Familial caregiving following stroke: findings from the comprehensive post-acute stroke services (COMPASS) pragmatic cluster-randomized transitional care study

ABSTRACT Background Stroke patients discharged home often require prolonged assistance from caregivers. Little is known about the real-world effectiveness of a comprehensive stroke transitional care intervention on relieving caregiver strain. Objectives To describe the effect of the COMPASS transitional care (COMPASS-TC) intervention on caregiver strain and characterize the types, duration, and intensity of caregiving. Methods The cluster-randomized COMPASS pragmatic trial evaluated the effectiveness of COMPASS-TC versus usual care with patients with mild stroke and TIA at 40 hospitals in North Carolina, USA. Of 5882 patients enrolled, 4208 (71%) identified a familial caregiver. A follow-up Caregiver Questionnaire, including the Modified Caregiver Strain Index, was administered at approximately three months post-discharge. Demographics and frequency, duration, and intensity of caregiving were compared between groups. Results 1228 caregivers (29%) completed the questionnaire. Completion was positively associated with older patient age, white race, and spousal relationship. One-third of the caregivers provided ≥30 hours of care per week and 889 (79%) provided care ≥9 weeks. Average standardized caregiver strain was 21.9 (0–100), increasing with stroke severity and comorbidity burden. Women caregivers reported higher strain than men. Treatment allocation was not associated with caregiver strain. Conclusions This sample of mild stroke and TIA survivors received significant assistance from familial caregivers. However, caregiver strain was relatively low. Findings support the importance of familial caregiving in stroke, the continued disproportionate burden on women within the family, and the need for future research on caregiver support.


Introduction
Stroke patients discharged home often require prolonged assistance from caregivers. However, sustained caregiving can result in an increased risk for depression, anxiety, and social isolation, negatively affecting well-being. 1,2 Caregivers want to feel supported by health care providers, be prepared to provide the necessary care, and expect the health care system to be accountable for coordination and follow-up care. [3][4][5][6] Studies have found that transitional care (TC) interventions for caregivers and stroke survivors leaving the hospital can improve patient and caregiver outcomes, especially when the interventions are tailored to the specific needs of each member of the dyad. 1,[7][8][9][10] Caregiving is particularly important for stroke survivors who are discharged directly home from acute care. A large proportion of those patients experience significant mobility impairments, which limit their ability for independent living. 11 Globally, 50% of survivors experience chronic disabilities post-stroke. 12 Given the significant impact of stroke on global public health and the strain on caregivers, interventions to address the needs of both stroke survivors and caregivers are needed. 1,10 The needs of family members caring for stroke survivors differ from those of other caregivers in that stroke occurs unexpectedly, requiring that family members assume the caregiving role suddenly. Depending on the needs of the stroke survivor, caregivers may assume a range of responsibilities, from providing minimal assistance with instrumental activities of daily living (IADLs) to significant assistance with IADLs and basic activities of daily living (ADLs). They may also have to assume the roles and responsibilities that the stroke survivor performed prior to the stroke event. 13,14 Addressing the post-discharge needs of stroke survivors is often complex. In addition to the traditional TC management strategies suggested by Coleman et al and others, [15][16][17][18][19][20] stroke patients may need referrals to community-based resources and outpatient therapies. Further, family caregivers of stroke survivors typically need support and education to become better prepared to assist with ADLs and IADLs at home. Little is currently known about the real-world effectiveness of a comprehensive stroke TC intervention on relieving caregiver strain, and previous studies have demonstrated mixed results. 1,21 In order to improve post-discharge outcomes, the Centers for Medicare and Medicaid Services (CMS) implemented a TC management (TCM) reimbursement protocol in the US in 2013. Use of TCM billing codes requires patient contact via phone, e-mail, or in person within two days of discharge and an in-person visit by an advanced practice provider (nurse practitioner or physician's assistant) or physician within 7 to 14 days post-discharge. Services supported by TCM codes include assessing the need for diagnostic tests, the patient's functional status, caregiving support, and establishing and/or reestablishing treatment including rehabilitation and community services. TCM providers are expected to be aware of community services and communicate and coordinate with the agencies, as needed. The primary goal of TCM is to coordinate patient care among providers for up to 30 days post-discharge to reduce preventable readmissions. 22 COMPASS-TC was designed to provide TCM in a way that is consistent with CMS billing requirements and that comprehensively addresses both patient and caregiver needs.

The COMPASS Study
The COMPASS study was a large clusterrandomized controlled pragmatic trial in 40 North Carolina hospitals. The first large-scale, multicenter pragmatic trial of a comprehensive post-acute stroke TC model in the US, COMPASS tested effects of implementing an evidence-based post-acute stroke TC program, COMPASS-TC, on patient and caregiver outcomes in real-world settings in a population of mild stroke and transient ischemic attack (TIA) survivors discharged directly home. 23 COMPASS-TC was designed with stakeholder input specifically to address the gaps in postdischarge TC for patients with stroke and their caregivers. The intervention is based on the early supported discharge (ESD) model that has been widely implemented in the UK and European countries. In a 2012 Cochrane review of 14 trials in Europe (N = 1,957), patients with mild to moderate stroke severity who received ESD had decreased hospital length of stay, accelerated functional recovery, improved patient/caregiver satisfaction, avoided long-term nursing home placement, and reduced hospital readmission with no adverse events. [24][25][26] Some studies have also found that caregiver burden was lower with ESD. 27,28 ESD is now the standard of care in the UK 26 and Canada. 6,29 The COMPASS study design and methods have been published. 30,31 Briefly, in this pragmatic cluster randomized controlled trial 40 hospital units (2 hospitals functioned as one administrative unit) were randomized to implement COMPASS-TC (N = 20) or maintain their usual care (N = 20) for patients with stroke and TIA discharged home. Hospitals were stratified by number of stroke patients discharged annually and stroke center certification status, and were randomized centrally by a study statistician. 23,31 The COMPASS-TC model COMPASS-TC included a 2-day follow-up telephone call and a clinic visit within 30 days. Details on the full intervention model have been published. 31,32 During the clinic visit, eligible patients discharged from intervention hospitals and their caregivers received standardized comprehensive assessments, which included a 10-item caregiver assessment evaluating the patient's need for caregiving as well as caregiver availability, assistance provided, general health, and stress and strain. 31 The findings from the comprehensive assessment were used to develop a tailored care plan with referrals to targeted existing community resources, including caregiver-specific support services from the local Area Agency on Aging (e.g. home delivered meals, support groups, and in-home assistance); outpatient rehabilitation services; and community pharmacy assistance to help with medication management. For example, if a caregiver indicated on the assessment they needed meal support, a referral was made to the local home-delivered meals program. In this way, the COMPASS-TC care model and individualized care plan were designed to identify and respond to the specific needs of both patients and caregivers. Usual care hospitals continued their current standard of post-acute care.
Patients and caregivers in the intervention group received additional telephone follow-up calls at 30 and 60 days' post-discharge. Caregivers of patients discharged from hospitals in the usual care group were contacted only at 95-110 days after patient discharge home and did not receive the tailored COMPASS-TC care plan with referrals to existing support services that were provided to the caregivers of patients from the intervention hospitals. A key, prespecified secondary outcome in the COMPASS trial was caregiver strain. The purpose of this study was to address the secondary outcome of the COMPASS trial by describing the effect of COMPASS-TC on caregiver strain and to characterize the types, duration, and intensity of caregiving in the study sample.

Participants
The study population consisted of all COMPASS study patient-caregiver dyads for which an unpaid caregiver who provided assistance to the patient after returning home was identified. Patients were eligible if they were English-or Spanish-speaking, 18 years of age or older, and were discharged directly home following a stroke or TIA. Patients with subarachnoid or aneurysmal stroke were excluded because these types of stroke are generally more severe and require facility-based post-acute care. The intervention was designed for stroke survivors with TIA or mild/moderate stroke. At baseline enrollment, prior to hospital discharge, patients were asked to identify whether they had a primary caregiver (unpaid family member or friend) who would assist them with activities upon returning home. Caregiver contact information, gender and relationship to the patient were collected by trained post-acute nurse coordinators. Additional information about the patient's stroke severity, medical history and demographics was also captured.

Data collection
Patient outcomes data were collected at 90-days post-discharge using a standardized tool administered by blinded telephone interviewers. 32 During this interview, patients were asked to confirm the name of their primary familial caregiver and identified caregivers were mailed a paper follow-up Caregiver Questionnaire at approximately 95-110 days after the patient was discharged from the hospital. Caregivers who did not respond to the first mailed questionnaire were contacted by telephone and sent an additional paper questionnaire.

Instruments
The follow-up Caregiver Questionnaire was developed specifically for this study and captured caregiver demographic information, the 13-item Modified Caregiver Strain Index (MCSI), 33,34 and types, duration, and intensity of caregiver assistance with ADLs and IADLs. The follow-up Caregiver Questionnaire was amended in May 2017, approximately seven months into data collection, after receiving feedback from caregiver participants, and in an effort to increase the response rate. Version 2 included all items in Version 1 but also asked caregivers if they had provided assistance to the patient prior to the stroke or TIA hospitalization. Table 1 includes the categories and specific types of items included in the follow-up Caregiver Questionnaire. See the Supplementary Materials, Figure 2 for the complete follow-up Caregiver Questionnaire, Version 2.
The MCSI, a 13-item, validated, evidence-based tool used to assess caregiver strain, has been used in stroke [35][36][37] and caregiver research. 34,38,39 The modified version has an internal reliability of a = 0.90 vs. a = 0.86 for the original and a test-retest reliability of 0.88. 33,40 In a study of caregivers of stroke survivors, the MCSI had test-retest reliability of 0.93. 41 It is designed to assess several domains of strain including financial, physical, psychological, and social. Sample items include: "My sleep is disturbed" and "Caregiving is a physical strain." Responses range from 0 to 2, (No = 0, Sometimes = 1, Yes = 2), with a total score range of 0 to 26. Higher scores indicate higher strain. 33,40

Ethical considerations
Institutional review board (IRB) approval was granted through Wake Forest University Health Sciences (central IRB) or through local hospital IRBs. The study met criteria for a waiver of consent and HIPAA authorization; therefore, eligible patients were enrolled at hospital discharge without consent. At 90 days postdischarge, patients or their proxies provided verbal informed consent over the telephone for collection of outcomes. 42 The trial was periodically reviewed by an independent Data and Safety Monitoring Board.

Data analysis
Analyses were conducted using SAS 9.4 and in accordance with a pre-specified statistical analysis plan, which has been published. 23 Baseline characteristics of patient-caregiver dyads were summarized by treatment group (i.e. intervention and usual care). Descriptive statistics were calculated for type, duration, and intensity of caregiving tasks. Analyses of the effect of COMPASS-TC were performed as intention-totreat (ITT). To best represent the underlying construct of caregiver burden within the multidimensional MCSI scale and account for itemlevel missingness, MCSI scores were standardized to range from 0 to 100, with higher scores representing greater strain. Questionnaires were scored if at least 10 of the 13 items were answered. 43 Additional details have been published in Section 5.2.1 of the Statistical Analysis Plan. 23 Mean difference in caregiver strain (Intervention versus Usual Care) and 95% confidence intervals (CI) were estimated using weighted linear mixed models that included a hospital-specific random effect and adjusted for patient age, race, NIH Stroke Severity (NIHSS) score, diagnosis (stroke, TIA), caregiver-patient relationship, and randomization stratum. Some caregivers did not provide outcomes. To address potential selection bias resulting from incomplete outcome ascertainment, inverse probability weights were estimated using conditional logistic regression and included in the linear mixed model. 23,[44][45][46] Additional details have been published. 23,47  Caregivers were defined as unpaid family members or friends who assisted the patient with ADLs and IADLs. The majority were women (68%) and spouses (53%). Mean age for the Intervention group was 64.2 (standard deviation [SD] 13.2) years and Usual Care 63.1 (SD 13.7) years. Table 2 describes caregiver characteristics. Of caregivers in the Intervention group (n = 1722), 20% (n = 353) had a completed caregiver assessment and received a tailored care plan during the in clinic visit.

Outcomes
Of the 4208 caregivers identified at enrollment, 1228 completed the follow-up Caregiver Questionnaire (29% overall; 33% Intervention and 26.5% Usual Care). Questionnaire completion was positively associated with older patient age, male patient sex, white race, and spousal relationship with the patient. Caregivers who completed the follow-up Caregiver Questionnaire reported extended durations of caregiving post-stroke, with 79% providing care for 9 weeks or longer, and 33% for 30 hours or more per week. On average, caregivers reported assisting with 1.0 (SD 1.5) ADL and 6.3 (SD 3.3) IADLs. For both groups, providing assistance with dressing was the most frequent ADL and assisting with transportation, shopping, preparing meals, and housework were the most frequent IADLs reported by caregivers. Only 10% of caregivers reported being in excellent health. Table 3 describes caregiving type, duration, and intensity of caregiving activities and caregiver selfrated health.
The average standardized caregiver strain was 21.9 (SD 23.3) (range: 0-100) and was not significantly different between Intervention 21.9 (SD 23.5) and Usual Care 21.9 (SD 23.1). Characteristics associated with higher strain were higher patient stroke severity and patient comorbidities. Female caregivers reported higher strain than males. In ITT analysis, treatment group was not associated with decreased strain (adjusted mean difference 0.65; 95% CI −1.99, 3.29). Fifty-one percent of caregivers in both groups responded "yes" or "sometimes" to the MCSI item on feeling overwhelmed.
As a pragmatic trial, this study was designed to incorporate stakeholder-driven refinements throughout the course of the trial. 31,48 As such, based on caregiver feedback on the first version of the followup Caregiver Questionnaire, we revised it to include questions about whether the caregiver had been providing assistance with ADLs and IADLs tasks before as well as after the stroke event. Of 1228 caregivers, 771 (62.8%) completed Version 2 of the follow-up Caregiver Questionnaire, and 457 (37.3%) completed Version 1. Table 4 and Figures 1 and 2 compare ADLs and IADLs assistance provided before and after the stroke. Of the 771 caregivers responding to Version 2, 19.7% to 26.9% assisted with each ADL pre-stroke,   and 42.5% to 66.4% assisted with each IADL prestroke. The average number of ADLs provided prestroke was 0.4 (SD 1.1) and IADLs was 4.4 (SD 3.6); post-stroke these numbers were 1 (SD 1.5) ADL and 6 (SD 3.4) IADLs. Of note, more than 70% of caregivers assisting with ADLs post-stroke had not assisted with ADLs pre-stroke. For eight of the IADLs, caregivers were more likely to have provided assistance with each task before and after the stroke. The only IADL that was new to approximately 60% of the caregivers was medication management, indicating that this IADL and the five ADLs tasks were novel to most of the caregivers post-stroke.

Discussion
Caring for a stroke survivor has been described as "a full-time job" by stroke caregivers.' 2 It is a life changing, long-term event for many, which can create fear about the potential for another, more severe stroke, and uncertainty about the impact of stroke on the lives of both the caregiver and the stroke survivor. 2,14,49 It can also cause sudden and unexpected changes in the stroke survivor's relationship with a spouse and other family members. 2,50,51 New or increased duties associated with the caregiving role, especially those that involve both physical and psychological demands such as  assisting with ADLs and medication management, can be overwhelming and impact physical, emotional, and mental health, 2,52,53 leaving family caregivers of stroke survivors at higher risk for depression, anxiety, social isolation, and reduced quality of life. 2,54-57 Caregiving can have a cumulative negative effect on caregiver well-being, especially for caregivers who have other existing responsibilities and changes in lifestyle, such as an unexpected exit from the workforce, reduction in socialization, and changes in future plans. 58 In this pragmatic study of survivors with mild stroke and TIA, COMPASS-TC was not associated with caregiver strain, even though one-third of the caregivers reported providing more than 30 hours of care per week. Caregivers primarily assisted with IADLs, with an average of 6.9 IADLs per week poststroke. While they only minimally assisted with ADLs post-stroke, with an average of 1 ADL weekly, more than 70% reported that these tasks were new post-stroke. Additionally, 79% reported providing assistance for nine weeks or longer. While we did not observe an association between the type, duration, and intensity of caregiving and caregiver strain, the high number of hours of care provided weekly, the length of time providing care, and the response on the MCSI that more than 50% of the caregivers felt overwhelmed at least some of the time indicates that caregivers are shouldering increased responsibilities, putting them at risk for negative outcomes over time.
Research suggests that the more dependent the stroke survivor is on the caregiver, the greater the caregiver strain. 59,60 Caregiver stress and well-being are associated with stroke survivor functional status, such as low physical, cognitive, behavioral, and emotional function. 56,[59][60][61][62] Given that the stroke survivors in this study primarily experienced mild strokes and TIAs, it is not surprising that their caregivers reported assisting with more IADLs than ADLs. In most cases the IADLs tasks they reported performing (e.g. meal preparation, shopping, and transportation) were tasks that they had been performing prior to the stroke. The only IADL that more caregivers reported performing after the stroke was medication management. This is consistent with previous reports on the stroke survivor participants in the COMPASS study, where 38% had cognitive impairments and 73% had polypharmacy. 63 These findings lead us to question how well MCSI scores represent the types, intensity, and duration of tasks performed by caregivers in this study. Did the caregivers view these tasks as part of their role as the spouse, partner, or child of the stroke survivor? Are there other tools that are more discriminant in measuring caregiver strain? Our findings are congruent with the meta-analysis by Kokorelias et al., 2 indicating that even in the case of mild stroke and TIA caregiving is a "full-time job" suggesting caregivers may require additional support and resources, especially if they have to continue in this role for the foreseeable future.

Strengths and Limitations
This study has notable strengths. COMPASS was the first large-scale, multicenter pragmatic trial of a comprehensive post-acute stroke TC model, designed with stroke survivor and caregiver input. The COMPASS research team included a stakeholder advisory board with both stroke survivors and caregivers. 48 These stakeholders helped the research team develop and refine the intervention and the follow-up Caregiver Questionnaire. COMPASS-TC was implemented in a wide variety of communities, from urban areas with broad community resources to small rural communities. 64,65 This study also has limitations. One of the main findings from the highly pragmatic COMPASS trial was that CMS policies and reimbursements are not sufficient incentives to drive TC uptake. 31 Only 35% (n = 928) of eligible stroke survivors completed the intervention protocol, and approximately three-quarters of those identified a family caregiver. 23 While poor intervention uptake and low rates of outcome capture may be considered weaknesses in explanatory randomized clinical trials, pragmatic trials, such as COMPASS, are explicitly designed to give intervention sites flexibility in the way an intervention is implemented, within real-world settings and among real-world patient populations. 23,66 As such, patients and sites that are not fully adherent are not excluded.
Further, financial assistance for implementation is not provided, nor incentives for caregiver survey completion. In keeping with its highly pragmatic design, COMPASS provided a realistic measure of TC uptake and its effectiveness for stroke and TIA survivors. 23 These factors may have contributed to a low caregiver response rate and an inability to demonstrate significant differences between groups. The low response rate may have also led to non-response bias, as healthier and/or more capable or less overburdened caregivers may have been more likely to respond to the questionnaire. Providing incentives to caregivers may have improved response rates. The data may also be subject to recall bias, as caregivers may not remember how much caregiving they were providing at the beginning of the stroke survivor's recovery. Further, COMPASS-TC was designed to address the needs of mild stroke or TIA survivors discharged home; therefore, results are not generalizable to all stroke survivor-caregiver dyads. As this was a pragmatic trial, data collection was limited to what was practically feasible. In this case, follow-up with caregivers was limited to the follow-up Caregiver Questionnaire mailed approximately three months post-discharge.

Implications for rehabilitation and future research
Although we did not find a difference in caregiver strain associated with COMPASS-TC, differences may become more evident as the caregiver sustains the role over time, particularly when providing care for stroke survivors with more extensive functional limitations or as their functional status declines over time. These factors and recent recommendations related to supporting caregivers, suggest that reevaluating the intensity and types of caregiving assistance, the impact of caregiving, and the changing needs of caregivers over the caregiving trajectory are important gaps in research and practice. 2,67,68 Future research should focus on conducting a comprehensive assessment of the caregiver's needs at regular intervals post-discharge, tailoring care plans to address needs as they arise, and developing standardized measures that are sensitive to the impact of caregiving over time. Regular assessment for the need for community-based caregiver support services should also be considered. 14,6814 Future studies should assess which elements of both caregiver and stroke survivor interventions are most useful to caregivers, and what caregiver-survivor dyad elements influence the impact of comprehensive interventions. Importantly, our results suggest that healthcare system-level changes are not sufficient to alter the strain on familial caregivers.

Conclusion
In this study, COMPASS-TC was not associated with caregiver strain. One-third of caregivers of mild stroke and TIA survivors provided more than 30 hours of assistance per week in the first 90 days after discharge from the hospital, and 79% reported providing care for nine weeks or more. Overall caregiver strain remained relatively low during this time period; however, the impact of this caregiver engagement with stroke survivors may not be reflected in standardized assessments of caregiver strain.