Abstract
To encourage person-centered care, the Centers for Medicare and Medicaid require nursing homes to measure resident preferences using the Preferences Assessment Tool (PAT). No known research has examined the implications of respondent type (i.e., resident, proxy, staff) on preference importance; therefore, the purpose of this study was to compare the importance of preferences depending on which respondent completed the PAT. Participants included 16,111 Veterans discharged to community-based skilled nursing facilities after hospitalization for heart failure. A majority (95%) of residents completed the PAT compared to proxy (3%) and staff (2%). Proxy responders were both more and less likely to indicate individual preferences as important compared to residents. Staff members were consistently less likely to indicate all preferences as important compared to residents. Findings from this study emphasize the need for proxy and staff to find methods to better understand residents’ preferences when residents are not able to participate in assessments.
Few proxies or staff complete the Preference Assessment Tool (PAT) compared to nursing home residents. Proxies may not accurately rate the importance of residents' care preferences. Staff persons seem to consistently underestimate the importance of residents' care preferences.
Even when proxies and staff complete the PAT, often, residents were documented as able to understand others/be understood. A better understanding of how and why different responders complete the PAT is needed. Proxy and staff responders might underestimate what is most important to residents, so it’s important that residents' preferences be continuously re-assessed, especially when staff and proxies complete the PAT. Observation of residents' affect and behaviors may be a useful way to understand their care preferences, especially when residents do not or cannot complete the PAT.What this paper adds
Applications of study findings
Introduction
Nursing homes and other forms of long-term services and supports (LTSS), under the guidance of the Centers for Medicare and Medicaid (CMS), are increasingly transitioning away from task-focused, provider-driven care to requiring a more person-centered, preference-based care approach. A person-centered care (PCC) model encourages resident autonomy and facilitates care delivery that is congruent with resident preferences (Duan et al., 2020; Van Haitsma et al., 2020; Wagner et al., 2021; Yevchak et al., 2017). Measuring and honoring resident preferences is associated with quality of life, quality of care, life satisfaction, and behavioral health (Abbott et al., 2018; Andrew & Meeks, 2018; Bangerter et al., 2016; Shippee et al., 2013; Van Haitsma et al., 2013).
In the US, nursing homes reimbursed by CMS are required to measure resident preferences using the Preferences Assessment Tool (PAT; Section F) on the Minimum Data Set (MDS) 3.0 (Kane, 2003; Saliba & Buchanan, 2008). The PAT assesses the importance of 16 daily routine and activity preferences that can impact quality of life for residents (e.g., ability to use a phone in private; participate in religious services). Previous research using the PAT has confirmed the relationship between the importance of residents’ preferences and individual/organizational characteristics (i.e., sex; race; cognitive status; depressive symptoms; physical function; rurality; fewer staffing hours; Duan, Shippee, et al., 2020; Heid et al., 2016; Housen et al., 2008; Roberts et al., 2018; Roberts & Saliba, 2019); but limited research exists on the measurement properties of the PAT across residents with different diagnoses and abilities to convey their preferences.
A unique aspect distinguishing the PAT from other MDS measures is it can be filled out by either the resident, a proxy respondent (typically a family member or significant other), or a staff member depending on the resident’s cognitive capacity and proxy availability. To date, no known study has examined if there are any differences in the importance of preferences depending on if the resident, a proxy, or a staff member completes the survey. If differences exist, this has implications for the validity of the PAT and the provision of PCC for nursing home (NH) residents. Indeed, it may well be that proxy or staff respondents do not accurately assess residents’ preferences for care, which could have a downstream effect on reduced PCC. As such, the purpose of this study was to compare the importance of preferences among residents, proxy respondents, and staff members completing the PAT.
Methods
Study Design and Data Sources
In this cross-sectional secondary data analysis, we linked Veterans Health Administration (VHA) electronic medical records to Minimum Data Set (MDS) version 3.0 assessments and Medicare claims using unique Veterans Administration (VA) identifiers. This study was approved by the Institutional Review Board at the Providence VA Medical Center.
Participants
Our study population included Veterans who were hospitalized for heart failure (HF) at a U.S. VA hospital between October 1, 2010 and September 30, 2015 and were subsequently discharged to a non-VA nursing home. We examined non-VA nursing homes as data restrictions prevented us from reliably ascertaining MDS records from VA-affiliated nursing homes. Veterans hospitalized for HF were identified through hospitalization claims if an International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) code for HF was documented in the primary position with codes of 402.01, 402.11, 402.91, 404.01, 404.03, 404.11, 404.13, 404.91, 404.93, 428.0, 428.1, 428.20, 428.21, 428.22, 428.23, 428.30, 428.31, 428.32, 428.33, 428.40, 428.41, 428.42, 428.43, or 428.9. Discharges to nursing homes were captured if an admission MDS assessment was documented up to 14 days after the date of discharge from the index HF hospitalization. Veterans were excluded if they were: (1) considered “long-term” residents (i.e., nursing home stay totaled 90 days or more), (2) not admitted to a nursing home during 0–7 days after discharge from a hospital, (3) missing data on admission assessment from 1–14 days, (4) missing data beyond admission assessment, (5) missing data on two or more items on the PAT, and/or (6) missing data on related covariates. When administering the PAT to the resident, the user is instructed to stop the assessment and proceed with a proxy or staff member if there are two or more items the resident cannot answer. Therefore, we excluded assessments with missing data for two or more items on the PAT and imputed data for participants missing one item on the PAT (Saliba & Buchanan, 2008). For Veterans with multiple eligible HF hospitalizations with subsequent discharges to nursing home, only the first hospital admission was included.
Measures
The MDS is a federally mandated, standardized resident assessment administered at scheduled intervals in Medicare and Medicaid-certified nursing homes. For this study, residents’ MDS assessments collected at the time of residents’ admission to the nursing home were used. MDS-derived variables included demographic information (i.e., marital status), functional and cognitive status (e.g., dementia; delirium; cognitive function; activities of daily living), and ability to understand and be understood by others (i.e., speech clarity; make self understood; ability to understand others). Information extracted from VHA Electronic medical records and hospitalization claims included demographics such as age, sex, race/ethnicity, comorbidities (i.e., summary of the Elixhauser index; Elixhauser et al., 1998), prior healthcare utilization (i.e., length of hospital stay), and clinical condition (i.e., ejection fraction).
Preference importance was assessed using Section F of the MDS (the PAT), which assesses the importance of 16 daily care and activity preferences (Housen et al., 2009). We used Section F of the MDS assessment completed on residents’ admission to the nursing home. Residents are asked to complete the PAT, but if they cannot be understood a proxy respondent is asked to complete the PAT on the resident’s behalf. If a proxy respondent is not available, or the resident/proxy did not complete three or more items on the PAT, a staff member completes the PAT. If residents or proxy respondents complete the PAT, it is on a 4-point Likert-type scale, ranging from Very Important (1) to Not important at all (4) with a fifth option of Important, but can’t do or no choice. If staff members complete the PAT, it is dichotomized into whether the resident prefers or does not prefer the activity. For this analysis, PAT items were dichotomized into two categories: preference or no preference. Preference included items rated as very important, somewhat important, important but can’t do or no choice; no preference included items rated as not very important and not important at all (Roberts & Saliba, 2019).
Data Analysis Plan
First, descriptive analysis was conducted to compare different characteristics of residents by the source of responses for daily preferences (resident, proxy, and staff) using Chi-square tests for dichotomous or categorical variables and ANOVA for continuous variables. Second, frequency was performed to examine differences between a preference across the 16 PAT items using Chi-square tests and Cramér’s V effect sizes. Third, logistic regressions were conducted to estimate the odds of PAT items associated with each source of responses using the residents as the reference group. We adjusted for age, sex, race/ethnicity, marital status, length of stay, dementia, delirium, cognitive function, speech clarity, ability to make self understood, ability to understand others, summary of Elixhauser index, activities of daily living, and ejection fraction in the logistic regression. We assessed for multicollinearity by consulting the variance inflation factor (VIF; Gareth et al., 2013). VIF did not exceed 5 in any model, suggesting no issues of multicollinearity. For the frequency comparisons and logistic regression among the PAT items, we emphasized confidence intervals and effect sizes to determine the magnitude of the effects (Cumming, 2014), as we expected the analyses to be over-powered due to the large sample size.
Missing data were handled by comparing the characteristics of participants missing data on the PAT with participants missing no data. Participants were excluded from the cohort if they were missing two or more items on the PAT or any covariates; however, if participants were only missing one item on the PAT they were kept in the cohort. Comparing the two groups helps us understand any patterns or possible reasons for those missing data. Then, the combination of multiple imputation techniques and missing indicators were performed to address missing data for all variables included in the analyses (source of responses for daily preferences, 16 PAT items, and the covariates). This approach gives minimal bias in most scenarios (Sperrin & Martin, 2020). The missing indicator was created and coded as 1 for missing and 0 for non-missing, then was also included in the multiple imputation models to fill in the missing values for all the variables included in logistic regression models (Sperrin & Martin, 2020). Sensitivity analyses were performed to understand the differences in the estimates of the association of a preference across the PAT items compared to complete case analysis. Analyses were conducted in SAS v9.4 (SAS Institute Inc., Cary, NC).
Results
Resident Characteristics by Section F Respondent Type.
Notes: p-value refers to the omnibus test across all three categories.
pairwise comparison between resident & staff significant at p < .05,
pairwise comparison between resident & proxy significant at p < .05,
pairwise comparison between proxy & staff significant at p < .05.
Importance of PAT Items by Respondent Type.
Odds Ratios and 95% Confidence Intervals Comparing Proxy and Staff to Residents Indicating PAT Items as Important.
Discussion
Although there are measures in place to facilitate identification and incorporation of resident preferences in clinical care, whether reporting source impacts the importance of identified preferences is unknown. We examined this question in an epidemiological sample of Veterans hospitalized for HF in community-based nursing homes. Overall, few proxies or staff completed the PAT. Instead, most residents were able to report for themselves what preferences were important to them. This may indicate that the PAT is a useful tool for a wide range of residents. Some research suggests this may be the case as residents living with cognitive impairment are able to complete the tool accurately and consistently ((Carey et al., 2018); Saliba & Buchanan, 2008).
Targeted research on cognitive function and ability to report preferences is needed, but in the meantime, the PAT remains a useful tool to try to understand residents’ important preferences. Interestingly, while most residents were able to complete the PAT themselves, for those that were not able to complete the PAT, staff still reported many residents as being able or usually to be understood and understand others. This is an interesting finding that requires future study to ascertain what the differentiating factor is when staff determine who is the most appropriate person to complete the assessment. This discrepancy could be due to diagnosis, the assessment algorithm, or even resident preference; but, our analysis limits us from providing an answer. Future research on how staff make decisions to administer the PAT might have important implications for staff training and tool use.
Our results indicated that a small, but significant number of resident preferences were provided by proxy or staff respondents. Analysis of the pattern of endorsement indicates that proxy preference reporters were more likely to rate preferences pertaining to autonomy as important, but less likely to prioritize preferences related to self-care or meaningful activities in comparison to resident respondents, whereas staff were overall less likely to rate any preferences as important. The circumstances dictating whether proxies versus residents answered are not clear from the existing dataset. It is likely that proxies completed assessments for residents who were unable to provide their own responses. Therefore, residents with more severe symptoms, such as significant cognitive or physical impairments, may be more reliant on proxies to provide this information. This reliance on proxies and staff members, however, could lead to incorrect or inconsistent reporting of care preferences.
There were key differences in the preferences endorsed by reporting source that have implications for care provision and resident quality of life. We found that proxies were more likely to indicate preferences as important (compared to residents) when they related to how residents spent their time (i.e., spending time with family/friends and choosing bedtimes) or what was available to them (i.e., having snacks available)—reflecting a preference for residents’ autonomy in their daily routines. In contrast, proxies rated preferences related to meaningful activities, (i.e., using a phone in private, having reading materials available, keeping up with the news, or going outside when the weather is good), and self-care (i.e., choosing what clothes to wear, choosing how to bathe, having a place to lock things up, or taking care of their belongings) as less important than resident respondents. These differences indicate that the information gained from proxies may not accurately reflect resident preferences. This finding is similar to previous work in the nursing home (Heid et al., 2020) and community setting (Carpenter et al., 2006) which demonstrated how caregivers tend to underestimate residents’ preference for enrichment and growth activities. One explanation for this phenomenon could be due to residents’ loss of sensory abilities, especially hearing and vision, causing proxies' perception of the resident’s level of interest in such activities to be altered. Proxies’ level of involvement, their knowledge of residents’ personality traits, and their own sense of well-being may also influence their perception of resident preferences (Heid et al., 2017).
Proxys’ perceptions of residents’ preferences may be based on knowledge of residents’ past preferences which may change over time along with residents’ ability to communicate new preferences (Carstensen et al., 1999; Heid et al., 2020). Previous work indicates that although preferences are relatively stable over a 3 month time period (Abbott et al., 2018), situational experiences (Heid et al., 2017) within the person (e.g., functional ability, personal schedule), facility environment (e.g., facility schedule, facility policy), social environment (e.g., quality and type of interactions), and global environment (e.g., weather, current events) can affect the reporting of the importance of a preference at any given point in time (Abbott et al., 2018). In clinical practice, especially when residents are not able to effectively communicate, it may be more appropriate to identify preferences based on observation of resident behaviors and indicators of engagement while involved in the activities.
Adding observed indicators of engagement to the PAT might facilitate better understanding of residents’ preferences and improve the delivery of person-centered care. Using preference satisfaction ratings via assessments such as ComPASS (Care Preference Assessment of Satisfaction; https://compass.linkedsenior.com/) is an alternative strategy for those who can participate more fully in the assessment process (Van Haitsma et al., 2014). Preference satisfaction ratings use information from the PAT to evaluate how well important preferences are satisfied over time (Heid et al., 2020) and have been linked to residents’ overall satisfaction with care (Madrigal et al., 2022). Both objective, observation-based measures and subjective, survey-based measures give residents and staff the opportunity to communicate about preferences and improve the delivery of person-centered care.
In our study, when comparing staff indicated preferences to resident indicated preferences, staffs were less likely to indicate each PAT item as important. Staff can only indicate one of two choices on the PAT, whether the resident does or does not prefer the preference in question. This likely impacts the results in our study and also the accuracy of documentation reflecting residents’ preferences. Frequency of staff contact with the resident and recall of perceived preferences may also explain the difference. A resident’s level of engagement with an activity may also change and fluctuate with time, environment, and resident’s internal state. This again points to the need of having an objective assessment of a resident’s level of engagement with an activity which is spread over time and informs the assessment and fulfillment of PAT preferences.
Discrepancies between residents, proxies, and staff perceptions of resident preferences should be considered within the context of other research in this area. Previous research highlights the complexity and variability in resident preferences including: who is involved in care discussions, attitudes of staff, residents’ perception of choice, resources, and functional ability (Bangerter et al., 2016). Some resident preferences might be situational and their rating of importance might change if residents feel they have more of a voice in decision-making and/or specific barriers are understood. For example, some preferences such as desiring choice in roommates may be misinterpreted by staff as an issue of privacy when residents are actually interested in having a voice in decision-making. This highlights the potential tension between resident preferences and the economic pressures/competing demands nursing home staff and administrators face (e.g., roommate assignments may be made by NHs to maximize occupancy rather than resident preference; Jaen et al., 1994; Krist et al., 2017). Improved communication throughout the person-centered care planning process while acknowledging the complex nature of the nursing home work environment and potential limitations to staff time and resources is needed (Abbott et al., 2018). These insights are critically important toward balancing the need to incorporate person-centered care with the need for the use of evidence-based medicine as these two approaches are not always easily reconciled (Siminoff, 2013).
This study was not without limitations. Generalization of findings is limited by the use of a highly specific sample (i.e., Veterans living with HF recently admitted to community nursing homes); however, the large sample size provides a well-founded start into the exploration of reporting source and preference assessment. Future studies with larger, more diverse samples may provide further insight into proxy and staff understanding of residents’ preferences. Another limitation relates to how the assessments (resident, proxy, staff) were completed at admission to the nursing home. Preference assessments (MDS Section F) are required to be completed annually, so more research is needed on whether staff, proxy, and resident agreement improves over time as staff gain familiarity with residents and residents adapt to living in a nursing home. Other research suggests residents’ preferences stay consistent for at least 3 months’ time (Abbott, et al., 2018) but more research is needed to confirm these findings and see how they extend to proxy reporters. Assessments were also not completed for the same resident–primary studies focused on understanding the differences between staff, proxy, and resident report for the same individual would be useful to better understand the nuance between preference importance and reporting source. An additional limitation includes the differences in cognitive functioning and ability of staff, proxies, and residents which we cannot account for in this study but may have influenced both if and how proxies or staff members completed the preferences survey. Moreover, whereas statistically significant differences were found for some of these demographic characteristics, not all were practically meaningful. For example, the differences in length of stay were about 1 day and the differences in age were one 5-year interval. Interpretation of the demographics table should be done with caution when looking for differences between these three groups.
Conclusion and Implications
Findings from this study emphasize the need to align nursing home care delivery with residents’ preferences. To account for those who cannot accurately or effectively report their important preferences, proxies and staff members are employed to report important preferences on the resident’s behalf; however, it is likely that proxies and staff are misrepresenting, typically underreporting, residents’ important preferences. Future studies should focus on using objective measures of preference ratings and engagement, strengthening proxy and staff understanding of residents’ preferences, and exploring proxy and staff understanding of change related to preferences over time. In the meantime, these findings can help family members reflect on areas they might be underreporting importance or need to think more critically about/survey other family members or friends to get a more accurate profile of the residents’ preferences. Findings are also relevant for staff who need to consider the possibility that they are underreporting the importance of residents’ preferences and need to use strategies to improve preference identification such as talking to family, observing residents’ behavioral responses and affect when presented with choices, and asking residents yes or no questions related to their preferences. Honoring preferences is an essential component of person-centered care, and this work is a first step toward improving our understanding and use of current assessments and preference reporting sources.
Supplemental Material
Supplemental Material - Comparing Resident, Proxy, and Staff Respondents for Nursing Home Residents’ Preferences for Everyday Living
Supplemental Material for Comparing Resident, Proxy, and Staff Respondents for Nursing Home Residents’ Preferences for Everyday Living by Zachary J. Kunicki, Caroline Madrigal, Lien T. Quach, Melissa R. Riester, Lan Jiang, Matthew S. Duprey, Melanie Bozzay, Andrew R. Zullo, Mriganka Singh, John McGeary, Wen-Chih Wu, and James L. Rudolph in Journal of Applied Gerontology
Footnotes
Acknowledgments
Dr. Madrigal acknowledges support from the VA’s Office of Academic Affiliation for her Advanced Fellowship in Health Services Research. Dr. Wu acknowledges support from two SMA grants, the Group Medical Visits in Heart Failure for Post-Hospitalization Follow-Up Award (5I01HX001800-05), and the Implementation Trial Evaluating On-site In-person Versus Remote Video-Assisted Facilitation to Train Providers on the Implementation (IRP 20-003).
Declaration of conflicting interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by the This work was supported by the VA Health Services Research and Development Center of Innovation in Long Term Services and Supports (CIN 13-419 and C19 20-213), the VA QUERI-Geriatrics and Extended Care Partnered Evaluation Center for Community Nursing Homes (PEC 15-465). The funders did not play a role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication. The authors retained full independence in the conduct and communication of this research. The statements and opinions expressed are those of the authors and do not represent the official policy or procedures of the United States Government or the Department of Veterans Affairs.
IRB Protocol
This study was approved by the Institutional Review Board at the Providence VA Medical Center, approval #1633554.
Supplemental Material
Supplemental material for this article is available online.
References
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.
