivariate correlations between Quality of Life and predictor variables.

		Cognitive Impairment*	Functional Limitation†	Psychological Distress‡	Pain§	Fatigue∥	White Race
QoL**	Pearson Correlation	.065	−.239	−.430***	−.210	−.396***	.218
	Sig. (2-tailed)	.604	.051	.000	.091	.001	.076
	N	67	67	67	66	67	67

^***Correlation is significant at the 0.01 level (2-tailed).

^*Cognitive impairment (CI) was assessed with the Montreal Cognitive Assessment (MoCA), scored from 0 to 30.

^†Functional limitation was assessed with the OARS questionnaire; higher scores indicate greater functional limitation, with a maximum of 26 points.

^‡Psychological distress was assessed with the PHQ-4; higher scores indicate greater psychological distress, with of a maximum of 12 points.

^§Pain was assessed with the Iowa Pain Thermometer; higher scores indicate more pain, with a maximum of 12 points.

^∥Fatigue was assessed with the FACIT; higher scores indicating more fatigue, with a maximum of 52 points.

^**Quality of life was assessed with the QoL-AD; higher scores indicate higher quality of life, with a maximum of 60 points.

Multivariable Regression Analysis

Backward stepwise regression analyses were conducted to select a best-fitting model of QoL (Table 3). Based on findings from preliminary bivariate analysis and aims of this study, the initial set of predictor variables entered into backward stepwise regression analysis included level of CI, functional limitation, psychological distress, pain, fatigue, and race. Pain, CI, and functional limitation did not meet inclusion criteria (p ≤ 0.10) and were sequentially removed, producing a final model of QoL in terms of psychological distress (β=−.28, p=.032), fatigue (β=−.26, p=.048), and race (β=.21, p=.063). Variables retained in the final model accounted for 27% of the variation in QoL.

Table 3.

Multivariable model for QoL. Potential predictor variables entered in backward stepwise linear regression analysis were CI, functional limitation, psychological distress, pain, fatigue, and white race. The final model accounts for 27% of the variation in QoL in terms of Psychological Distress, Fatigue, and white race. CI was assessed with the MoCA, scored from 0 to 30. Functional limitation was assessed with the Older Americans Resources and Services Multidimensional Assessment; higher scores indicate greater functional limitation, with a maximum of 26 points. Psychological distress was assessed with the Patient Health Questionnaire-4; higher scores indicate greater psychological distress, with of a maximum of 12 points. Pain was assessed with the Iowa Pain Thermometer; higher scores indicate more pain, with a maximum of 12 points. Fatigue was assessed with the Functional Assessment of Chronic Illness Therapy Fatigue Scale; higher scores indicate more fatigue, with a maximum of 52 points.

Predictor Variables:
	Unstandardized Coefficients		Standardized Coefficients
	B	Std. Error	Beta	t	Significance
Psychological Distress	−.671	.305	−.279	−2.200	.032
Fatigue	−.186	.092	−.255	−2.020	.048
White Race	3.009	1.590	.208	1.892	.063

Model Characteristics:
R	R Squared	Adjusted R Squared	Standard Error of the Estimate
.517d	.268	.232	6.351

DISCUSSION

The high incidence of CI in this diverse sample of AL residents at EOL underscores the importance of understanding factors that shape QoL for AL residents with CI approaching EOL. As hypothesized based on the literature, level of CI was not significantly correlated with QoL. These findings support our assertion that AL residents with CI have potential for positive QoL at EOL.

Psychological distress was most strongly negatively associated with QoL in both bivariate and multivariable analysis. This observation is consistent with established associations between QoL and mental health in people with CI.^{(9, 13-17)} The slight association found between degree of CI and QoL in bivariate analysis might be attributed to residents experiencing psychological distress related to their cognitive limitations.

The significant negative association between fatigue and QoL reinforces the well-established role of fatigue at EOL,^{(26, 27)} and extends it to the population of AL residents with CI at EOL. Fatigue may be easily overlooked in the AL environment, where napping and inactivity are normalized. Yet the negative correlation between fatigue and QoL suggests it is not benign.

The positive association between QoL and race was an unexpected finding, which was retained in the final multivariable model, though it did not reach statistical significance. This study is notable for its racially diverse sample, which was 45% non-white (n=30). Research in AL generally includes primarily white samples, with some exceptions. ^{(28, 29)} This project benefited from a diverse sample of AL residents; however, the small sample size limited its capacity for complex analyses. The nested structure of these data can impact the standard errors, although preliminary analyses indicated that clustering did not need to be accounted for in estimation. The difference in self-reported QoL according to race emphasizes the need for further research into potential racial disparities in EOL care in AL. Our qualitative data show that Non-White residents who scored low on QoL tended to have smaller social support networks and lack family support compared with other residents in the sample who scored higher on this measure. Some referred to their own history of caring for older family members at home, indicating that AL placement and lack of support from family, including some estranged family relationships, were at odds with their expectations for care at end of life.⁽³⁰⁾ Other possible underlying factors include differences related to healthcare access or quality or direct effects of racism. Although socioeconomic status may be a confounding variable in this relationship, there was only a slight non-significant correlation between QoL and highest level of education, which is a proxy indicator of socioeconomic status.⁽³¹⁾

Assessing QoL among people with CI is inherently limited by communication barriers in the late stages of dementia. Consequently, the perspectives of people with CI are underrepresented in research and healthcare delivery. Existing literature indicates that caregivers do not accurately estimate the QoL of people with CI.^(8-11) Therefore, the self-reported measures used in this study were vital to the objective of understanding residents’ experiences within CI. Although the version of the QoL-AD used in this study has been validated in populations with mild to moderate cognitive impairment,⁽¹⁵⁾ the Cronbach’s α values in this analysis, which included residents with MoCA scores from 2 to 26, provide support it may be reliable for residents with more severe CI in AL. Notably, our ongoing qualitative research shows that many residents with scores this low are able to verbally convey their thoughts, feelings, and needs despite having some communication deficits.⁽²⁰⁾

Conclusions and Implications

As a cross-sectional study, this analysis cannot determine causal relationships. Yet, this study illuminates factors that are closely related to QoL among AL residents with CI at EOL, which merit further investigation. Most notably, CI itself was not the most significant factor associated with QoL, contrary to what caregivers may assume. Instead, modifiable factors including psychological distress and fatigue were more closely correlated with QoL. Interventions to provide psychological support and better manage fatigue among AL residents could substantially improve EOL experiences. As the US population ages, the number of people experiencing CI and EOL in the AL setting is steeply rising. To provide quality care for this population, recognizing their perceived QoL and identifying strategies to improve it will be vital.