Surveying Family Caregivers to Assess Quality of End-of-Life Care in Medicare Advantage

Abstract

Objectives:

As Medicare Advantage (MA) enrollment increases, measures are needed to monitor end-of-life care quality across settings. We aimed to use MA data to identify respondents for a survey of end-of-life care and derive quality measures from survey responses.

Study Design:

Cross-sectional study.

Methods:

We developed a survey that assesses end-of-life care experiences across settings and field-tested it among family caregivers of Kaiser Permanente members who died from July to November 2021. We used factor analysis to assess survey item performance and calculate composite measures, examined validity by regressing overall ratings on composites, and described care experiences across settings and care patterns through multivariable regression.

Results:

The response rate was 30%; 33% of decedents had cancer, 49% died at home, 24% in a hospital, and 5% in a nursing home. Confirmatory factor analyses supported five composite quality measures, assessing timeliness, communication, consistency of care with patient preferences, respect, and symptom palliation. Cronbach’s alpha indicated adequate internal consistency reliability of composites (range: 0.72–0.82). Care experiences were better for those who died at home with hospice compared with those who died in a nursing home and for those who received care from one setting compared with those who transitioned across settings.

Conclusion:

It is possible to sample family caregivers for a survey of end-of-life care experiences using administrative data from an MA organization. Survey-based quality measures can be used to assess and compare the quality of end-of-life care in MA.

Key Message

As Medicare Advantage enrollment increases, measures are needed to monitor the quality of end-of-life care across settings. Quality measures derived from the CAHPS End-of-Life Care Survey allow for reliable assessment of key aspects of end-of-life care, including consistency of care with patient preferences and help for symptoms.

Introduction

Value-based reimbursement, which considers quality and cost of care together, has been on the rise across the U.S. health care system and is increasingly being used to pay for end-of-life care.¹ Quality measures derived from patient and family care experience surveys are a cornerstone of value-based initiatives. Integrating the voices of patients and caregivers into assessments of the quality of serious illness and end-of-life care is critical,² given the importance of goal-concordant care and maximizing quality of life at the end of life.³

Most existing care experience surveys used by accountability initiatives in the U.S. focus on specific care types or settings, such as hospice or hospitals;^4,5 as a result, they do not examine the entire episode of end-of-life care. In addition, setting-specific surveys fail to capture transitions across care settings, which can be particularly burdensome to patients,⁶ and may impose a burden on those who are sampled for multiple surveys.⁷

Surveys of end-of-life care in the United Kingdom and United States have characterized end-of-life care experiences for populations by using death certificates to identify next of kin for survey sampling,^8,9 or identifying family caregivers of Medicare beneficiaries previously sampled using claims data to assess care in the last month of life.¹⁰ We developed a sampling methodology, survey instrument, and set of quality measures that could be implemented in a standardized fashion by health systems, Medicare Advantage (MA) organizations, or accountable care organizations to evaluate end-of-life care quality for value-based initiatives.

To do so, we developed and field-tested a survey of end-of-life care experiences among caregivers of those who died while enrolled in a large health system and MA organization. The Consumer Assessment of Healthcare Providers and Systems (CAHPS®) End-of-Life Care Survey assesses experiences across a variety of different care settings (e.g., hospice, nursing home, acute care hospital, other care facility, or home). Our goals were to identify quality measures designed to examine end-of-life care across the entire care episode, assess the degree to which these measures are internally reliable and valid, and examine variation in measure performance across patient subgroups.

Methods

We conducted an observational, cross-sectional field test of a draft version of the CAHPS End-of-Life Care Survey among bereaved family caregivers of patients who had recently died. We used field test survey response data to describe psychometric properties of proposed quality measures derived from survey responses.

Survey sample

To identify a sample for the field test of the CAHPS End-of-Life Care Survey, we worked with Kaiser Permanente, a large integrated health care system and MA organization, to identify the names and contact information of informal caregivers of members who had recently died. Caregivers were primary health care decision makers or agents verbally appointed by members when they participated in life care planning discussions, named in an advance directive, or listed as emergency contacts in electronic medical records.

We randomly sampled informal caregivers of patients age 18 or older who died from July to November 2021 while receiving care from Kaiser Permanente in the Denver and Boulder Metropolitan service areas of Colorado or Southern California (n = 1700 of 9743 decedents, split evenly across the five months of sampling; Supplementary Fig. A1). One caregiver was sampled per decedent; in instances in which a decedent had more than one listed caregiver, the primary health care decision maker was prioritized over the alternate health care decision maker, and the alternate health care decision maker was prioritized over other types of appointed health care agents. If no health care agent was found, then the first emergency contact with address and phone was sampled.

To be survey-eligible, patients needed to have had at least two health care visits in the last year of life and an informal caregiver living in the U.S. for whom a mailing address or telephone number was available (92 percent of all adult decedents). Caregivers were eligible to complete the survey if they indicated that they “sometimes,” “usually,” or “always” oversaw or took part in the patient’s care; 12 caregivers were excluded due to lack of involvement in the patient’s care. Caregivers were considered ineligible if they indicated in response to the survey that the patient did not receive care from a health care provider in the last month of life or that the patient died of an accident or injury and did not have a chronic underlying condition (n = 38). Caregiver respondents who were deceased at the time of survey administration, were unable to complete the survey due to a mental or physical incapacity, did not read or speak the language in which the survey was being administered, or were representatives of an institution (e.g., funeral home) rather than a family member or friends were also considered ineligible (n = 31).

Survey instrument

The draft survey instrument included evaluative items assessing aspects of care deemed important by caregivers, health care providers, and quality improvement leaders interviewed and convened during survey development and identified by national consensus clinical practice guidelines for palliative care.¹¹ Topics included timeliness of care, communication with the health care team, consistency of care with patient preferences, treating patients with respect, symptom palliation, and emotional support. To identify candidate questions for inclusion in the survey, we conducted a systematic review of the peer-reviewed and gray literature and compiled a bank of survey questions. Wherever possible, items for the field test survey were derived or adapted from surveys for which there were published assessments of reliability and validity. To promote alignment with national initiatives, items from the CAHPS Hospice Survey or other CAHPS surveys were prioritized.¹² We also adapted survey items from a survey previously conducted by Kaiser Permanente to identify opportunities to improve end-of-life care.¹³ Draft survey items and questionnaires were refined based on three rounds of cognitive interviews with caregivers of family members or friends who had died in the prior year.

Data collection

Data were collected in July and August 2022, 8 to 12 months after the death of the patient, using a mixed-mode protocol (mail with telephone follow-up). A survey packet was mailed to sampled caregivers; Spanish-language materials were included in the survey packet if the decedent’s primary spoken or written language was Spanish. Three weeks after the survey packet mailing, follow-up calls began to caregivers who had not previously completed the mail survey; over the course of five weeks, up to five telephone calls were made to try to complete the survey by telephone in English or Spanish. The study was approved by RAND’s Human Subjects Protection Committee.

Analyses

We tested differences in decedent/caregiver characteristics between respondents and non-respondents using chi-square tests. For each evaluative item, we calculated “top-box” scores, the percentage of caregivers that endorsed the most positive response option(s) (e.g., “Always” for questions using a Never/Sometimes/Usually/Always response scale). Top-box scores are often used for reporting results of care experience surveys, as they are easily interpretable and promote high standards of care quality.¹⁴

We conducted confirmatory factor analysis (CFA) to evaluate the factor structure of 14 evaluative items in the draft survey. We used mean- and variance-adjusted weighted least squares to account for the dichotomous nature of top-box item scores.¹⁵ We used a criterion of factor loadings ≥0.40 for inclusion within proposed factors,¹⁶ and assessed overall model fit using the Comparative Fit Index (CFI), the root mean square error of approximation (RMSEA), and weighted root mean square residual (WRMR). A model with good fit typically has a CFI >0.95, RMSEA <0.05, and WRMR <1.0,¹⁷ with WRMR being less critical.^18,19 We hypothesized five factors: getting timely care, communication, treating patient with respect, consistency of care with patient preferences, and getting help for symptoms. With the factorial structure confirmed, we calculated composite measure scores for each factor as the average of top-box-scored items included in that factor and assessed the degree to which these five composite measures assess distinct content domains by calculating their correlations. Correlations exceeding 0.80 suggest that composites are measuring aspects of care that are insufficiently distinct.¹⁶ We also calculated the internal consistency reliability of the composite measures using Cronbach’s alpha. Cronbach’s alphas of 0.70 or higher are considered adequate for group comparisons.²⁰

To assess construct validity, the degree to which the composite measures are related to respondents’ overall reported care experiences, we evaluated the associations of each composite measure’s top-box score with the top-box score of the respondents’ overall rating of care (a rating of nine or 10 on a scale of zero to 10). We estimated linear regression models with the overall rating as a dependent variable, adjusting for patient and caregiver characteristics that are known to be associated with systematic differences in survey response: decedent age, caregiver-reported decedent health conditions and length of time between last clinic visit or virtual or home visit/encounter and death, caregiver age, education, relationship to the decedent, language spoken at home/survey language, and response percentile (the length of time between patient death and survey response).²¹ We fit models that included only one composite at a time as a predictor, as well as a model that included all composites simultaneously as predictors to measure the unique association of each composite with the overall rating beyond all other composites.

To explore the degree to which evaluative items and composite quality measures can identify variation across subgroups, we used linear regressions to test differences in item and measure scores by health care utilization, health care transitions, and location of death, adjusting for all case-mix variables noted above and region. Location of death was categorized using caregiver-reported location of death and the administrative variable for receipt of hospice care in the last 12 months of life (home with hospice care within 12 months of death, home without hospice care within 12 months of death, assisted living facility, nursing home, hospital, hospice facility or hospice house, or other). Health care utilization variables included an administrative variable for receipt of hospice care in the last 12 months of life and a flag for care transitions, indicating whether the decedent had received care in more than one setting in the last 12 months of life. Missing data were rare (less than 5% for all predictors); predictors were imputed with the mean within the health system region, and if still missing, with the overall mean.

Results

Characteristics of decedents

The response rate was 30.1%. Nine percent of decedents for whom care experiences were reported in our data were non-Hispanic Black, 20.1% were Hispanic, and 61.3% were non-Hispanic White (Table 1). Approximately three in five decedents (62.3%) received hospice care in the year prior to death, with the percent of decedents using hospice varying considerably by cause of death. Forty-nine percent died at home, 24.4% in the hospital, and 5.3% in a nursing home or skilled nursing facility. Respondents were more likely than non-respondents to be caregivers of decedents age 80 or older, with shorter lengths of time between the last health care visit and death, who had heart disease or dementia, more diagnoses, and who received hospice care in the last year of life (Supplementary Table A1).

Table 1.

Characteristics of Decedents and Caregiver Respondents

	n (%)
Characteristics of Decedents
Sex^a
Female	240 (49.2%)
Male	248 (50.8%)
Age^a
18–64	42 (8.6%)
65–79	157 (32.2%)
80–89	180 (36.9%)
90 or older	109 (22.3%)
Race / Ethnicity^b
Hispanic	98 (20.1%)
Non-Hispanic Black or African American	46 (9.4%)
Non-Hispanic White	299 (61.3%)
All other	30 (6.1%)
Education^b
Less than high school	75 (15.4%)
High school graduate or GED	121 (24.8%
Some college or 2-year degree	106 (21.7%)
College degree or more	163 (33.4%)
Cause of Death^b,c
Accident or injury	28 (5.7%)
Alzheimer’s or other dementia	91 (18.6%)
Cancer	149 (30.5%)
COVID-19	23 (4.7%)
Diabetes or high blood sugar	36 (7.4%)
Emphysema, asthma, COPD, or other lung problems	90 (18.4%)
Flu or pneumonia	31 (6.4%)
Heart disease or heart attack	105 (21.5%)
Renal failure or kidney disease	75 (15.4%)
Stroke	35 (7.2%)
Other cause	144 (29.5%)
Hospice Care in Last 12 Months^a	304 (62.3%)
Location of Death^b
Home	238 (48.8%)
Hospital	119 (24.4%)
Hospice facility or hospice house	45 (9.2%)
Assisted living facility	38 (7.8%)
Nursing home or skilled nursing facility	26 (5.3%)
Other place	11 (2.3%)
Characteristics of Caregiver Respondents
Relationship to Decedent (Caregiver is decedent’s…)^b
Spouse or partner	201 (41.2%)
Child	206 (42.2%)
Friend	18 (3.7%)
All other	60 (12.3%)
Preferred Language (language spoken at home or used for survey response)^a,b
Survey and home language are English	425 (87.1%)
Survey or home language are Spanish	31 (6.4%)
Other	16 (3.3%)

All means and percentages are calculated including missing values. There were 3.1% of respondents missing race/ethnicity, 4.7% missing decedent education, 1.6% missing cause of death, 2.3% missing location of death, 0.6% missing caregiver relationship to decedent, and 3.3% missing caregiver-preferred language.

Variable from administrative data.

Variable from survey response data.

Note that respondents could give more than one response to this item. Percentages may add up to more than 100%.

Composite measures

The five-factor CFA model provided an excellent fit to the data, χ²₍₆₇₎ = 110.42, p < 0.001; CFI = 0.995; RMSEA = 0.036; WRMR = 0.776. Table 2 displays the factor loadings and corrected item-total correlation for the 14 evaluative items proposed for the five composite measures, along with Cronbach’s alpha internal consistency for each composite measure. The factor loadings ranged between 0.72 and 0.96 and corrected item-total correlations ranged between 0.50 and 0.68, suggesting these items are strong indicators of the corresponding factor. The five composites were moderately correlated (Table 3). Intercorrelation was highest between Communication and Treating Patient with Respect (r = 0.71).

Table 2.

Psychometric Properties of Survey of End-of-Life Care Quality Measures and Component Items

Composite measure and component items	Response options	Cronbach’salpha	Cronbach’salpha withitem deleted	Item-Totalcorrelation	Factorloading
Getting Timely Care		0.770
When you or your family member contacted a health care provider during regular office hours, how often did you get the help needed?^a	Never/Sometimes/Usually/Always		NA	0.631	0.887
When you or your family member contacted a health care provider during evenings, weekends, or holidays, how often did you get the help needed?^a	Never/Sometimes/Usually/Always		NA	0.631	0.940
Communication		0.818
During the last month of life, how often did your family member’s health care providers explain things in a way that was easy to understand?	Never/Sometimes/Usually/Always		0.772	0.638	0.867
During the last month of life, how often did your family member’s health care providers seem to know the important information about his or her medical history?	Never/Sometimes/Usually/Always		0.794	0.589	0.781
During the last month of life, how often were you and your family member kept informed about his or her condition?	Never/Sometimes/Usually/Always		0.768	0.647	0.850
During the last month of life, how often did health care providers listen carefully to you and your family member?	Never/Sometimes/Usually/Always		0.751	0.684	0.944
Care Preferences		0.719
During the last month of life, did health care providers involve you and your family member in decisions as much as you both wanted?	Yes, definitely/Yes, somewhat/No		0.621	0.547	0.914
Did health care providers do the best they could to respect your family member’s wishes?^a	Yes, definitely/Yes, somewhat/No		0.584	0.577	0.87
Did health care providers do the best they could to honor your family member’s desired location to pass away?	Yes, definitely/Yes, somewhat/No/Don’t Know		0.684	0.495	0.717
Treating Patient with Respect		0.768
During the last month of life, how often did health care providers treat your family member with dignity and respect?	Never/Sometimes/Usually/Always		NA	0.623	0.907
During the last month of life, how often did you feel that health care providers really cared about your family member?	Never/Sometimes/Usually/Always		NA	0.623	0.961
Getting Help for Symptoms		0.718
Did your family member get as much help with pain as he or she needed?^a	Yes, definitely/Yes, somewhat/No		0.553	0.599	0.895
How often did your family member get the help he or she needed for trouble breathing?^a	Never/Sometimes/Usually/Always		0.654	0.517	0.746
How often did your family member get the help he or she needed for feelings of anxiety or sadness?^a	Never/Sometimes/Usually/Always		0.675	0.500	0.806

Top-box response options shown in bold.

A screening question identifies respondents eligible to respond to this evaluative item.

Table 3.

Correlations Among Survey of End-of-Life Care Composite Measure Scores

	Getting timelycare	Communication	Carepreferences	Treating patientwith respect	Getting helpfor symptoms
Getting Timely Care	1
Communication	0.605	1
Care Preferences	0.461	0.627	1
Treating Patient with Respect	0.473	0.711	0.592	1
Getting Help with Symptoms	0.448	0.586	0.531	0.516	1

All correlations are significant at p < 0.0001.

Among the models that consider each composite’s association with overall rating individually, the Communication, Treating Patient with Respect, and Care Preferences composites were the strongest predictors of overall rating of care (β = 0.612, β = 0.559, and β = 0.554, respectively; Table 4, column 2). Since these composites were among the most correlated with one another, the coefficients for these measures were smaller, and the coefficient for communication was no longer statistically significant in a model containing all five composites (Table 4, column 3). The model including all composites accounted for 47.1% of the variance in overall rating of care.

Table 4.

Using Survey of End-of-Life Care Measures to Predict Overall Ratings of Care

	Models assessingcompositesone-by-one	Model includingall composites
	Estimate	Estimate
Getting Timely Care	0.502^**	0.207^**
Communication	0.612^**	0.095
Care Preferences	0.554^**	0.182^**
Treating Patient with Respect	0.559^**	0.183*
Getting Help for Symptoms	0.531^**	0.236^**

Table reports coefficients of regression models predicting overall ratings of care from quality measure composites. Models are adjusted for case mix, including decedent age, caregiver-reported decedent health conditions, and length of time between last clinic visit or virtual or home visit/encounter and death, caregiver age, education, relationship to the decedent, language spoken at home/survey language; and response percentile (the length of time between patient death and survey response).

p < 0.01.

p < 0.001.

Reported care experiences

Across all respondents, the composites showing the greatest room for improvement were Getting Timely Care (top-box score of 53.9), Getting Help for Symptoms (57.4), and Communication (58.5; Table 5). Just 37.9% of all respondents indicated that the decedent “always” got needed help for feelings of anxiety or sadness, 52.2% to 53.8% reported that the decedent “always” got the help they needed during regular office hours and on evenings, weekends, or holidays, and 54.3% reported that the decedent’s health care providers “always” seemed to know the important information about their medical history during their family member’s last month of life (Supplementary Table A1).

Table 5.

Top-Box Scores for Survey of End-of-Life Care Measures, Overall and for Subgroups of Interest

		Location of death			Health care transitionsin last month of life
	All(n = 488)	Home with hospice care within 12 months of death (n = 194)	Hospital(n = 119)	Nursinghome(n = 26)	One setting(n = 155)	More than one setting (n = 328)
Getting Timely Care	53.94	61.28 (ref)	48.26	23.07^***	57.60 (ref)	52.25
Communication	58.48	64.92 (ref)	56.38	43.11^*	61.51 (ref)	57.05
Care Preferences	76.30	83.10 (ref)	71.02^**	63.50^**	79.53 (ref)	74.74
Treating Patient with Respect	69.38	72.93 (ref)	67.74	61.82	71.55 (ref)	68.35
Getting Help for Symptoms	57.36	63.36 (ref)	54.28	48.95	64.34 (ref)	54.19^*
Emotional Support (Single Item)
During the last month of life, how much emotional support did you and your family member get from health care providers?	70.42	75.04 (ref)	67.43	58.63	75.14 (ref)	68.19
Overall Rating (Single Item)
Using any number from 0 to 10, where 0 is the worst health care possible and 10 is the best health care possible, what number would you use to rate your family member’s health care during the last month of life?	57.75	65.44 (ref)	58.57	30.14^**	66.96 (ref)	53.38^*

Top-box scores refer to the percent of respondents who endorsed the most positive response(s) to the survey item or composite quality measure. Models examined location of death (home with hospice care within 12 months of death, home without hospice care within 12 months of death, assisted living facility, nursing home, hospital, hospice facility / hospice house, and other) and health care transitions in last month of life (one setting and more than one setting); scores for some categories are not shown due to space constraints. Scores for each subgroup are adjusted for case mix and region; case mix variables include decedent age, caregiver-reported decedent health conditions, and length of time between last clinic visit or virtual or home visit/encounter and death, caregiver age, education, relationship to the decedent, language spoken at home/survey language; and response percentile (the length of time between patient death and survey response). Missing predictors are imputed with the mean within region and, if still missing, with the overall mean.

p < 0.05.

p < 0.01.

***

p < 0.0001.

In contrast, 83.6% of caregivers reported that health care providers “definitely” did their best to honor the decedent’s desired location to die, and 78.5% reported that health care providers “definitely” did the best they could to respect the decedent’s wishes.

Variation in reported care experiences by subgroup

There was notable variation in reported care experiences by decedent location of death and health care utilization patterns (Table 5). For example, caregivers of those who died at home with hospice care within 12 months of death reported substantially more positive care experiences than those whose loved ones died in a nursing home (range: top-box measures scores of 11 to 38 points higher with differences p < 0.05 or less for four of the seven composite and single-item measures).

Caregivers of decedents who transitioned from one setting of care to another in the last month of life had poorer care experiences than those who had received care in one setting, particularly with regard to getting help for symptoms (adjusted top-box score of 54.2 versus 64.3, p < 0.05); these caregivers were also substantially less likely to rate their care a nine or 10 out of 10 (53.4 vs. 67.0; p < 0.05). Caregivers of those who received hospice care in the last 12 months of life reported better care experiences than those who did not receive hospice care, particularly regarding respect for care preferences (top-box score of 79.7 versus 70.7, p < 0.05).

Discussion

Current federal policy focuses on value-based payment with goals that include constraining health care costs. Avoiding hospitalizations and costly treatments in the last month of life is important for achieving this goal. Reducing intensity of care in the last month of life is defensible and desirable if the care delivered is consistent with patients’ goals and responsive to their needs for symptom palliation and emotional support. Retrospective surveys of family caregivers are a valuable source of information regarding these aspects of quality of care at the end of life. Typically, these surveys rely on death certificates or caregiver information collected by hospices to identify the survey sample. Our study tested and confirmed that it is possible to use administrative data from an MA organization to identify a sample of decedents and their family caregivers for a survey of end-of-life care experiences across care settings. This approach allows for examination of a time period when most patients experience one or more health care transitions and one or more symptoms in need of palliation. The resulting survey-based measures are internally reliable and valid. They may be particularly useful for assessing quality of care in MA and in emerging value-based models.²²

In keeping with prior studies, we found that reported care experiences are better for those who die at home,²³ are worse for those who have experienced transitions across settings of care,⁶ and that there is the greatest room for improvement regarding help for anxiety or sadness and keeping the caregiver informed of the patient’s condition.²⁴ These findings suggest that measures from the CAHPS End-of-Life Care Survey can be used to identify variation in experiences across patient subgroups and aspects of care. Reported experiences reflect care received during the COVID-19 pandemic, and therefore must be interpreted considering pandemic-era restrictions on visitation, staffing shortages, and altered modes of communication between health care professionals, patients, and caregivers.^25,26

Our results should be interpreted in light of some limitations. First, we surveyed caregivers 8–12 months after the death of their loved one, and some caregivers may have also previously received the CAHPS Hospice Survey. Consequently, some sampled caregivers may have been more likely to respond than others, and our response rate was 30 percent, lower than desirable (although similar to that of a widely used national post-death survey of caregivers).²⁷ Low response rates can raise concerns about response bias (i.e., the degree to which responses are representative of the full population of interest); we adjust for differences in case mix to address concerns about possible nonresponse bias when comparing across subgroups of respondents.^28,29 In future applications of the survey, response rates may be improved by administering the survey closer to the death of the patient, without a previously administered care experience survey, and using more recent contact information for caregivers; earlier survey administration following patient death may also improve caregiver recall. Second, we conducted the field test in English and Spanish; further testing would be needed to incorporate perspectives of those who speak other languages. Third, respondents to the CAHPS End-of-Life Care Survey are caregivers rather than patients. These respondents act as proxies for some questions (e.g., symptoms), and in many instances, are involved directly in health care communication and decision-making. Surveying caregivers allows for assessment of care in the final period of life (which cannot be achieved via patient report), promotes parallelism in the time frame assessed (i.e., final episode of care), and reduces selection bias that would result from a patient survey, since many patients are too ill or cognitively impaired to respond to a survey at the end of life. Caregivers’ responses have moderate-to-high agreement with patient responses regarding observable symptoms, such as fatigue or shortness of breath, and quality of care, such as being kept informed by health care providers.^30,31 Although caregiver respondents may sometimes answer differently than patients would have, there is no evidence that these differences reduce the validity of comparisons across entities. Finally, we field-tested the survey in an integrated health system that had invested in data infrastructure that facilitated timely identification of decedents and caregivers; other MA organizations may need to make similar investments prior to fielding such a survey.

Conclusion

With the rapid growth of value-based initiatives and MA enrollment, accountability measures are needed to examine the quality of care at the end of life, an important and costly period in which many individuals receive care in more than one setting, facing health care transitions and difficult care planning decisions. The CAHPS End-of-Life Care Survey can be administered using MA data, assesses a broad range of care experiences that are important indicators of high-quality end-of-life care, and aligns with other surveys in national use. Quality measures derived from CAHPS End-of-Life Care Survey responses can be used to assess and compare care experiences in health systems, MA organizations, or accountable care organizations.

Footnotes

Acknowledgments

The authors gratefully acknowledge Margaret Wang for her assistance with the design of the survey field test, Eleanor Jensen and Daniel Johnson for participation in interviews to inform the content of the CAHPS End-of-Life Care Survey, and Julie Brown and Ron Hays for their comments on an earlier draft of the article.

Author Disclosure Statement

No competing financial interests exist.

Funding Information

This work was supported by a cooperative agreement from the Agency for Healthcare and Research Quality (AHRQ) (Contract number U18 HS025920).

Supplementary Material

References

Hospices Likely to Keep Moving Towards Value-Based Care. Available from: https://hospicenews.com/2022/09/09/hospices-likely-to-keep-moving-towards-value-based-care/ [Last accessed: January 31, 2025].

National Academies of Sciences Engineering and Medicine: Integrating the patient and caregiver voice into serious illness care: Proceedings of a workshop. Washington, DC: The National Academies Press; 2017.

Anhang Price

, Elliott

. Measuring patient-centeredness of care for seriously Ill individuals: Challenges and opportunities for accountability initiatives. J Palliat Med, 2018; 21(S2):S28–S35.

Centers for Medicare & Medicaid Services: CAHPS Hospice Survey website. Available from: https://www.hospicecahpssurvey.org [Last accessed: January 31, 2025].

U.S. Department of Veterans Affairs Center for Health Equity Research and Promotion: The Bereaved Family Survey. Available from: https://www.cherp.research.va.gov/PROMISE/The_Bereaved_Family_Survey.asp/ [Last accessed: January 31, 2025].

Makaroun

, Teno

, Freedman

, et al. Late transitions and bereaved family member perceptions of quality of end-of-life care. J Am Geriatr Soc, 2018; 66(9):1730–1736.

Evans

, Berman

, Burlingame

, et al. It’s time to take patient experience measurement and reporting to a new level: Next steps for modernizing and democratizing national patient surveys. Health Affairs Forefront, 2020.

Office for National Statistics: National Survey of Bereaved People (VOICES). Available from: https://www.ons.gov.uk/peoplepopulationandcommunity/healthandsocialcare/healthcaresystem/bulletins/nationalsurveyofbereavedpeoplevoices/england2015 [Last accessed: January 31, 2025].

Teno

, Clarridge

, Casey

, et al. Validation of toolkit after-death bereaved family member interview. J Pain Symptom Manage, 2001; 22(3):752–758.

10.

National Health & Aging Trends Study Methods & Documentation. Available from: https://nhats.org/researcher/nhats/methods-documentation [Last accessed: January 31, 2025].

11.

National Consensus Project for Quality Palliative Care: Clinical Practice Guidelines for Quality Palliative Care, 4th edition. National Coalition for Hospice and Palliative Care: Richmond, VA; 2018.

12.

Agency for Healthcare Research and Quality. CAHPS Patient Experience Surveys and Guidance. Available from: https://www.ahrq.gov/cahps/surveys-guidance/index.html [Last accessed: January 31, 2025].

13.

Wang

, Johnson

, Glass

, et al. Assessing end-of-life care across settings in an integrated health care delivery system. J Palliat Med, 2020; 23(6):825–828.

14.

Robert Wood Johnson Foundation: How to Report Results of the CAHPS Clinician & Group Survey. 2010.

15.

Flora

, Curran

. An empirical evaluation of alternative methods of estimation for confirmatory factor analysis with ordinal data. Psychol Methods, 2004; 9(4):466–491.

16.

Brown

. Confirmatory Factor Analysis for Applied Research. 2nd ed. Guilford Press: New York, NY; 2015.

17.

Muthén

. Mplus Technical Appendices. Muthén & Muthén: Los Angeles, CA; 1998–2004.

18.

DiStefano

, Liu

, Jiang

, et al. Examination of the weighted root mean square residual: Evidence for trustworthiness? Structural Equation Modeling: A Multidisciplinary Journal, 2018; 25(3):453–466.

19.

, Bentler

. Cutoff criteria for fit indexes in covariance structure analysis: Conventional criteria versus new alternatives. Structural Equation Modeling: A Multidisciplinary Journal, 1999; 6(1):1–55.

20.

Nunnally

. Bernstein IH: Psychometric Theory. McGraw-Hill: New York, NY; 1994.

21.

Parast

, Haas

, Tolpadi

, et al. Effects of caregiver and decedent characteristics on CAHPS Hospice Survey scores. J Pain Symptom Manage, 2018; 56(4):519–529.e511.

22.

Ankuda

, Aldridge

, Braun

, et al. Addressing serious illness care in medicare advantage. N Engl J Med, 2023; 388(19):1729–1732.

23.

Teno

, Clarridge

, Casey

, et al. Family perspectives on end-of-life care at the last place of care. JAMA, 2004; 291(1):88–93.

24.

Ankuda

, Kelley

, Morrison

, et al. Family and friend perceptions of quality of end-of-life care in medicare advantage vs traditional medicare. JAMA Netw Open, 2020; 3(10):e2020345.

25.

Feder

, Smith

, Griffin

, et al. Why Couldn’t I Go in To See Him?” Bereaved Families’ Perceptions of End-of-Life Communication During COVID-19. J American Geriatrics Society, 2021; 69(3):587–592.

26.

Kates

, Gerolamo

, Pogorzelska-Maziarz

. The impact of COVID-19 on the hospice and palliative care workforce. Public Health Nurs, 2021; 38(3):459–463.

27.

Centers for Medicare & Medicaid Services. CAHPS Hospice Survey website, Scoring and Analysis. Available from: https://hospicecahpssurvey.org/en/public-reporting/scoring-and-analysis/ [Last accessed: January 31, 2025].

28.

Elliott

, Edwards

, Angeles

, et al. Patterns of unit and item nonresponse in the CAHPS® hospital survey. Health Services Research, 2005; 40(6p2):2096–2119.

29.

Elliott

, Zaslavsky

, Goldstein

, et al. Effects of survey mode, patient mix, and nonresponse on CAHPS® hospital survey scores. Health Serv Res, 2009; 44(2 Pt 1):501–518.

30.

McPherson

, Addington-Hall

. Judging the quality of care at the end of life: Can proxies provide reliable information? Soc Sci Med, 2003; 56(1):95–109.

31.

Elliott

, Beckett

, Chong

, et al. How do proxy responses and proxy-assisted responses differ from what Medicare beneficiaries might have reported about their health care? Health Serv Res, 2008; 43(3):833–848.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.09 MB

0.01 MB