BABEL Advance Care Planning in Long-term Care

1.0 SUMMARY Most Canadian nursing home (NH) residents are elderly and frail, have multiple chronic health conditions and impairments, and have dementia. In 2014, 244,000 Canadians lived in NHs, including 6% of those ≥65 y.o., at a cost of \>$10 billion/yr. NH residents experience high rates of acute illness; approximately 33% have emergency department (ED) visits and 23% are hospitalized yearly. Many of these visits are avoidable, and expose residents to iatrogenic complications. In Manitoba \>1.5% of NH residents are admitted to intensive care units yearly, where they receive highly aggressive care. Approximately 30-50% of NH residents die each year, experiencing a progressive burden of severe symptoms leading up to death. Thus, there are serious concerns about Advance Care Planning (ACP) and end-of-life (EOL) care in NHs. Canadians in general have mediocre knowledge of, and engagement in ACP. Also, studies show that values such as quality of life and aversion to being dependent trump survival in determining care preferences. Among hospitalized octogenarians, 61% desired comfort care only, or just a brief trial of aggressive care. A U.S. study found that decisions for LTC residents to be sent to ED were frequently driven by families who felt unprepared for their loved ones' death, and insecure about the quality of NH care, where there had been little or no discussion about ACP. Systematic approaches to ACP in NHs have demonstrated benefits, including: increases in ACP uptake, higher compliance with EOL wishes, higher satisfaction with care and emotional well-being, reduced family stress and anxiety, and lower rates of hospitalization. Generally, multimodal ACP interventions have shown the most benefits. Thus, ACP can improve outcomes for NH residents, their families, and society. The goal of this proposal is to apply best practices in ACP, and demonstrate that it can be implemented it in a scalable, sustainable way across provinces. This will result from delivering the ACP intervention within the existing envelope of NH staffing, and by acquiring most of the data from the Resident Assessment Instrument (RAI), which is completed quarterly for NH residents in 9 provinces. As RAI contains information identifying NH residents at the highest risk for dying within 6-12 months, it will be used to target the ACP intervention to such individuals.

2.0 SPECIFIC AIMS 2.1. Aim 1. Improve intervention acceptability, buy-in, and usability by engaging a diverse group of stakeholders to collaboratively design key details of the study. * Aim 1A: The stakeholder group work to identify and agree upon key principles and practical aspects of ACP in the NH setting. * Aim 1B: Using the key principles and practical aspects, and existing evidence about beneficial approaches to ACP in NHs, the investigator group will devise an evidence-based, integrated ACP intervention that can be applied by existing NH personnel. * Aim 1C: Guided by the results of the stakeholder meeting, the investigator group will identify characteristics of NH residents -- available from RAI data routinely and repeatedly available in Canadian NHs -- to identify a cohort of residents who are at the highest risk of dying in the following 6-12 months. * Aim 1D: Informed by the results of the stakeholder meeting, and existing data, the investigator group, will agree upon the outcomes to be assessed in this study. 2.2. Aim 2. Design and implement training of NH personnel in application of the systematic approach to ACP devised in Aim#1B. 2.3. Aim 3. For each participating NH, devise an approach for real-time access to, and use of, the RAI data for residents. 2.4. Aim 4. Applying the systematic approach to ACP devised in Aim#1B, to high-risk NH residents identified in Aim#1C, will result in improvement in relevant outcomes identified in Aim#1D. 3.0 METHODOLOGICAL APPROACH 3.1. Aim 1A. During a 1 day meeting in Toronto, the investigators will convene stakeholders including: investigators, NH residents and families, advocacy groups for the elderly, legal experts, NH management, NH staff, an expert on inter-professional care, NH physician, geriatrician, palliative care physician, ethicist, ICU physician, and dietitian. Directed discussions and breakout sessions will be utilized to identify key principles and practical components of ACP in the NH setting. All sessions will be recorded. 3.2. Aim 1B. After the stakeholder meeting, the investigator group will conduct weekly telephone conferences to design an integrated, systematic, evidence-based ACP intervention -- which will be referred to as "The BABEL Approach to ACP". It will be informed by both: (a) the stakeholder meeting, and (b) existing, published evidence about approaches to ACP in the long-term care setting that have been demonstrated to improve relevant outcomes. These latter include, but are not limited to: (i) "Let Me Decide", (ii) the methods of Morrison et al., and (iii) "Respecting Choices". Criteria will include: likelihood to improve relevant outcomes, practicability, feasibility, scalability, sustainability and general acceptability. 3.3 Aim 1C. This aim will utilize national RAI data on NH residents, maintained at the University of Waterloo by Dr. John Hirdes. Preliminary work in this direction indicates that 44% of all NH residents in the country have one or more of the following, high-risk characteristics, all available from the RAI data: CHESS score ≥3, cancer, congestive heart failure, leave \>25% of their food uneaten. 3.4. Aim 1D. There will be discussion during the stakeholder meeting, specifically addressing which potential outcomes are relevant and should be tracked. Informed by those discussions, and outcomes studied in prior studies of ACP in long-term care, the investigator group will compile a final list of outcomes for the study. 3.5. Aim 2. After design of The BABEL Approach to ACP, the investigator group will continue weekly conference calls to develop a detailed plan to train NH staff in implementing it. The investigators and study coordinators in each province will provide the training, which is expected to include (but may not be limited to) applied learning, intentional case studies, role-playing exercises, targeted discussions and adjunctive print, video, and web-based materials. Criteria for the training methods include: practicability, feasibility, acceptability, scalability, and sustainability. Training will occur in the NHs, and will not include actual residents. 3.6. Aim 3. Local investigators and study coordinators will visit each participating NH, and working with NH personnel will: (i) locate how the NH staff conduct the quarterly RAI surveys, (ii) locate where the data is maintained in the NH, and (iii) devise a method, with the direct assistance of NH personnel, for using the RAI data to identify residents who meet targeting criteria (Aim#1C). 3.7. Aim 4. 3.7.A. GENERAL METHODS: This will be a cluster-randomized study in 24 NHs in Ontario, Manitoba, and Alberta, divided equally between intervention and control (total counts are: Ontario=13, Alberta=6, Manitoba=5). Informed consent will be obtained from cognitively intact residents, and from their substitute decision-maker (SDM). 3.7.B. ACP IN CONTROL NURSING HOMES Eligible residents in the 12 control NHs will receive the prevalent approach to ACP. No elements of The BABEL Approach to ACP will be introduced in the control homes. 3.7.C. ACP INTERVENTION NURSING HOMES The intervention comprises: (i) training on, and introduction of The BABEL Approach to ACP for targeted NH residents, and (ii) after these ACP discussions occur, notifying the resident's primary care physician of the residents' ACP preferences, as elicited by the ACP. Intervention homes will be asked to use the systematic, evidence-based approach to ACP for targeted residents. They will be free to use it in non-targeted residents, but it will not be specifically requested that they do so. 3.7.D. INFORMED CONSENT In both intervention and control homes informed consent will be sought from eligible residents (or surrogate consent from SDMs, for residents deemed to lack capacity to provide informed consent), and separately from SDMs. As the ACP approach of our intervention is not investigational, the investigators believe it does not require informed consent, per se. However, informed consent from residents is required to obtain their personal information, track the course of their ACP wishes/decisions, obtain their clinical outcomes, and include their data in group form in analyses, reports and manuscripts from the research study. Residents' clinical outcome information will be obtained from a combination of: their nursing home records, their physician, emergency departments, urgent care, and hospitals. A separate page on the consent form serves specifically as consent to obtain such medical information. Informed consent from SDMs is required to contact them separately in order to send them a survey assessing their satisfaction with care at end-of-life provided to the NH resident. 3.7.E. ANALYSIS OF RESIDENT OUTCOMES Most of these analyses will use hierarchical (2-level) random effects models to account for clustering within NHs, and possible differences between intervention and control groups that might occur, despite randomization. The provinces will be included as fixed effects; interaction terms between province and Intervention/Control will be included to account for possible influences of existing approaches to ACP undertaken by provinces (e.g. PoET in Ontario). Logistic regression will be used for binary outcomes, Poisson regression for rates, OLS regression for continuous variables. The time-to-event analysis will use Cox proportional hazards regression with fixed effects to represent the nursing homes. Resident-level covariates, obtained from NH records and RAI will be: age at NH admission (stratified), gender, marital status at study entry, elapsed time from NH admission to study entry (stratified), location immediately prior to NH entry (as home, assisted living, other LTC facility, hospital or other acute care facility, Chronic Continuing Care facility \[in ON only\]), medical conditions/diagnoses, CHESS score at study entry, whether at study entry resident was leaving ≥25% of NH food uneaten, and use of specific medications at study entry. Nursing home-level covariates will be: # of beds (stratified), ownership (public vs. private), and elements of pre-existing ACP approaches as derived from the Nursing Home Environmental Scan (specific variables still to be determined, but will do so before doing any analysis). 3.7.F. ANALYSIS OF SDM OUTCOMES Analysis will use hierarchical (2-level) mixed models to account for clustering within NHs, and possible differences between intervention and control groups that might occur, despite randomization. The provinces will be included as fixed effects; interaction terms between province and Intervention/Control will be included to account for possible influences of existing provincial approaches to ACP (e.g. PoET in Ontario). Covariates, obtained from the SDM participants, NH records, and RAI will include all of the covariates from Section 3.7.E, and in addition: SDM age at resident death (stratified), SDM gender, and SDM relationship to deceased resident. 3.7.G. ANALYSIS OF NH STAFF OUTCOMES Analysis of NH staff self-efficacy in ACP will compare the pre-study to post-study difference in self-efficacy ratings between intervention and control NHs. Analysis, in intervention NHs only, of NH staff perceptions of The BABEL Approach to ACP, and to the training in it's use will be mainly descriptive. Analysis of the semi-structured interview of NH personnel will involve thematic analysis, compared between Control and Intervention homes. 3.7.H. ANALYSIS OF PROCESS OUTCOME OF THE RATE OF DOCUMENTED ACP DISCUSSIONS Analysis will use the hierarchical (2-level) random effects Poisson modeling strategy of section 3.7.E. 3.7.I. SAMPLE SIZE CALCULATION * For this calculation, use was made of PASS 2008 sample size software (NCSS, LLC, Kaysville, Utah). Specifically, the calculation for Two Independent Means with a cluster-randomized design. * This calculation was for the co-primary outcome of the ACP Audit survey given 6 weeks after study entry, specifically the fraction of items #1-7 answered 'Yes'. Thus, the scale goes from from 0-7. * It used the following assumptions: • Mean±SD score: * In Control homes = 5.0±1.5 (72% 'yes') (33, 34) * In Intervention homes = 5.6±1.5 (85% 'yes') * 5% Type I error rate * Average of 150 residents in each NH, of whom 0.5% are in respite care (2) * 44% of NH residents are at high risk of death or other acute outcomes, defined as any of: CHESS ≥ 3, cancer, congestive heart failure, or leave \>25% of their food uneaten. (personal communication, Dr. John Hirdes, University of Waterloo) * 75% rate of consent among eligible NH residents * Equal proportion of NHs in intervention and control groups * Intraclass correlation coefficient (ICC) for the primary outcome = 20% Using these assumptions, 12 NHs in each group (24 total) gives power of 83% (total N=1164).