BMJ Open Is Committed to Open Peer Review. As Part of This Commitment We Make the Peer Review History of Every Article We Publish Publicly Available

BMJ Open: first published as 10.1136/bmjopen-2019-036631 on 23 April 2020. Downloaded from

BMJ Open is committed to open peer review. As part of this commitment we make the peer review history of every article we publish publicly available.

When an article is published we post the peer reviewers’ comments and the authors’ responses online. We also post the versions of the paper that were used during peer review. These are the versions that the peer review comments apply to.

The versions of the paper that follow are the versions that were submitted during the peer review process. They are not the versions of record or the final published versions. They should not be cited or distributed as the published version of this manuscript.

BMJ Open is an open access journal and the full, final, typeset and author-corrected version of record of the manuscript is available on our site with no access controls, subscription charges or pay-per-view fees (http://bmjopen.bmj.com).

If you have any questions on BMJ Open’s open peer review process please email [email protected] http://bmjopen.bmj.com/ on September 28, 2021 by guest. Protected copyright. BMJ Open BMJ Open: first published as 10.1136/bmjopen-2019-036631 on 23 April 2020. Downloaded from

Dyadic approach to supervised community rehabilitation participation in an Asian setting post-stroke: exploring role of caregiver and patient characteristics in a prospective cohort study

ForJournal: peerBMJ Open review only Manuscript ID bmjopen-2019-036631

Article Type: Original research

Date Submitted by the 26-Dec-2019 Author:

Complete List of Authors: Tyagi, Shilpa; National University Singapore Saw Swee Hock School of Public Health Koh, Gerald; National University of Singapore, Saw Swee Hock School of Public Health Luo, Nan; National University Singapore Saw Swee Hock School of Public Health, Tan, Kelvin; Government of Singapore Ministry of Health, Policy Research & Evaluation Division Hoenig, Helen; Durham VA Medical Center, Physical Medicine and Rehabilitation Service

Matchar, David; Duke University, Internal Medicine; Duke-NUS Medical http://bmjopen.bmj.com/ School, Health Services and Systems Research Yoong, Joanne; National University Singapore Saw Swee Hock School of Public Health Chan, Angelique; Duke-NUS Graduate Medical School, Centre for Ageing Research and Education Lee, Kim En; Lee Kim En Neurology Pte Ltd Venketasubramanian, Narayanaswamy; Raffles Hospital, Raffles Neuroscience Centre

Menon, Edward; St. Andrew’s Community Hospital on September 28, 2021 by guest. Protected copyright. Chan, Kin Ming; Mount Alvernia Hospital De Silva, Deidre Anne; National Neuroscience Institute - Singapore General Hospital Campus Yap, Philip; Khoo Teck Puat Hospital, Tan , Boon Yeow; St. Luke's Hospital, Singapore Chew, Effie; National University Hospital, Department of Rehabilitation Medicine Young, Sherry H.; Changi General Hospital, Department of Rehabilitation Medicine Ng, Yee Sien; Singapore General Hospital, Department of Rehabilitation Medicine Tu, Tian Ming; Tan Tock Seng Hospital, National Neuroscience Institute Ang, Yan Hoon; Khoo Teck Puat Hospital, Geriatric Medicine Kong, Keng He; Tan Tock Seng Hospital, Department of Rehabilitation Medicine Singh, Rajinder; Tan Tock Seng Hospital, National Neuroscience Institute Merchant, Reshma; National University Singapore Yong Loo Lin School of Medicine, Department of Medicine

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 1 of 36 BMJ Open BMJ Open: first published as 10.1136/bmjopen-2019-036631 on 23 April 2020. Downloaded from 1 2 3 Chang, Hui Meng; National Neuroscience Institute - Singapore General 4 Hospital Campus 5 Yeo, Tseng Tsai; National University Hospital, Department of 6 Neurosurgery 7 Ning, Chou; National University Hospital, Department of Neurosurgery 8 Cheong, Angela; National University Singapore Saw Swee Hock School of 9 Public Health Tan, Chuen Seng; National University of Singapore, Saw Swee Hock 10 School of Public Health 11 12 Rehabilitation medicine < INTERNAL MEDICINE, Stroke < NEUROLOGY, 13 Keywords: SOCIAL MEDICINE, PUBLIC HEALTH, Organisation of health services < 14 HEALTH SERVICES ADMINISTRATION & MANAGEMENT 15

16 For peer review only 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31

32 http://bmjopen.bmj.com/ 33 34 35 36 37 38 39

40 on September 28, 2021 by guest. Protected copyright. 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 2 of 36

1 2 3

4 BMJ Open: first published as 10.1136/bmjopen-2019-036631 on 23 April 2020. Downloaded from 5 6 7 8 9 I, the Submitting Author has the right to grant and does grant on behalf of all authors of the Work (as defined 10 in the below author licence), an exclusive licence and/or a non-exclusive licence for contributions from authors 11 who are: i) UK Crown employees; ii) where BMJ has agreed a CC-BY licence shall apply, and/or iii) in accordance 12 with the terms applicable for US Federal Government officers or employees acting as part of their official 13 duties; on a worldwide, perpetual, irrevocable, royalty-free basis to BMJ Publishing Group Ltd (“BMJ”) its 14 licensees and where the relevant Journal is co-owned by BMJ to the co-owners of the Journal, to publish the 15 Work in this journal and any other BMJ products and to exploit all rights, as set out in our licence. 16 17 The Submitting Author accepts and understands that any supply made under these terms is made by BMJ to 18 the Submitting Author Forunless you peer are acting as review an employee on behalf only of your employer or a postgraduate 19 student of an affiliated institution which is paying any applicable article publishing charge (“APC”) for Open 20 Access articles. Where the Submitting Author wishes to make the Work available on an Open Access basis (and 21 intends to pay the relevant APC), the terms of reuse of such Open Access shall be governed by a Creative 22 Commons licence – details of these licences and which Creative Commons licence will apply to this Work are set 23 out in our licence referred to above. 24 25 Other than as permitted in any relevant BMJ Author’s Self Archiving Policies, I confirm this Work has not been 26 accepted for publication elsewhere, is not being considered for publication elsewhere and does not duplicate 27 material already published. I confirm all authors consent to publication of this Work and authorise the granting 28 of this licence. 29 30 31 32 33 34 35 36 http://bmjopen.bmj.com/ 37 38 39 40 41 42 43 44 on September 28, 2021 by guest. Protected copyright. 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 3 of 36 BMJ Open

1 1 2 3 Dyadic approach to supervised community rehabilitation 4 BMJ Open: first published as 10.1136/bmjopen-2019-036631 on 23 April 2020. Downloaded from 5 6 participation in an Asian setting post-stroke: exploring role of 7 8 caregiver and patient characteristics in a prospective cohort study 9 1 1 1 10 Shilpa TYAGI , Gerald Choon-Huat KOH *, Nan LUO , Kelvin Bryan 11 2 3 4,5 1 12 TAN , Helen HOENIG , David B. MATCHAR , Joanne YOONG , 13 4 6 7 14 Angelique CHAN , Kim En LEE , N. VENKETASUBRAMANIAN , 15 8 9 10 16 Edward MENON , Kin Ming CHAN , Deidre Anne DE SILVA , Philip 17 11 12 13 14 18 YAP , Boon ForYeow TANpeer, Effie review CHEW , Sherry only H. YOUNG , Yee 19 15 16 17 18 20 Sien NG , Tian Ming TU , Yan Hoon ANG , Keng He KONG , 21 16 19 10 22 Rajinder SINGH , Reshma A. MERCHANT , Hui Meng CHANG , 23 20 20 1 24 Tseng Tsai YEO , Chou NING , Angela CHEONG , Chuen Seng 25 1 26 TAN , PhD 27 *Corresponding Author 28 29 30 31 1Saw Swee Hock School of Public Health, National University of Singapore, 32 33 Singapore, Singapore; 2Policy Research & Economics Office, Ministry of Health, 34 3 35 Singapore, Singapore; Physical Medicine and Rehabilitation Service, Durham VA 36 Medical Centre, USA; 4Program in Health Services and Systems Research, Duke- http://bmjopen.bmj.com/ 37 38 NUS Graduate Medical School, Singapore, Singapore; 5Department of Medicine 39 40 (General Internal Medicine), Duke University Medical Center, Durham, NC, USA; 41 6 7 42 Lee Kim En Neurology Pte Ltd, Singapore, Singapore; Raffles Neuroscience 43 Centre, Raffles Hospital, Singapore, Singapore; 8St. Andrew’s Community Hospital, 44 on September 28, 2021 by guest. Protected copyright. 45 Singapore, Singapore; 9Mount Alvernia Hospital, Singapore, Singapore; 10National 46 47 Neuroscience Institute, Singapore General Hospital campus, Singapore, Singapore; 48 11Geriatric Centre, Khoo Teck Puat Hospital, Singapore, Singapore; 12St. Luke's 49 50 Hospital, Singapore, Singapore; 13Department of Rehabilitation Medicine, National 51 52 University Hospital, Singapore, Singapore; 14Department of Rehabilitation Medicine, 53 15 54 Changi General Hospital, Singapore, Singapore; Department of Rehabilitation 55 Medicine, Singapore General Hospital, Singapore, Singapore; 16Department of 56 57 Neurology, National Neuroscience Institute, Tan Tock Seng Hospital, Singapore, 58 17 59 Singapore; Geriatric Medicine, Khoo Teck Puat Hospital, Singapore, Singapore; 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 4 of 36

2 1 2 3 18Department of Rehabilitation Medicine, Tan Tock Seng Hospital, Singapore, 4 BMJ Open: first published as 10.1136/bmjopen-2019-036631 on 23 April 2020. Downloaded from 5 Singapore; 19Department of Medicine, Yong Loo Lin School of Medicine, National 6 20 7 University of Singapore, Singapore, Singapore; Department of Neurosurgery, 8 National University Hospital, Singapore, Singapore. 9 10 11 12 *Corresponding author 13 14 Dr Gerald CH Koh 15 Saw Swee Hock School of Public Health 16 17 National University of Singapore 18 For peer review only 19 12 Science Drive 2 20 21 #10-01 22 Singapore 117549 23 24 Tel (65) 6516 4979 25 26 Fax (65) 6779 1489 27 Email: [email protected] 28 29 Word count = 2903 [from introduction to conclusion] 30 31 32 33 34 35 36 http://bmjopen.bmj.com/ 37 38 39 40 41 42 43 44 on September 28, 2021 by guest. Protected copyright. 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 5 of 36 BMJ Open

3 1 2 3 4 BMJ Open: first published as 10.1136/bmjopen-2019-036631 on 23 April 2020. Downloaded from 5 6 ABSTRACT 7 8 9 Objective: To study the association of caregiver factors and stroke patient 10 factors with supervised community rehabilitation (SCR) participation over first 11 12 3 months and subsequent 3-12 months post-stroke in an Asian setting. 13 14 15 Design: prospective cohort study 16 17 18 Setting: communityFor setting peer review only 19 20 Participants: We recruited stroke patients and their caregivers into our yearlong 21 22 cohort. Caregiver and patient variables were collected over 3-monthly intervals. We 23 performed logistic regression with outcome variable being SCR participation post- 24 25 stroke. 26 27 Outcome measures: SCR participation over first 3 months and subsequent 3-12 28 29 months post-stroke 30 Results: 251 stroke patient-caregiver dyads were available for current analysis. Mean 31 32 age of caregivers was 50.1 years with majority being female, married and co-residing 33 34 with the stroke patient. There were 61%, 28%, 4% and 7% of spousal, adult-child, 35 sibling and others as caregivers. The odds of SCR participation decreased by about 36 http://bmjopen.bmj.com/ 37 15% for every unit increase in caregiver-reported stroke patient disruptive behaviour 38 39 score (OR: 0.845; 95% CI: 0.769, 0.929). For every 1-unit increase in caregiver’s 40 41 positive management strategy score, the odds of using SCR service increased by 42 about 4% (OR: 1.039; 95% CI: 1.011, 1.068). 43 44 on September 28, 2021 by guest. Protected copyright. 45 Conclusion: We established that SCR participation is jointly determined by 46 47 both caregiver and stroke patient factors, with factors varying over early and 48 49 late post-stroke period. Our results support adoption of a dyadic approach for 50 studying utilization of community rehabilitation services. 51 52 53 Keywords: Stroke rehabilitation, health services, stroke, family caregivers 54 55 56 57 58 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 6 of 36

4 1 2 3 Article summary 4 BMJ Open: first published as 10.1136/bmjopen-2019-036631 on 23 April 2020. Downloaded from 5 - We studied the association of caregiver factors along with stroke patient 6 7 factors with supervised community rehabilitation participation over first 3 8 months and subsequent 3-12 months post-stroke in a prospective yearlong 9 10 cohort study. 11 12 - We are among the first to demonstrate the role of caregivers in stroke patient’s 13 14 supervised community rehabilitation substantiating the rationale for adoption 15 of a stroke patient-caregiver dyadic approach to studying post-stroke 16 17 outcomes. 18 For peer review only 19 - Another strength is the comprehensiveness of caregiver variables considered 20 21 which enabled us to explore the role of caregivers in-depth. 22 - Our study sample included stroke patients surviving first post-stroke year, 23 24 excluding deaths within the follow-up period (<5%) limiting the 25 26 generalizability of our findings to those stroke patients who are alive at the 27 end of first year post-stroke. 28 29 - There is a possibility of information bias related to limited recall of supervised 30 31 community rehabilitation participation by stroke patients and their caregivers, 32 33 which was addressed by keeping relatively shorter recall period. 34 35 36 http://bmjopen.bmj.com/ 37 38 List of abbreviations 39 40 CES-D: Centre for Epidemiological Studies Depression scale; DRCs: day 41 rehabilitation centers; FDW: foreign domestic worker; mRS: modified Rankin scale; 42 43 NIHSS: National Institute of Health Scale; rDMSS: Revised Dementia Management 44 on September 28, 2021 by guest. Protected copyright. 45 Strategies scale; SCR: supervised community rehabilitation. 46 47 48 49 50 INTRODUCTION 51 52 Stroke is associated with significant mortality burden globally.(1) However, recent 53 epidemiological trends with an increased incidence in the younger population and 54 55 decreasing mortality rates over the years highlight the importance of functional 56 57 recovery post-stroke.(2) While rehabilitation is essential for functional recovery and 58 59 re-integration back in the community, it is conditional on stroke patients taking the 60 initiative to attend such rehabilitation services after a stroke event. Transition from

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 7 of 36 BMJ Open

5 1 2 3 inpatient settings into the community is challenging for the stroke patient-caregiver 4 BMJ Open: first published as 10.1136/bmjopen-2019-036631 on 23 April 2020. Downloaded from 5 dyad.(3) They move from a well-supported setting with a multi-disciplinary team 6 7 providing care to a setting where they are on their own trying to maintain the care 8 continuum and seeking community services including supervised community 9 10 rehabilitation (SCR). Concept of duality of stroke crisis with first crisis coinciding 11 12 with the stroke and the second occurring during the discharge from inpatient setting to 13 14 home is reported previously.(4) 15 The role of family caregivers becomes highly relevant post-stroke, considering more 16 17 than half of stroke patients are discharged home with differing degrees of residual 18 For peer review only 19 physical impairments.(5) However, neither is this caregiving role explicitly 20 21 acknowledged nor is the caregiver’s capacity and commitment to provide care 22 assessed.(6) This could result in a mismatch between expected caregiving 23 24 responsibilities and caregiver’s ability to fulfil these, potentially leading to adverse 25 26 consequences for the stroke patient-caregiver dyad. High reliance on caregivers to 27 assist the stroke patients in the community implies that caregiver factors like coping, 28 29 perceived stress, support system and well-being can in turn influence stroke patient’s 30 31 outcomes like participation in supervised rehabilitation. This highlights the 32 33 importance of adopting a stroke patient-caregiver dyadic approach to studying 34 supervised rehabilitation in the community. 35 36 Recognizing the relevance of caregivers and caregiving context in stroke patient’s http://bmjopen.bmj.com/ 37 38 recovery process, researchers have attempted to study the association of caregiver 39 40 availability and some socio-demographic characteristics with functional outcomes 41 post-stroke in patients attending inpatient rehabilitation services.(7-11) While a study 42 43 in US(12) reported positive role of spouse in post-stroke recovery of stroke patients, 44 on September 28, 2021 by guest. Protected copyright. 45 another study in Canada(8) reported caregiver support being associated with higher 46 functional gain as compared to those without caregiver support. A recent study based 47 48 in China explored the role of family member’s positive and negative attitudes in post- 49 50 stroke functional and cognitive recovery, with more positive attitudes being 51 52 associated with more cognitive functional gains after rehabilitation.(13) Increasingly 53 over the years, literature seems to be in favour of caregivers playing an important role 54 55 in stroke patients’ functional and cognitive outcomes post rehabilitation across 56 57 different settings and context. However, none of the studies so far have focused on the 58 association of caregiver characteristics with SCR participation. 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 8 of 36

6 1 2 3 Addressing the above mentioned gaps, we aimed to study the association of caregiver 4 BMJ Open: first published as 10.1136/bmjopen-2019-036631 on 23 April 2020. Downloaded from 5 factors along with stroke patient factors with SCR participation over first 3 months 6 7 and subsequent 3-12 months post-stroke in an Asian setting. 8 9 10 MATERIALS AND METHODS 11 12 Study setting 13 14 In Singapore, after stabilization in a tertiary hospital, stroke patients are assessed for 15 rehabilitation eligibility and based on this assessment, they may undergo intensive 16 17 rehabilitation in an inpatient setting.(14) Another option either in succession to above 18 For peer review only 19 or as an alternative is SCR, often delivered at the day rehabilitation centres or DRCs. 20 21 DRCs are run either within the premises of a step-down facility or as stand-alone 22 centres, mainly providing physiotherapy and occupational therapy.(15) DRCs along 23 24 with nursing homes (i.e. long-term residential care settings within the community) 25 26 and home-based rehabilitation fall under the broad umbrella of SCR. 27 28 29 Participants 30 31 Our participants were part of the prospective Singapore Stroke Study. Singaporeans 32 33 or permanent residents 40 years and above, who suffered a stroke or experienced 34 symptoms within 4 weeks of admission to any of the five tertiary hospitals in 35 36 Singapore during recruitment period (December 2010 to September 2013) with http://bmjopen.bmj.com/ 37 38 confirmed stroke diagnosis were recruited along with their caregivers. Caregivers 39 40 could be an immediate or extended family member or friend who provided care or 41 assistance of any kind and took responsibility of the stroke patient and were 42 43 recognized by the patient, not fully paid for caregiving. Study details are reported 44 on September 28, 2021 by guest. Protected copyright. 45 elsewhere.(16, 17) Ethical approval was obtained from the relevant institutional 46 ethical review boards. 47 48 49 50 Independent variables (caregiver) 51 52 With a primary focus on caregiver factors, following caregiver variables were 53 54 considered for current analysis: socio-demographic characteristics, marital status, 55 56 relationship with caregiver (caregiver being spouse, adult-child, sibling or others 57 58 including distant relatives and friends), number of chronic ailments, co-residing 59 status, caregiver burden, family conflict, social support, caregiver reported patient 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 9 of 36 BMJ Open

7 1 2 3 behavioural issues and adopted caregiver management approaches. Under caregiving 4 BMJ Open: first published as 10.1136/bmjopen-2019-036631 on 23 April 2020. Downloaded from 5 burden, we incorporated measures of both objective and subjective burden measured 6 7 by Oberst Caregiving Burden Scale(18) and Zarit’s Burden Interview(19, 20) 8 respectively. Family caregiving conflict scale recommended by Pearlin and 9 10 colleagues was used to capture family conflict.(21) Adapting Pearlin and colleagues’ 11 12 description of construct of social support, we tried to incorporate both “instrumental” 13 14 and “expressive” dimensions of social support, with former captured by presence of 15 paid help or foreign domestic worker (FDW) for general household tasks or 16 17 specifically for stroke patient and latter captured by Pearlin’s 8-item perceived social 18 For peer review only 19 support instrument.(21) Caregiver reported occurrence of problematic behaviour by 20 21 stroke patients was recorded using the Revised Memory and Behavioural Problem 22 Checklist.(22) Separate summated scores were calculated across the 3 domains of 23 24 disruptive, depressive and memory related behavioural problems with Cronbach’s 25 26 alpha for each domain being 0.73, 0.87 and 0.90 respectively. To capture the care 27 management or coping approaches by stroke patient’s caregiver, we used the revised 28 29 dementia management strategies scale (rDMSS). Previously validated in 30 31 Singapore,(23) the scale has two sub-components of positive and negative types of 32 33 management strategies with good reported internal consistency (Cronbach’s alpha 34 0.89 and 0.87 respectively). 35 36 http://bmjopen.bmj.com/ 37 38 Independent variables (patient) 39 40 Following baseline stroke patient variables were considered: socio-demographic 41 42 characteristics, marital status, ward class as a proxy of socio-economic status, 43 44 Charlson Comorbidity Index, type of stroke (ischemic or non-ischemic), recurrent or on September 28, 2021 by guest. Protected copyright. 45 first stroke, stroke severity measured on National Institute of Health Scale or NIHSS, 46 47 functional status measured on modified Rankin scale or mRS, impairment in 48 49 cognition measured on the Mini-Mental State examination or MMSE, discharge status 50 51 and depression measured on the 11-item version of the Centre for Epidemiological 52 Studies Depression scale (CES-D). For scales with more than 10 missing cases 53 54 (NIHSS, MMSE, Revised memory and behaviour checklist), we used the person 55 56 mean substitution approach to impute for missing values for cases with less than half 57 58 constituting items missing.(24) 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 10 of 36

8 1 2 3 Outcome variables 4 BMJ Open: first published as 10.1136/bmjopen-2019-036631 on 23 April 2020. Downloaded from 5 6 The outcome of interest was uptake of SCR which comprised of any of the following: 7 8 rehabilitation at home, DRC or nursing homes. This information was captured in the 9 survey at 3 monthly intervals by asking the caregiver, “Has the stroke patient at any 10 11 time during the last 3 months received rehabilitation? Please include any 12 13 rehabilitation at home, DRC and nursing homes”. For subsequent 3 to 12 months, we 14 15 created a variable for capturing the SCR uptake information collected at 6, 9 and 12- 16 month interviews. It was coded as yes if the caregiver reported any uptake at either of 17 18 the time points andFor no if peerthe caregiver review reported no uptake only across all time points. 19 20 21 Data analysis 22 23 24 Univariate analysis was conducted to describe the caregiver and stroke patient 25 26 characteristics. We conducted bivariate analysis to examine the associations between 27 independent variables (caregiver and stroke patient factors) and uptake of SCR. 28 29 Independent variables with p values < 0.1 on bivariate analysis were chosen as 30 31 potential predictors for the multivariable regression model. With these potential 32 33 predictor variables, we built the most parsimonious model using a backward variable 34 selection approach. At each model building stage, the most insignificant variable was 35 36 removed until we were left with variables having a p-value < 0.05, except for age, http://bmjopen.bmj.com/ 37 38 gender, ethnicity and ward class of stroke patient which were kept in the model. 39 40 Logistic regression was used at both bivariate and model building stages and we 41 reported the unadjusted and adjusted odds ratio (OR) estimates with 95% confidence 42 43 intervals (CI). We ran separate models for uptake of SCR across two time periods of 44 on September 28, 2021 by guest. Protected copyright. 45 first 3 months post-stroke and subsequent 3 to 12 months. Significance level was set 46 at 5%. With the most parsimonious model, we performed diagnostics for model fit 47 48 using Hosmer and Lemeshow’s goodness-of-fit test [0-3 months Model: p-value= 49 50 0.663; 3-12 months Model: p-value= 0.778], checked for model misspecifications, 51 52 multi-collinearity and influential observations. All analysis was performed in Stata 53 version 14.1.(25) 54 55 56 57 Patient and Public Involvement 58 59 This research was done without patient involvement. Patients were not invited to 60 comment on the study design and were not consulted to develop patient relevant

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 11 of 36 BMJ Open

9 1 2 3 outcomes or interpret the results. Patients were not invited to contribute to the writing 4 BMJ Open: first published as 10.1136/bmjopen-2019-036631 on 23 April 2020. Downloaded from 5 or editing of this document for readability or accuracy. 6 7 8 RESULTS 9 10 251 stroke patient-caregiver dyads were available for current analysis, after exclusion 11 12 of patients with deaths within the follow up period and limiting to complete cases. 13 14 (Please refer figure 1 for study flowchart) The follow-up rates for caregivers were 15 87.2% and 73.4% at 3 and 12 months. The prevalence of any SCR participation was 16 17 49% over 0-3 month and 25% over 3-12 month time period. The mean age of 18 For peer review only 19 caregivers was 50.1 years with majority being female, married and co-residing with 20 21 the stroke patient. There were 61%, 28%, 4% and 7% of spousal, adult-child, sibling 22 and others as caregivers. Mean scores for summated values for memory related, 23 24 depressive and disruptive behaviour problems of stroke patients were 5.01, 3.15 and 25 26 2.67 respectively. 34.27 (10.85) and 11.07 (4.60) were the mean (SD) scores for 27 caregiver reported positive and negative care management strategies respectively. The 28 29 stroke patients had mean age of 61.8 years with majority being male (65%), of 30 31 Chinese ethnicity (59%) and married (81%). About 89% had ischemic index stroke 32 33 and for 17% the index stroke was a recurrent one. Out of all stroke patients, 57%, 34 38% and 5% had mild, moderately severe and severe type of index stroke as measured 35 36 on NIHSS. More than half (59%) had moderate to severe disability (3-5) on MRS. http://bmjopen.bmj.com/ 37 38 (Please refer table 1 for baseline characteristics) 39 40 41 [Figure 1 near here] 42 43 44 on September 28, 2021 by guest. Protected copyright. 45 Table 1. Descriptive characteristics of any supervised community rehabilitation users 46 47 post-stroke 48 49 All stroke Any Users*, Any Users †, 50 51 * 52 patients , 3 months 3-12 months 53 54 No. (%) No. (%) No. (%) 55 56 57 Variable Category 58 59 CAREGIVER FACTORS 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 12 of 36

10 1 2 3 Age of caregiver (in 50.13 (13.09) 49.03 (12.97) 51.16 (10.76) 4 BMJ Open: first published as 10.1136/bmjopen-2019-036631 on 23 April 2020. Downloaded from 5 6 years) mean (SD) 7 8 Gender of caregiver Male 61 (23) 29 (24) 9 (16) 9 10 Female 199 (77) 93 (76) 49 (84) 11 12 13 Ethnicity of caregiver Chinese 151 (58) 66 (54) 35 (60) 14 15 Non-Chinese 109 (42) 56 (46) 23 (40) 16 17 Marital Status of Married 205 (79) 93 (76) 47 (81) 18 For peer review only 19 caregiver 20 21 22 Single 55 (21) 29 (24) 11 (19) 23 24 Caregiver identity Spouse 159 (61) 71 (58) 32 (55) 25 26 Adult-child 74 (28) 36 (30) 17 (29) 27 28 29 Sibling 10 (4) 6 (5) 4 (7) 30 31 Others 17 (7) 9 (7) 5 (9) 32 33 Co-residing with patient Yes 231 (89) 109 (89) 49 (84) 34 35 36 No 29 (11) 13 (11) 9 (16) http://bmjopen.bmj.com/ 37 38 Caring for multiple care Yes 108 (42) 49 (40) 22 (38) 39 40 recipients 41 42 No 152 (58) 73 (60) 36 (62) 43 44 on September 28, 2021 by guest. Protected copyright. 45 Revised memory and behaviour checklist 46 47 Memory problems mean (SD) 5.01 (5.93) 4.95 (6.30) 5.09 (6.06) 48 49 Depressive behaviour mean (SD) 3.15 (4.82) 2.72 (4.21) 3.61 (5.21) 50 51 52 problems 53 54 Disruptive behaviour mean (SD) 2.67 (3.62) 1.97 (2.86) 2.10 (2.83) 55 56 problems 57 58 59 Caregiver Burden 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 13 of 36 BMJ Open

11 1 2 3 Oberst Caregiving mean (SD) 31.71 (12.63) 32.94 (12.02) 35.18 (13.20) 4 BMJ Open: first published as 10.1136/bmjopen-2019-036631 on 23 April 2020. Downloaded from 5 6 Burden Scale 7 8 Zarit Burden Interview mean (SD) 8.73 (7.86) 8.25 (6.83) 8.74 (8.13) 9 10 Family conflict 11 12 13 Attitude towards patient mean (SD) 11.42 (4.49) 11.58 (4.46) 12.00 (4.12) 14 15 Attitude towards mean (SD) 11.63 (4.37) 11.68 (4.40) 12.26 (3.91) 16 17 caregiver 18 For peer review only 19 Social Support (instrumental) 20 21 22 FDW for general help Yes 212 (82) 98 (80) 42 (72) 23 24 No 48 (18) 24 (20) 16 (28) 25 26 FDW for stroke patient Yes 33 (13) 17 (14) 11 (19) 27 28 29 No 227 (87) 105 (86) 47 (81) 30 31 Social Support mean (SD) 26.33 (4.90) 26.49 (5.03) 25.95 (4.61) 32 33 (perceived) 34 35 36 Care Management Strategies http://bmjopen.bmj.com/ 37 38 Positive strategies mean (SD) 34.27 (10.85) 36.52 (10.19) 35.88 (10.92) 39 40 Negative strategies mean (SD) 11.07 (4.60) 10.56 (4.34) 11.36 (4.46) 41 42 PATIENT FACTORS 43 44 on September 28, 2021 by guest. Protected copyright. 45 Age of patient mean (SD) 61.77 (10.42) 60.86 (10.63) 62.29 (10.53) 46 47 (in years) 48 49 Gender of patient Male 169 (65) 77 (63) 37 (64) 50 51 52 Female 91 (35) 45 (37) 21 (36) 53 54 Ethnicity of patient Chinese 153 (59) 65 (53) 35 (60) 55 56 Non-Chinese 107 (41) 57 (47) 23 (40) 57 58 59 Marital Status of patient Married 210 (81) 97 (80) 45 (78) 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 14 of 36

12 1 2 3 Single 50 (19) 25 (20) 13 (22) 4 BMJ Open: first published as 10.1136/bmjopen-2019-036631 on 23 April 2020. Downloaded from 5 6 Ward Class Unsubsidized 21 (8) 11 (9) 7 (12) 7 8 Subsidized 235 (92) 110 (91) 51 (88) 9 10 Charlson Comorbidity 1 - 3 52 (20) 22 (18) 13 (22) 11 12 13 Index 14 15 4 - 6 165 (63) 79 (65) 31 (53) 16 17 >= 7 43 (17) 21 (17) 14 (24) 18 For peer review only 19 Stroke type Ischemic 231 (89) 106 (87) 48 (83) 20 21 22 Non-ischemic 29 (11) 16 (13) 10 (17) 23 24 Recurrent stroke Yes 43 (17) 16 (13) 8 (14) 25 26 No 217 (83) 106 (87) 50 (86) 27 28 29 National Institute of Mild (0-4) 149 (57) 60 (49) 23 (40) 30 31 Health Scale 32 33 Moderately 97 (38) 55 (45) 28 (48) 34 35 36 severe (5-14) http://bmjopen.bmj.com/ 37 38 Severe (15-24) 14 (5) 7 (6) 7 (12) 39 40 Modified Rankin Scale No or slight 106 (41) 38 (31) 11 (19) 41 42 disability (0-2) 43 44 on September 28, 2021 by guest. Protected copyright. 45 Moderate or 154 (59) 84 (69) 47 (81) 46 47 severe 48 49 disability (3-5) 50 51 52 Mini-Mental State No cognitive 150 (58) 70 (57) 30 (52) 53 54 Examination impairment 55 56 (24-30) 57 58 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 15 of 36 BMJ Open

13 1 2 3 Mild cognitive 65 (25) 33 (27) 15 (26) 4 BMJ Open: first published as 10.1136/bmjopen-2019-036631 on 23 April 2020. Downloaded from 5 6 impairment 7 8 (18-23) 9 10 Severe 45 (17) 19 (16) 13 (22) 11 12 13 cognitive 14 15 impairment (1- 16 17 17) 18 For peer review only 19 Discharge to Step-down Yes 66 (25) 33 (27) 19 (33) 20 21 22 facility (Community 23 24 Hospital) 25 26 No 194 (75) 89 (73) 39 (67) 27 28 29 Centre for mean (SD) 6.31 (5.61) 6.59 (5.51) 6.60 (5.64) 30 31 Epidemiological Studies 32 33 Depression Scale 34 35 36 http://bmjopen.bmj.com/ 37 *N = 251, †N=237 38 39 40 Abbreviations: No.: number; SD: standard deviation; FDW: foreign domestic worker 41 42 43 44 on September 28, 2021 by guest. Protected copyright. 45 Supervised community rehabilitation utilization (0-3 months post-stroke) 46 47 48 Table 2 depicts the results of association of caregiver characteristics with odds of 49 50 using SCR across 3 months post-stroke. The variables that entered the final adjusted 51 model of odds of using any SCR service in first 3 months post-stroke were caregiver 52 53 reported disruptive behaviour of patient, positive management strategy by caregiver 54 55 and patient’s functional status (Please refer table 2). For every 1-unit increase in 56 57 caregiver reported stroke patient disruptive behaviour score, the odds of using SCR 58 service decreased by about 15% (OR: 0.845; 95% CI: 0.769, 0.929). For every 1-unit 59 60 increase in caregiver reported positive management strategy score, the odds of using

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 16 of 36

14 1 2 3 SCR service increased by about 4% (OR: 1.039; 95% CI: 1.011, 1.068). Compared to 4 BMJ Open: first published as 10.1136/bmjopen-2019-036631 on 23 April 2020. Downloaded from 5 stroke patients with no or mild functional disability, those with moderate to severe 6 7 functional disability had 2.76 times the odds of using SCR service over first 3 months 8 post-stroke (p=0.001). We further explored the joint influence of positive care 9 10 management strategies and caregiver reported disruptive behaviour of the stroke 11 12 patient on SCR participation by introducing an interaction term in the final adjusted 13 14 model, which was not statistically significant. 15 16 17 Supervised community rehabilitation utilization (3-12 months post-stroke) 18 For peer review only 19 20 Table 3 depicts the results of association of caregiver characteristics with odds of 21 using SCR across 3-12 months post-stroke. The only variable that entered the final 22 23 adjusted model of odds of using any SCR service in 3-12 months post-stroke was 24 25 functional status with odds of using SCR service being 4.19 times in those with 26 moderate to severe functional disability when compared to those with none to mild 27 28 disability (OR: 4.234; 95% CI: 2.034, 8.812) (Please refer table 3). 29 30 31 32 33 34 35 36 http://bmjopen.bmj.com/ 37 38 39 40 41 42 43 44 on September 28, 2021 by guest. Protected copyright. 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open: first published as 10.1136/bmjopen-2019-036631 on 23 April 2020. Downloaded from

Page 17 of 36 BMJ Open

15 1 2 3 4 5 Table 2. Association of caregiver and patient characteristics with supervised community rehabilitation (any use) across 3 months post-stroke. 6 7 Any Rehabilitation (3 months) 8 9 Variable Reference category OR (95% CI) P-value aOR* (95% CI) P-value 10 11 12 CAREGIVER FACTORS For peer review only 13 14 Age (in years) 0.990 (0.971, 1.009) 0.303 15 16 Gender Male 0.889 (0.492, 1.605) 0.696 http://bmjopen.bmj.com/ 17 18 19 Ethnicity Non-Chinese 0.631 (0.380, 1.049) 0.076 20 21 Marital Status Single 0.810 (0.445, 1.474) 0.489 22 23 Caregiver identity Spouse 0.848 24 on September 28, 2021 by guest. Protected copyright. 25 26 Adult-child 1.096 (0.627, 1.918) 27 28 Sibling 1.690 (0.458, 6.232) 29 30 Others 1.268 (0.464, 3.461) 31 32 Comorbid Conditions None 0.076 33 34 35 1 2.136 (1.131, 4.035) 36 37 2 0.697 (0.276, 1.761) 38 39 40 41 42 43 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 BMJ Open: first published as 10.1136/bmjopen-2019-036631 on 23 April 2020. Downloaded from

BMJ Open Page 18 of 36

16 1 2 3 4 5 3 and more 1.119 (0.465, 2.694) 6 7 Co-residing with patient No 1.187 (0.545, 2.584) 0.665 8 9 Caring for multiple care recipients No 0.875 (0.530, 1.446) 0.602 10 11 12 Caregiver relationship For peer review1.131 (0.976, 1.309) only0.101 13 14 with patient 15 16 Revised memory and behaviour checklist http://bmjopen.bmj.com/ 17 18 19 Memory problems 1.004 (0.963, 1.047) 0.856 20 21 Depressive behaviour problems 0.961 (0.911, 1.013) 0.136 22 23 Disruptive behaviour problems 0.889 (0.819, 0.964) 0.005 0.845 (0.769, 0.929) <0.001 24 on September 28, 2021 by guest. Protected copyright. 25 26 Caregiver burden 27 28 Oberst Caregiving Burden Scale 1.013 (0.993, 1.033) 0.202 29 30 Zarit Burden Interview 0.984 (0.953, 1.016) 0.318 31 32 Family conflict 33 34 35 Attitude towards patient 1.029 (0.973, 1.087) 0.202 36 37 Attitude towards caregiver 1.018 (0.962, 1.078) 0.529 38 39 40 41 42 43 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 BMJ Open: first published as 10.1136/bmjopen-2019-036631 on 23 April 2020. Downloaded from

Page 19 of 36 BMJ Open

17 1 2 3 4 5 Social support (instrumental) 6 7 FDW for general help No 0.933 (0.497, 1.751) 0.830 8 9 FDW for stroke patient No 1.143 (0.550, 2.379) 0.720 10 11 12 Social Support (perceived) For peer review1.020 (0.970, 1.073) only0.436 13 14 Care management strategies 15 16 Positive strategies 1.043 (1.018, 1.068) 0.001 http://bmjopen.bmj.com/ 1.039 (1.011, 1.068) 0.006 17 18 19 Negative strategies 0.947 (0.896, 1.002) 0.057 20 21 PATIENT FACTORS 22 23 Age (in years) 0.982 (0.959, 1.006) 0.148 24 on September 28, 2021 by guest. Protected copyright. 25 26 Gender Male 1.169 (0.696, 1.964) 0.556 27 28 Ethnicity Non-Chinese 0.551 (0.330, 0.919) 0.022 29 30 Marital Status Single 0.933 (0.502, 1.733) 0.826 31 32 Ward Class Unsubsidized 0.769 (0.307, 1.928) 0.576 33 34 35 Charlson Comorbidity Index 1 - 3 0.673 36 37 4 - 6 1.335 (0.707, 2.523) 38 39 40 41 42 43 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 BMJ Open: first published as 10.1136/bmjopen-2019-036631 on 23 April 2020. Downloaded from

BMJ Open Page 20 of 36

18 1 2 3 4 5 >= 7 1.258 (0.557, 2.843) 6 7 Stroke type Non-ischemic 0.679 (0.307, 1.502) 0.340 8 9 Recurrent stroke No 0.628 (0.317, 1.244) 0.182 10 11 12 National Institute of Health ForMild (0-4) peer review only0.038 13 14 Scale 15 16 Moderately severe (5-14) 1.974 (1.165, 3.345) http://bmjopen.bmj.com/ 17 18 19 Severe (15-24) 1.633 (0.523, 5.105) 20 21 Modified Rankin Scale No or slight disability (0-2) 0.003 0.001 22 23 Moderate or severe disability (3-5) 2.177 (1.300, 3.645) 2.759 (1.532, 4.969) 24 on September 28, 2021 by guest. Protected copyright. 25 26 Mini-Mental State No cognitive impairment 0.480 27 28 Examination (24-30) 29 30 Mild cognitive impairment (18-23) 1.263 (0.693, 2.300) 31 32 Severe cognitive impairment (1-17) 0.783 (0.399, 1.538) 33 34 35 36 37 38 39 40 41 42 43 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 BMJ Open: first published as 10.1136/bmjopen-2019-036631 on 23 April 2020. Downloaded from

Page 21 of 36 BMJ Open

19 1 2 3 4 5 Discharge to Step-down facility No 1.172 (0.664, 2.069) 0.584 6 7 (Community 8 9 Hospital) 10 11 12 Centre for Epidemiological For peer review1.014 (0.969, 1.060) only0.555 13 14 Studies Depression Scale 15 16 Abbreviations: OR: odds ratio; aOR: adjusted odds ratio; CI: confidence interval; FDW: foreign domestic worker http://bmjopen.bmj.com/ 17 18 *Model adjusted for patient’s age, gender, ethnicity and ward class 19 20 21 22 23 24 on September 28, 2021 by guest. Protected copyright. 25 Table 3. Association of caregiver and patient characteristics with supervised community rehabilitation (any use) across 3-12 months post-stroke. 26 27 Any Rehabilitation (3 - 12 months) 28 29 Variable Reference category OR (95% CI) P-value aOR* (95% CI) P-value 30 31 32 CAREGIVER FACTORS 33 34 Age (in years) 1.008 (0.985, 1.031) 0.500 35 36 Gender Male 1.828 (0.832, 4.016) 0.133 37 38 39 40 41 42 43 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 BMJ Open: first published as 10.1136/bmjopen-2019-036631 on 23 April 2020. Downloaded from

BMJ Open Page 22 of 36

20 1 2 3 4 5 Ethnicity Non-Chinese 1.097 (0.600, 2.008) 0.763 6 7 Marital Status Single 1.230 (0.584, 2.589) 0.587 8 9 Caregiver identity Spouse 0.485 10 11 12 Adult-child For peer review1.236 (0.628, 2.432) only 13 14 Sibling 2.375 (0.632, 8.931) 15 16 Others 1.781 (0.568, 5.586) http://bmjopen.bmj.com/ 17 18 19 Comorbid Conditions None 0.349 20 21 1 1.036 (0.505, 2.124) 22 23 2 0.432 (0.121, 1.543) 24 on September 28, 2021 by guest. Protected copyright. 25 26 3 and more 0.432 (0.121, 1.543) 27 28 Co-residing with patient No 0.534 (0.222, 1.284) 0.161 29 30 Caring for multiple care recipients No 0.828 (0.451, 1.521) 0.543 31 32 Caregiver relationship 1.002 (0.846, 1.187) 0.979 33 34 35 with patient 36 37 Revised memory and behaviour checklist 38 39 40 41 42 43 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 BMJ Open: first published as 10.1136/bmjopen-2019-036631 on 23 April 2020. Downloaded from

Page 23 of 36 BMJ Open

21 1 2 3 4 5 Memory problems 1.004 (0.954, 1.056) 0.888 6 7 Depressive behaviour problems 1.027 (0.968, 1.089) 0.377 8 9 Disruptive behaviour problems 0.941 (0.855, 1.035) 0.209 10 11 12 Caregiver burden For peer review only 13 14 Oberst Caregiving Burden Scale 1.026 (1.003, 1.049) 0.025 15 16 Zarit Burden Interview 1.000 (0.963, 1.038) 0.997 http://bmjopen.bmj.com/ 17 18 19 Family conflict 20 21 Attitude towards patient 1.030 (0.962, 1.102) 0.399 22 23 Attitude towards caregiver 10.33 (0.963, 1.108) 0.369 24 on September 28, 2021 by guest. Protected copyright. 25 26 Social support (instrumental) 27 28 FDW for general help No 0.406 (0.198, 0.834) 0.014 29 30 FDW for stroke patient No 2.758 (1.175, 6.477) 0.020 31 32 Social Support (perceived) 0.971 (0.914, 1.033) 0.356 33 34 35 Care management strategies 36 37 Positive strategies 1.016 (0.988, 1.044) 0.278 38 39 40 41 42 43 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 BMJ Open: first published as 10.1136/bmjopen-2019-036631 on 23 April 2020. Downloaded from

BMJ Open Page 24 of 36

22 1 2 3 4 5 Negative strategies 1.009 (0.948, 1.074) 0.778 6 7 PATIENT FACTORS 8 9 Age (in years) 1.010 (0.981, 1.040) 0.491 10 11 12 Gender ForMale peer review1.126 (0.606, 2.090) only0.708 13 14 Ethnicity Non-Chinese 1.072 (0.586, 1.963) 0.821 15 16 Marital Status Single 0.725 (0.350, 1.502) 0.387 http://bmjopen.bmj.com/ 17 18 19 Ward Class Unsubsidized 0.439 (0.159, 1.212) 0.112 20 21 Charlson Comorbidity Index 1 - 3 0.186 22 23 4 - 6 0.687 (0.324, 1.457) 24 on September 28, 2021 by guest. Protected copyright. 25 26 >= 7 1.356 (0.547, 3.364) 27 28 Stroke type Non-ischemic 0.471 (0.201, 1.106) 0.084 29 30 Recurrent stroke No 0.735 (0.318, 1.700) 0.472 31 32 National Institute of Health Mild (0-4) 0.003 33 34 35 Scale 36 37 Moderately severe (5-14) 2.235 (1.186, 4.212) 38 39 40 41 42 43 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 BMJ Open: first published as 10.1136/bmjopen-2019-036631 on 23 April 2020. Downloaded from

Page 25 of 36 BMJ Open

23 1 2 3 4 5 Severe (15-24) 5.681 (1.747, 6 7 18.473) 8 9 Modified Rankin Scale No or slight disability (0-2) <0.001 <0.001 10 11 12 Moderate or severe disability (3-5)For peer review4.132 (2.013, 8.479) only 4.234 (2.034, 8.812) 13 14 Mini-Mental State No cognitive impairment 0.303 15 16 Examination (24-30) http://bmjopen.bmj.com/ 17 18 19 Mild cognitive impairment (18-23) 1.279 (0.627, 2.609) 20 21 Severe cognitive impairment (1-17) 1.833 (0.842, 3.994) 22 23 Discharge to Community No 1.808 (0.939, 3.480) 0.076 24 on September 28, 2021 by guest. Protected copyright. 25 26 Hospital 27 28 Centre for Epidemiological 1.009 (0.957, 1.065) 0.730 29 30 Studies Depression Scale 31 32 Abbreviations: OR: odds ratio; aOR: adjusted odds ratio; CI: confidence interval; FDW: foreign domestic worker 33 34 *Model adjusted for patient’s age, gender, ethnicity and ward class 35 36 37 38 39 40 41 42 43 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 BMJ Open Page 26 of 36

24 1 2 3 DISCUSSION 4 BMJ Open: first published as 10.1136/bmjopen-2019-036631 on 23 April 2020. Downloaded from 5 We are among the first to establish that participation in SCR is jointly determined by 6 7 both caregiver factors and stroke patient factors. We also demonstrated that SCR 8 9 determinants vary across different time periods post-stroke, with caregiver 10 determinants being significant over early post-stroke period (0-3 months) and stroke 11 12 patient determinants being significant over both early and late post-stroke period (3- 13 14 12 months). 15 16 Past literature has acquainted us fairly well with the role of caregiver factors, such as, 17 caregiver availability,(10) support(8, 11) and psychosocial health(7) in functional 18 For peer review only 19 recovery or gains post-stroke. Clark and colleagues reported stronger chance beliefs 20 21 of the caregivers to be associated with decreased likelihood of their stroke patient 22 23 attending outpatient medicine and rehabilitation therapy appointments.(26) 24 Researchers have reported caregiver factors to be associated with delayed discharge 25 26 from inpatient settings.(27, 28) Another study exploring the caregiver determinants of 27 28 post-stroke inpatient rehabilitation reported co-residing with caregivers to be 29 associated with decreased utilization of inpatient rehabilitation.(29) We did not find 30 31 any significant association between co-residing caregiver status and SCR 32 33 participation. While authors in these studies demonstrated the role of caregivers in 34 35 inpatient rehabilitation setting, our study adds new knowledge on role of caregivers in 36 SCR participation once they are discharged back home. http://bmjopen.bmj.com/ 37 38 We are the first to study the role of caregiver factors in SCR participation over early 39 40 and late post-stroke periods and have demonstrated that the association of caregiver 41 factors with SCR participation varied over these two time periods. Specifically, 42 43 caregiver factors played a significant role during the early post-stroke period, with 44 on September 28, 2021 by guest. Protected copyright. 45 stroke patient’s functional status being the only significant factor in the late post- 46 47 stroke period. Possible explanation could be related to the transition to community 48 and adaptation related challenges in early post-stroke period with caregivers playing a 49 50 crucial role in the rehabilitation journey during this phase. The importance of this 51 52 finding is further emphasized considering higher participation in supervised 53 54 rehabilitation in early post-stroke period is reported to be associated with better 55 functional outcomes at 1 year post-stroke.(30) 56 57 We found that the odds of SCR participation increased with increase in the positive 58 59 care management strategy score. Possible explanation could be caregivers practicing 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 27 of 36 BMJ Open

25 1 2 3 positive coping adapt better to their new role with lower psychological issues like 4 BMJ Open: first published as 10.1136/bmjopen-2019-036631 on 23 April 2020. Downloaded from 5 anxiety and are better able to care for stroke patients including facilitation of SCR 6 7 participation. Along the same lines, a cross-sectional study on patients suffering 8 cerebrovascular accidents reported poorer functional outcomes in patients of 9 10 caregivers suffering from anxiety.(7) We found that the odds of SCR participation 11 12 decreased with increase in caregiver reported stroke patient’s disruptive behaviour 13 14 score. Researchers have reported that managing problematic behaviours post-stroke 15 can be difficult for the caregivers resulting in caregiver strain which might make them 16 17 unable to continue providing care(31) and, in some circumstances, resulting in 18 For peer review only 19 institutionalization of stroke patients.(32) Our finding of stroke patient’s disruptive 20 21 behaviour being a predictor of decreased SCR participation may be explained by 22 caregivers being stressed and unable to comply with their caregiving obligations, 23 24 including facilitation of SCR participation. 25 26 Following are the practical implications from our work. Efforts should be directed 27 towards promoting positive care management strategies among caregivers by 28 29 optimizing their efficacy in caregiving related tasks so that they adapt well. Currently, 30 31 caregiver competency training is mainly focused on physical assistance. However, the 32 33 scope should also include mastering skills to manage behavioural issues post-stroke. 34 Our results support adoption of a family-centred approach to post-stroke rehabilitation 35 36 providing due recognition to the family caregivers. A review on family-centred http://bmjopen.bmj.com/ 37 38 approach towards post-stroke rehabilitation recommended keeping the caregivers 39 40 informed, involving them in setting rehabilitation goals, teaching coping skills and 41 improving self-efficacy.(33) Another practical recommendation would be moving 42 43 from a reactive to a pro-active approach towards ensuring the stroke patient-caregiver 44 on September 28, 2021 by guest. Protected copyright. 45 dyads adjust well in the community. Caregiver readiness can be improved by doing an 46 assessment of the stroke patient’s care needs and the caregiver’s capacity and 47 48 commitment to care.(6) 49 50 51 Study limitations 52 53 Following are the limitations. There is a possibility of information bias related to 54 55 limited recall of SCR utilization. To address this, we kept our recall period to 3 56 57 months as past literature recommends shorter recall periods to ensure greater accuracy 58 59 of reporting utilization.(34) We limited the current sample to stroke patients surviving 60 first post-stroke year, excluding deaths within the follow-up period (<5%). Since

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 28 of 36

26 1 2 3 those that survived the first post-stroke year versus those who died during this period 4 BMJ Open: first published as 10.1136/bmjopen-2019-036631 on 23 April 2020. Downloaded from 5 might be systematically different, our findings would be generalizable to those stroke 6 7 patients who are alive at the end of first year post-stroke. Another limitation is related 8 to the temporality across caregiver characteristics and SCR participation over the first 9 10 three months after stroke as both were determined simultaneously at the end of 3 11 12 months. However, we did have the temporality across caregiver characteristics and 13 14 SCR participation over 3-12 months after stroke. 15 16 17 Study strengths 18 For peer review only 19 Our study has some strengths. We are among the first, to the best of our knowledge, 20 21 to demonstrate the role caregivers play in stroke patient’s SCR participation. Our 22 results have substantiated to some extent the rationale for adoption of a stroke patient- 23 24 caregiver dyadic approach(16) to studying post-stroke SCR utilization. We reported 25 26 the relative importance of caregiver factors in early as compared to late post-stroke 27 period. Another strength was the comprehensiveness of caregiver variables considered 28 29 which enabled us to explore the role of caregivers in-depth. 30 31 32 CONCLUSION 33 34 With the aim to study the caregiver determinants of SCR after stroke, our study 35 36 demonstrated that the decision of community rehabilitation participation is not http://bmjopen.bmj.com/ 37 38 singularly dependent on the stroke patient’s clinical or functional characteristics but 39 40 rather influenced by both caregiver and stroke patient’s characteristics. We found that 41 caregiver’s positive care management strategies increased the odds of SCR 42 43 participation and caregiver reported disruptive behaviour of the stroke patient 44 on September 28, 2021 by guest. Protected copyright. 45 decreased the odds of SCR participation over 3 months post-stroke. Moving forward, 46 our results make a case for adoption of a dyadic approach for studying post-stroke 47 48 utilization of community rehabilitation services or planning for such services or 49 50 resource allocation. 51 52 53 Acknowledgements 54 55 We would like to thank the medical staff at the public tertiary hospitals for 56 57 assisting with the recruitment of patients and their caregivers. We would 58 59 also like to thank all the participants in our study for their participation and 60 cooperation.

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 29 of 36 BMJ Open

27 1 2 3 Declaration of interest statement 4 BMJ Open: first published as 10.1136/bmjopen-2019-036631 on 23 April 2020. Downloaded from 5 The authors report no conflicts of interest. 6 7 8 Funding 9 10 This work was supported by Health Services Research Competitive Research Grant 11 12 from the National Medical Research Council, Singapore. 13 14 15 Data availability 16 17 The dataset used and analysed during the current study is available from the 18 For peer review only 19 corresponding author on reasonable request. 20 21 22 Authors’ contribution 23 24 ST was involved in conceptualization and design of the study, analysis and 25 26 interpretation of data, original draft preparation and incorporating revisions in 27 28 manuscript based on critical inputs from other co-authors. GCHK was involved in 29 conceptualization and design of the study, acquisition of data, drafting of the 30 31 manuscript and providing critical inputs to revision of manuscript along with 32 33 supervision of the study. NL was involved in conceptualization and design of the 34 35 study, acquisition of data, drafting of the manuscript and providing critical inputs to 36 revision of manuscript. KBT was involved in conceptualization and design of the http://bmjopen.bmj.com/ 37 38 study, acquisition of data, drafting of the manuscript and providing critical inputs to 39 40 revision of manuscript. HH made substantial contributions to conception and design 41 42 of the study specifically with provision of expertise in medical domain and was 43 involved in revising the manuscript critically for intellectual content. DBM made 44 on September 28, 2021 by guest. Protected copyright. 45 substantial contributions to conception and design of the study specifically, with 46 47 provision of expertise in medical domain and was involved in revising the manuscript 48 critically for intellectual content. JY made substantial contributions to conception and 49 50 design of the study specifically with provision of expertise in financial domain and 51 52 was involved in revising the manuscript critically for intellectual content. AC made 53 54 substantial contributions to conception and design of the study specifically with 55 provision of expertise in social domain and was involved in revising the manuscript 56 57 critically for intellectual content. KEL was involved in acquisition of data and in 58 59 revising the manuscript critically for intellectual content. NV was involved in 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 30 of 36

28 1 2 3 acquisition of data and in revising the manuscript critically for intellectual content. 4 BMJ Open: first published as 10.1136/bmjopen-2019-036631 on 23 April 2020. Downloaded from 5 EM was involved in acquisition of data and in revising the manuscript critically for 6 7 intellectual content. KMC was involved in acquisition of data and in revising the 8 manuscript critically for intellectual content. DADS was involved in acquisition of 9 10 data and in revising the manuscript critically for intellectual content. PY was involved 11 12 in acquisition of data and in revising the manuscript critically for intellectual content. 13 14 BYT was involved in acquisition of data and in revising the manuscript critically for 15 intellectual content. EC was involved in acquisition of data and in revising the 16 17 manuscript critically for intellectual content. SHY was involved in acquisition of data 18 For peer review only 19 and in revising the manuscript critically for intellectual content. YSN was involved in 20 21 acquisition of data and in revising the manuscript critically for intellectual content. 22 TMT was involved in acquisition of data and in revising the manuscript critically for 23 24 intellectual content. YHA was involved in acquisition of data and in revising the 25 26 manuscript critically for intellectual content. KHK was involved in acquisition of data 27 and in revising the manuscript critically for intellectual content. RS was involved in 28 29 acquisition of data and in revising the manuscript critically for intellectual content. 30 31 RAM was involved in acquisition of data and in revising the manuscript critically for 32 33 intellectual content. HMC was involved in acquisition of data and in revising the 34 manuscript critically for intellectual content. TTY was involved in acquisition of data 35 36 and in revising the manuscript critically for intellectual content. CN was involved in http://bmjopen.bmj.com/ 37 38 acquisition of data and in revising the manuscript critically for intellectual content. 39 40 AC was involved in acquisition of data and in revising the manuscript critically for 41 intellectual content. CST was involved in conceptualization and design of the study, 42 43 analysis and interpretation of data, drafting of the manuscript and providing critical 44 on September 28, 2021 by guest. Protected copyright. 45 inputs to revision of manuscript. All the authors have read and approved the final 46 version of the manuscript to be published and are agreeable to take accountability of 47 48 all aspects of the work involved in the manuscript. 49 50 51 52 53 REFERENCES 54 55 56 57 1. Strong K, Mathers C, Bonita R. Preventing stroke: saving lives around the 58 59 world. The Lancet Neurology. 2007;6(2):182-7. 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 31 of 36 BMJ Open

29 1 2 3 2. Singapore Stroke Registry. Trends in Stroke in Singapore 2005-2012. 4 BMJ Open: first published as 10.1136/bmjopen-2019-036631 on 23 April 2020. Downloaded from 5 Singapore: National Registry of Diseases Office, Ministry of Health. 6 7 3. Cameron JI, Tsoi C, Marsella A. Optimizing stroke systems of care by 8 enhancing transitions across care environments. Stroke. 2008;39(9):2637-43. 9 10 4. Lutz BJ, Young ME, Cox KJ et al. The crisis of stroke: experiences of patients 11 12 and their family caregivers. Topics in stroke rehabilitation. 2011;18(6):786-97. 13 14 5. Thom T, Haase N, Rosamond W et al. Heart disease and stroke statistics-- 15 2006 update: a report from the American Heart Association Statistics Committee and 16 17 Stroke Statistics Subcommittee. Circulation. 2006;113(6):e85-151. 18 For peer review only 19 6. Lutz BJ, Young ME, Creasy KR et al. Improving stroke caregiver readiness 20 21 for transition from inpatient rehabilitation to home. The Gerontologist. 22 2016;57(5):880-9. 23 24 7. Em S, Bozkurt M, Caglayan M et al. Psychological health of caregivers and 25 26 association with functional status of stroke patients. Topics in stroke rehabilitation. 27 2017;24(5):323-9. 28 29 8. Harris JE, Eng JJ, Miller WC et al. The role of caregiver involvement in 30 31 upper-limb treatment in individuals with subacute stroke. Physical therapy. 32 33 2010;90(9):1302-10. 34 9. Koh GC, Chen C, Cheong A et al. Trade-offs between effectiveness and 35 36 efficiency in stroke rehabilitation. International journal of stroke : official journal of http://bmjopen.bmj.com/ 37 38 the International Stroke Society. 2012;7(8):606-14. 39 40 10. Koh GCH, Wee LE, Chen C et al. Caregivers and their impact on inpatient 41 rehabilitation efficiency and effectiveness amongst recent stroke survivors in an 42 43 urbanised Asian society. Am Heart Assoc; 2012. 44 on September 28, 2021 by guest. Protected copyright. 45 11. Tsouna-Hadjis E, Vemmos KN, Zakopoulos N et al. First-stroke recovery 46 process: the role of family social support. Arch Phys Med Rehabil. 2000;81(7):881-7. 47 48 12. Baker AC. The spouse's positive effect on the stroke patient's recovery. 49 50 Rehabilitation nursing : the official journal of the Association of Rehabilitation 51 52 Nurses. 1993;18(1):30-3. 53 13. Fang Y, Tao Q, Zhou X et al. Patient and Family Member Factors Influencing 54 55 Outcomes of Poststroke Inpatient Rehabilitation. Arch Phys Med Rehabil. 56 57 2017;98(2):249-55 e2. 58 14. Chen C, Koh GC, Naidoo N et al. Trends in length of stay, functional 59 60 outcomes, and discharge destination stratified by disease type for inpatient

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 32 of 36

30 1 2 3 rehabilitation in Singapore community hospitals from 1996 to 2005. Arch Phys Med 4 BMJ Open: first published as 10.1136/bmjopen-2019-036631 on 23 April 2020. Downloaded from 5 Rehabil. 2013;94(7):1342-51 e4. 6 7 15. Intermediate and long-term care (ILTC) services. Ministry of Health (MOH), 8 Singapore. [cited 2018 November 5, 2018]. Available from: 9 10 https://http://www.moh.gov.sg/our-healthcare-system/healthcare-services-and- 11 12 facilities/intermediate-and-long-term-care-(iltc)-services 13 14 16. Tyagi S, Koh GC, Luo N et al. Dyadic approach to post-stroke 15 hospitalizations: role of caregiver and patient characteristics. BMC neurology. 16 17 2019;19(1):267. 18 For peer review only 19 17. Tyagi S, Koh GC-H, Luo N et al. Healthcare utilization and cost trajectories 20 21 post-stroke: role of caregiver and stroke factors. BMC health services research. 22 2018;18(1):881. 23 24 18. Bakas T, Austin JK, Jessup SL et al. Time and difficulty of tasks provided by 25 26 family caregivers of stroke survivors. Journal of Neuroscience Nursing. 27 2004;36(2):95. 28 29 19. Bédard M, Molloy DW, Squire L et al. The Zarit Burden Interview: a new 30 31 short version and screening version. The gerontologist. 2001;41(5):652-7. 32 33 20. Seng BK, Luo N, Ng WY et al. Validity and reliability of the Zarit Burden 34 Interview in assessing caregiving burden. Ann Acad Med Singapore. 2010;39:758-63. 35 36 21. Pearlin LI, Mullan JT, Semple SJ et al. Caregiving and the stress process: An http://bmjopen.bmj.com/ 37 38 overview of concepts and their measures. The gerontologist. 1990;30(5):583-94. 39 40 22. Teri L, Truax P, Logsdon R et al. Assessment of behavioral problems in 41 dementia: the revised memory and behavior problems checklist. Psychology and 42 43 aging. 1992;7(4):622. 44 on September 28, 2021 by guest. Protected copyright. 45 23. Tan L, Yap P, Ng WY et al. Exploring the use of the Dementia Management 46 Strategies Scale in caregivers of persons with dementia in Singapore. Aging & mental 47 48 health. 2013;17(8):935-41. 49 50 24. Downey RG, King C. Missing data in Likert ratings: A comparison of 51 52 replacement methods. J Gen Psychol. 1998;125(2):175-91. 53 25. StataCorp. 2015. Stata Statistical Software: Release 14. College Station, TX: 54 55 StataCorp LP. 56 57 26. Clark AN, Sander AM, Pappadis MR et al. Caregiver characteristics and their 58 relationship to health service utilization in minority patients with first episode stroke. 59 60 NeuroRehabilitation. 2010;27(1):95-104.

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 33 of 36 BMJ Open

31 1 2 3 27. Lai W, Buttineau M, Harvey JK et al. Clinical and psychosocial predictors of 4 BMJ Open: first published as 10.1136/bmjopen-2019-036631 on 23 April 2020. Downloaded from 5 exceeding target length of stay during inpatient stroke rehabilitation. Topics in stroke 6 7 rehabilitation. 2017;24(7):510-6. 8 28. Tan WS, Chong WF, Chua KS et al. Factors associated with delayed 9 10 discharges after inpatient stroke rehabilitation in Singapore. Ann Acad Med 11 12 Singapore. 2010;39(6):435-41. 13 14 29. Hinojosa MS, Rittman M, Hinojosa R. Informal caregivers and racial/ethnic 15 variation in health service use of stroke survivors. Journal of rehabilitation research 16 17 and development. 2009;46(2):233-41. 18 For peer review only 19 30. Koh GC, Saxena SK, Ng TP et al. Effect of duration, participation rate, and 20 21 supervision during community rehabilitation on functional outcomes in the first 22 poststroke year in Singapore. Arch Phys Med Rehabil. 2012;93(2):279-86. 23 24 31. Cameron JI, Cheung AM, Streiner DL et al. Stroke survivors' behavioral and 25 26 psychologic symptoms are associated with informal caregivers' experiences of 27 depression. Arch Phys Med Rehabil. 2006;87(2):177-83. 28 29 32. Stephens S. Who's There?: When stroke or Alzheimer's changes a person's 30 31 behavior, caregiving can become extreme. Here, experienced caregivers, patients, and 32 33 experts share their stories and advice. Neurology Now. 2009;5(4):26-9. 34 33. Visser-Meily A, Post M, Gorter JW et al. Rehabilitation of stroke patients 35 36 needs a family-centred approach. Disability and rehabilitation. 2006;28(24):1557-61. http://bmjopen.bmj.com/ 37 38 34. Bhandari A, Wagner T. Self-reported utilization of health care services: 39 40 improving measurement and accuracy. Medical Care Research and Review. 41 2006;63(2):217-35. 42 43 44 on September 28, 2021 by guest. Protected copyright. 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 34 of 36

32 1 2 3 Figure 1. Study flowchart 4 BMJ Open: first published as 10.1136/bmjopen-2019-036631 on 23 April 2020. Downloaded from 5 6 7 8 9 10 11 12 13 14 15 16 17 18 For peer review only 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 http://bmjopen.bmj.com/ 37 38 39 40 41 42 43 44 on September 28, 2021 by guest. Protected copyright. 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 35 of 36 BMJ Open BMJ Open: first published as 10.1136/bmjopen-2019-036631 on 23 April 2020. Downloaded from 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 For peer review only 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31

32 http://bmjopen.bmj.com/ 33 34 35 36 37 38 39

40 on September 28, 2021 by guest. Protected copyright. 41 42 43 44 45 Study Flowchart 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 36 of 36

1 STROBE Statement—Checklist of items that should be included in reports of cohort studies 2 3

4 Item Page No BMJ Open: first published as 10.1136/bmjopen-2019-036631 on 23 April 2020. Downloaded from 5 No Recommendation 6 Title and abstract 1 (a) Indicate the study’s design with a commonly used term in the title or Title page, Page-1 7 the abstract 8 (b) Provide in the abstract an informative and balanced summary of what Abstract, 9 Page-3 10 was done and what was found 11 Introduction 12 Introduction, 13 Background/rationale 2 Explain the scientific background and rationale for the investigation Page-4-6 14 being reported 15 Objectives 3 State specific objectives, including any prespecified hypotheses Introduction, 16 Page-6 17 Methods 18 For peer review only Methods, 19 Study design 4 Present key elements of study design early in the paper Page-6 20 Setting 5 Describe the setting, locations, and relevant dates, including periods of Methods, 21 Page-6 22 recruitment, exposure, follow-up, and data collection 23 Participants 6 (a) Give the eligibility criteria, and the sources and methods of selection Methods, 24 Page-6 of participants. Describe methods of follow-up 25 NA 26 (b) For matched studies, give matching criteria and number of exposed 27 and unexposed 28 Variables 7 Clearly define all outcomes, exposures, predictors, potential confounders, Methods, 29 Page 7-8 and effect modifiers. Give diagnostic criteria, if applicable 30 31 Data sources/ 8* For each variable of interest, give sources of data and details of methods Methods, Page 6-8 32 measurement of assessment (measurement). Describe comparability of assessment 33 methods if there is more than one group 34 Bias 9 Describe any efforts to address potential sources of bias Discussion, 35 Page-25-26

36 http://bmjopen.bmj.com/ Study size 10 Explain how the study size was arrived at Based on 37 availability 38 of data. 39 Quantitative variables 11 Explain how quantitative variables were handled in the analyses. If Methods 40 (analysis), applicable, describe which groupings were chosen and why 41 Page-8 42 Statistical methods 12 (a) Describe all statistical methods, including those used to control for Methods 43 (analysis), 44 confounding Page-8 on September 28, 2021 by guest. Protected copyright. 45 (b) Describe any methods used to examine subgroups and interactions NA 46 NA 47 (c) Explain how missing data were addressed 48 (d) If applicable, explain how loss to follow-up was addressed Follow-up 49 rates are 50 given under Results, 51 Page-9. 52 Not done. 53 (e) Describe any sensitivity analyses 54 Results 55 Participants 13* (a) Report numbers of individuals at each stage of study—eg numbers Results, 56 Page-9 & potentially eligible, examined for eligibility, confirmed eligible, included 57 figure 1 for 58 in the study, completing follow-up, and analysed study 59 flowchart. 60 (b) Give reasons for non-participation at each stage Figure 1 (c) Consider use of a flow diagram Figure 1

1 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 37 of 36 BMJ Open

1 Descriptive data 14* (a) Give characteristics of study participants (eg demographic, clinical, Results, 2 Page 9 & social) and information on exposures and potential confounders 3 Table 1

4 (b) Indicate number of participants with missing data for each variable of Table 1 BMJ Open: first published as 10.1136/bmjopen-2019-036631 on 23 April 2020. Downloaded from 5 interest 6 (c) Summarise follow-up time (eg, average and total amount) Follow-up 7 of 1 year 8 after index 9 stroke. 10 Outcome data 15* Report numbers of outcome events or summary measures over time Results, 11 Page-9 12 13 14 15 16 17 18 For peer review only 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35

36 http://bmjopen.bmj.com/ 37 38 39 40 41 42 43 44 on September 28, 2021 by guest. Protected copyright. 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

2 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 38 of 36

1 Main results 16 (a) Give unadjusted estimates and, if applicable, confounder-adjusted estimates and Results, 2 Table 2 & their precision (eg, 95% confidence interval). Make clear which confounders were 3 Table 3.

4 adjusted for and why they were included BMJ Open: first published as 10.1136/bmjopen-2019-036631 on 23 April 2020. Downloaded from 5 (b) Report category boundaries when continuous variables were categorized Table 1. 6 (c) If relevant, consider translating estimates of relative risk into absolute risk for a NA 7 meaningful time period 8 9 Other analyses 17 Report other analyses done—eg analyses of subgroups and interactions, and Results, Page-14 10 sensitivity analyses 11 12 Discussion 13 Key results 18 Summarise key results with reference to study objectives Discussion, 14 Page-24 15 Limitations 19 Discuss limitations of the study, taking into account sources of potential bias or Discussion, Page-25-26 16 imprecision. Discuss both direction and magnitude of any potential bias 17 Interpretation 20 Give a cautious overall interpretation of results considering objectives, limitations, Conclusion, 18 For peer review only Page-26 19 multiplicity of analyses, results from similar studies, and other relevant evidence 20 Generalisability 21 Discuss the generalisability (external validity) of the study results Discussion, 21 under study 22 limitations Page-25-26 23 24 Other information 25 Funding 22 Give the source of funding and the role of the funders for the present study and, if Funding 26 Page - 27 applicable, for the original study on which the present article is based 27 28 29 *Give information separately for exposed and unexposed groups. 30 31 Note: An Explanation and Elaboration article discusses each checklist item and gives methodological background and 32 33 published examples of transparent reporting. The STROBE checklist is best used in conjunction with this article (freely 34 available on the Web sites of PLoS Medicine at http://www.plosmedicine.org/, Annals of Internal Medicine at 35 http://www.annals.org/, and Epidemiology at http://www.epidem.com/). Information on the STROBE Initiative is

36 http://bmjopen.bmj.com/ available at http://www.strobe-statement.org. 37 38 39 40 41 42 43 44 on September 28, 2021 by guest. Protected copyright. 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

3 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open BMJ Open: first published as 10.1136/bmjopen-2019-036631 on 23 April 2020. Downloaded from

Dyadic approach to supervised community rehabilitation participation in an Asian setting post-stroke: exploring the role of caregiver and patient characteristics in a prospective cohort study

ForJournal: peerBMJ Open review only Manuscript ID bmjopen-2019-036631.R1

Article Type: Original research

Date Submitted by the 24-Feb-2020 Author:

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 1 of 40 BMJ Open BMJ Open: first published as 10.1136/bmjopen-2019-036631 on 23 April 2020. Downloaded from 1 2 3 Chang, Hui Meng; National Neuroscience Institute - Singapore General 4 Hospital Campus 5 Yeo, Tseng Tsai; National University Hospital, Department of 6 Neurosurgery 7 Ning, Chou; National University Hospital, Department of Neurosurgery 8 Cheong, Angela; National University Singapore Saw Swee Hock School of 9 Public Health Tan, Chuen Seng; National University of Singapore, Saw Swee Hock 10 School of Public Health 11 12 Primary Subject Health services research 13 Heading: 14 Secondary Subject Heading: Neurology, Health services research, Rehabilitation medicine 15 16 For peerRehabilitation reviewmedicine < INTERNAL only MEDICINE, Stroke < NEUROLOGY, 17 Keywords: SOCIAL MEDICINE, PUBLIC HEALTH, Organisation of health services < 18 HEALTH SERVICES ADMINISTRATION & MANAGEMENT 19 20 21 22 23 24 25 26 27 28 29 30 31

32 http://bmjopen.bmj.com/ 33 34 35 36 37 38 39

1 2 3

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 3 of 40 BMJ Open

1 1 2 3 1 Dyadic approach to supervised community rehabilitation 4 BMJ Open: first published as 10.1136/bmjopen-2019-036631 on 23 April 2020. Downloaded from 5 2 6 participation in an Asian setting post-stroke: exploring the role of 7 3 8 caregiver and patient characteristics in a prospective cohort study 9 4 1 1 1 10 Shilpa TYAGI , Gerald Choon-Huat KOH *, Nan LUO , Kelvin Bryan 11 2 3 4,5 1 12 5 TAN , Helen HOENIG , David B. MATCHAR , Joanne YOONG , 13 4 6 7 14 6 Angelique CHAN , Kim En LEE , N. VENKETASUBRAMANIAN , 15 8 9 10 16 7 Edward MENON , Kin Ming CHAN , Deidre Anne DE SILVA , Philip 17 8 11 12 13 14 18 YAP , Boon ForYeow TANpeer, Effie review CHEW , Sherry only H. YOUNG , Yee 19 15 16 17 18 20 9 Sien NG , Tian Ming TU , Yan Hoon ANG , Keng He KONG , 21 16 19 10 22 10 Rajinder SINGH , Reshma A. MERCHANT , Hui Meng CHANG , 23 20 20 1 24 11 Tseng Tsai YEO , Chou NING , Angela CHEONG , Chuen Seng 25 1 26 12 TAN , PhD 27 13 *Corresponding Author 28 29 14 30 31 15 1Saw Swee Hock School of Public Health, National University of Singapore, 32 33 16 Singapore, Singapore; 2Policy Research & Economics Office, Ministry of Health, 34 3 35 17 Singapore, Singapore; Physical Medicine and Rehabilitation Service, Durham VA 36 18 Medical Centre, USA; 4Program in Health Services and Systems Research, Duke- http://bmjopen.bmj.com/ 37 38 19 NUS Graduate Medical School, Singapore, Singapore; 5Department of Medicine 39 40 20 (General Internal Medicine), Duke University Medical Center, Durham, NC, USA; 41 6 7 42 21 Lee Kim En Neurology Pte Ltd, Singapore, Singapore; Raffles Neuroscience 43 22 Centre, Raffles Hospital, Singapore, Singapore; 8St. Andrew’s Community Hospital, 44 on September 28, 2021 by guest. Protected copyright. 45 23 Singapore, Singapore; 9Mount Alvernia Hospital, Singapore, Singapore; 10National 46 47 24 Neuroscience Institute, Singapore General Hospital campus, Singapore, Singapore; 48 25 11Geriatric Centre, Khoo Teck Puat Hospital, Singapore, Singapore; 12St. Luke's 49 50 26 Hospital, Singapore, Singapore; 13Department of Rehabilitation Medicine, National 51 52 27 University Hospital, Singapore, Singapore; 14Department of Rehabilitation Medicine, 53 15 54 28 Changi General Hospital, Singapore, Singapore; Department of Rehabilitation 55 29 Medicine, Singapore General Hospital, Singapore, Singapore; 16Department of 56 57 30 Neurology, National Neuroscience Institute, Tan Tock Seng Hospital, Singapore, 58 17 59 31 Singapore; Geriatric Medicine, Khoo Teck Puat Hospital, Singapore, Singapore; 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 4 of 40

2 1 2 3 32 18Department of Rehabilitation Medicine, Tan Tock Seng Hospital, Singapore, 4 BMJ Open: first published as 10.1136/bmjopen-2019-036631 on 23 April 2020. Downloaded from 5 33 Singapore; 19Department of Medicine, Yong Loo Lin School of Medicine, National 6 20 7 34 University of Singapore, Singapore, Singapore; Department of Neurosurgery, 8 35 National University Hospital, Singapore, Singapore. 9 10 36 11 12 37 *Corresponding author 13 14 38 Dr Gerald CH Koh 15 39 Saw Swee Hock School of Public Health 16 17 40 National University of Singapore 18 For peer review only 19 41 12 Science Drive 2 20 21 42 #10-01 22 43 Singapore 117549 23 24 44 Tel (65) 6516 4979 25 26 45 Fax (65) 6779 1489 27 46 Email: [email protected] 28 29 47 Word count = 4450 [from introduction to conclusion] 30 31 32 33 34 35 36 http://bmjopen.bmj.com/ 37 38 39 40 41 42 43 44 on September 28, 2021 by guest. Protected copyright. 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 5 of 40 BMJ Open

3 1 2 3 49 4 BMJ Open: first published as 10.1136/bmjopen-2019-036631 on 23 April 2020. Downloaded from 5 6 50 ABSTRACT 7 8 9 51 Objective: To study the association of caregiver factors and stroke survivor 10 52 factors with supervised community rehabilitation (SCR) participation over the 11 12 53 first 3 months and subsequent 3-12 months post-stroke in an Asian setting. 13 14 15 54 Design: prospective cohort study 16 17 18 55 Setting: communityFor setting peer review only 19 20 56 Participants: We recruited stroke survivors and their caregivers into our yearlong 21 22 57 cohort. Caregiver and stroke survivor variables were collected over 3-monthly 23 58 intervals. We performed logistic regression with the outcome variable being SCR 24 25 59 participation post-stroke. 26 27 60 Outcome measures: SCR participation over the first 3 months and subsequent 3-12 28 29 61 months post-stroke 30 62 Results: 251 stroke survivor-caregiver dyads were available for the current analysis. 31 32 63 The mean age of caregivers was 50.1 years, with the majority being female, married, 33 34 64 and co-residing with the stroke survivor. There were 61%, 28%, 4% and 7% of 35 65 spousal, adult-child, sibling and others as caregivers. The odds of SCR participation 36 http://bmjopen.bmj.com/ 37 66 decreased by about 15% for every unit increase in caregiver-reported stroke 38 39 67 survivor’s disruptive behaviour score (OR: 0.845; 95% CI: 0.769, 0.929). For every 1- 40 41 68 unit increase in the caregiver’s positive management strategy score, the odds of using 42 69 SCR service increased by about 4% (OR: 1.039; 95% CI: 1.011, 1.068). 43 44 on September 28, 2021 by guest. Protected copyright. 45 70 Conclusion: We established that SCR participation is jointly determined by 46 47 71 both caregiver and stroke survivor factors, with factors varying over the early 48 49 72 and late post-stroke period. Our results support the adoption of a dyadic or more 50 73 inclusive approach for studying the utilization of community rehabilitation 51 52 74 services, giving due consideration to both the stroke survivors and their 53 54 75 caregivers. Adopting a stroke survivor-caregiver dyadic approach in practice 55 76 settings should include promotion of positive care management strategies, 56 57 77 comprehensive caregiving training including both physical and behavioural 58 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 6 of 40

4 1 2 3 78 dimensions, active engagement of caregivers in rehabilitation journey and 4 BMJ Open: first published as 10.1136/bmjopen-2019-036631 on 23 April 2020. Downloaded from 5 79 conducting regular caregiver needs assessments in the community. 6 7 8 80 Keywords: Stroke rehabilitation, health services, stroke, family caregivers 9 81 10 11 82 12 13 83 14 15 84 16 85 Article summary 17 18 86 - We studiedFor the association peer of reviewcaregiver factors onlyalong with stroke survivor 19 20 87 factors with supervised community rehabilitation participation over the first 3 21 22 88 months and subsequent 3-12 months post-stroke in a prospective yearlong 23 89 cohort study. 24 25 90 - We are among the first to demonstrate the role of caregivers in stroke 26 27 91 survivor’s supervised community rehabilitation substantiating the rationale for 28 92 the adoption of a stroke survivor-caregiver dyadic approach to studying post- 29 30 93 stroke outcomes. 31 32 94 - Another strength is the comprehensiveness of caregiver variables considered, 33 34 95 which enabled us to explore the role of caregivers in-depth. 35 96 - Our study sample included patients with stroke surviving the first post-stroke 36 http://bmjopen.bmj.com/ 37 97 year, excluding deaths within the follow-up period (<5%) limiting the 38 39 98 generalizability of our findings to those stroke survivors who are alive at the 40 41 99 end of the first year post-stroke. 42 100 - There is a possibility of information bias related to the limited recall of 43 44

101 supervised community rehabilitation participation by patients with stroke and on September 28, 2021 by guest. Protected copyright. 45 46 102 their caregivers, which was addressed by keeping a relatively shorter recall 47 103 period. 48 49 104 50 51 105 52 53 106 List of abbreviations 54 107 CES-D: Centre for Epidemiological Studies Depression scale; DRCs: day 55 56 108 rehabilitation centres; FDW: foreign domestic worker; mRS: modified Rankin scale; 57 58 109 NIHSS: National Institute of Health Scale; rDMSS: Revised Dementia Management 59 110 Strategies scale; SCR: supervised community rehabilitation. 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 7 of 40 BMJ Open

5 1 2 3 111 4 BMJ Open: first published as 10.1136/bmjopen-2019-036631 on 23 April 2020. Downloaded from 5 6 7 112 INTRODUCTION 8 113 Stroke is associated with a significant mortality burden globally.(1) However, recent 9 10 114 epidemiological trends with an increased incidence in the younger population and 11 12 115 decreasing mortality rates over the years highlight the importance of functional 13 14 116 recovery post-stroke.(2) While rehabilitation is essential for functional recovery and 15 117 re-integration back in the community, it is conditional on patients with stroke taking 16 17 118 the initiative to attend such rehabilitation services after a stroke event. The transition 18 For peer review only 19 119 from inpatient settings into the community is challenging for the stroke survivor- 20 21 120 caregiver dyad.(3) They move from a well-supported setting with a multi-disciplinary 22 121 team providing care and facilitating rehabilitation to a setting where they are on their 23 24 122 own trying to maintain the care continuum and seeking community services including 25 26 123 supervised community rehabilitation (SCR). This challenging transition is further 27 124 described by the concept of the duality of stroke crisis with first crisis coinciding with 28 29 125 the stroke and the second occurring during the discharge from an inpatient setting to 30 31 126 home.(4) During this second crisis, stroke survivors often feel unprepared and 32 33 127 overwhelmed to navigate post-stroke recovery journey. Many stroke survivors rely on 34 128 family caregivers’ assistance to continue their recovery journey and re-integration 35 36 129 back into community. http://bmjopen.bmj.com/ 37 38 130 39 40 131 The role of family caregivers becomes highly relevant post-stroke, considering more 41 132 than half of the stroke survivors are discharged home with differing degrees of 42 43 133 residual physical impairments.(5) However, neither is this caregiving role explicitly 44 on September 28, 2021 by guest. Protected copyright. 45 134 acknowledged nor is the caregiver’s capacity and commitment to providing care 46 135 assessed.(6) This could result in a mismatch between the expected caregiving 47 48 136 responsibilities and the caregiver’s ability to fulfil these, potentially leading to 49 50 137 adverse consequences for the stroke survivor-caregiver dyad. High reliance on 51 52 138 caregivers to assist the stroke survivors in the community implies that caregiver 53 139 factors like coping, perceived stress, available support system and well-being can in 54 55 140 turn, influence the stroke survivor’s outcomes like participation in supervised 56 57 141 rehabilitation. This highlights the importance of adopting a stroke survivor-caregiver 58 59 142 dyadic approach to studying and implementing supervised rehabilitation in the 60 143 community, which includes provision of physical rehabilitation services by licensed

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 8 of 40

6 1 2 3 144 physiotherapists or occupational therapists in the community settings, such as day 4 BMJ Open: first published as 10.1136/bmjopen-2019-036631 on 23 April 2020. Downloaded from 5 145 rehabilitation centres (DRCs) or patient’s home. Dyadic approach is described as a 6 7 146 “holistic approach to post-stroke care provision by healthcare practitioners, giving due 8 147 importance to both patients with stroke and their caregivers, integrating caregivers in 9 10 148 the healthcare system to extend the care continuum to include informal care in the 11 12 149 community and provision of timely support for caregivers.”(7) Prior accounts in both 13 14 150 stroke(7-10) and non-stroke(11-13) population have included such stroke survivor- 15 151 caregiver dyadic approach in their narratives. In addition, giving due consideration to 16 17 152 both the stroke survivors and their caregivers in psycho-educational, skill-building 18 For peer review only 19 153 and support interventions is reported to improve stroke survivor outcomes.(8) 20 21 154 Recognizing the relevance of caregivers and caregiving context in stroke survivor’s 22 155 recovery process, researchers have attempted to study the association of caregiver 23 24 156 availability and some socio-demographic characteristics with functional outcomes 25 26 157 post-stroke in patients attending inpatient rehabilitation services.(14-18) While a 27 158 study in the US(19) reported a positive role of the spouse in the post-stroke recovery 28 29 159 of stroke survivors, another study in Canada(15) reported caregiver support is 30 31 160 associated with a higher functional gain as compared to those without caregiver 32 33 161 support. A recent study in China explored the role of family member’s positive and 34 162 negative attitudes in the post-stroke functional and cognitive recovery of stroke 35 36 163 survivors, with more positive attitudes being associated with more cognitive http://bmjopen.bmj.com/ 37 38 164 functional gains after rehabilitation.(20) Increasingly over the years, literature seems 39 40 165 to be in favour of caregivers playing an important role in stroke survivors’ functional 41 166 and cognitive outcomes post-rehabilitation across different settings and contexts. 42 43 167 However, none of the studies so far have focused on the association of caregiver 44 on September 28, 2021 by guest. Protected copyright. 45 168 characteristics with SCR participation. 46 169 Addressing the above mentioned gaps, we aimed to study the association of caregiver 47 48 170 factors along with stroke survivor factors with SCR participation over the first 3 49 50 171 months and subsequent 3-12 months post-stroke in an Asian setting. 51 52 53 172 MATERIALS AND METHODS 54 55 173 Study setting 56 57 174 In Singapore, after stabilization in a tertiary hospital, patients with stroke are assessed 58 59 175 for rehabilitation eligibility, and based on this assessment, they may undergo intensive 60 176 rehabilitation in an inpatient setting.(21) Another option either in succession to above

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 9 of 40 BMJ Open

7 1 2 3 177 or as an alternative is SCR, often delivered at the DRCs. DRCs are run either within 4 BMJ Open: first published as 10.1136/bmjopen-2019-036631 on 23 April 2020. Downloaded from 5 178 the premises of a step-down facility or as stand-alone centres, mainly providing 6 7 179 physiotherapy and occupational therapy.(22) DRCs along with nursing homes (i.e., 8 180 long-term residential care settings within the community) and home-based 9 10 181 rehabilitation fall under the broad umbrella of SCR. 11 12 13 182 Participants 14 15 183 Our participants were part of the Singapore Stroke Study, a prospective observational 16 17 184 study with recruitment of stroke survivors and caregivers over a period extending 18 For peer review only 19 185 from December 2010 to September 2013. Singaporeans or permanent residents 40 20 21 186 years and above who suffered a stroke or experienced symptoms within 4 weeks of 22 187 admission to any of the five tertiary hospitals in Singapore during the recruitment 23 24 188 period with confirmed stroke diagnosis were recruited along with their caregivers. 25 26 189 Caregivers could be an immediate or extended family member or friend who provided 27 190 care or assistance of any kind and took the responsibility for the stroke survivor and 28 29 191 were recognized by the stroke survivor, not fully paid for caregiving. The on-site 30 31 192 research nurses reviewed the list of patients with stroke on a daily basis to screen for 32 33 193 eligible participants. They conducted recruitment from screened eligible stroke 34 194 survivors across all five tertiary hospitals in Singapore during the study period 35 36 195 increasing the representativeness of the sample. All participants (i.e., stroke survivors http://bmjopen.bmj.com/ 37 38 196 and their caregivers) were explained the study purpose and procedures in their 39 40 197 preferred language, and written informed consent was taken and documented. 41 198 Participants were informed that they could withdraw from the study at any point 42 43 199 during the follow-up period of one year, if they wished. 44 on September 28, 2021 by guest. Protected copyright. 45 200 46 201 Data was collected at 3-monthly intervals, via in-person interviews at baseline, 3- 47 48 202 month and 12-month time points, and via telephone interviews at 6-month and 9- 49 50 203 month time points. Trained interviewers collected data under three broad domains 51 52 204 covering health, social and financial factors. Several measures were taken to ensure 53 205 good compliance and minimize attrition, such as sending reminders prior to scheduled 54 55 206 interviews, scheduling interviews over weekends or evenings during weekdays, 56 57 207 multiple contact attempts (upto 3) before categorizing as lost to follow-up. To ensure 58 59 208 the standardization and high quality of data collection, main investigators engaged in 60 209 the training of research assistants. The training sessions were video-recorded, and

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 10 of 40

8 1 2 3 210 these recordings were used to facilitate the training sessions of all the research 4 BMJ Open: first published as 10.1136/bmjopen-2019-036631 on 23 April 2020. Downloaded from 5 211 assistants covering the content and appropriate method of data collection along with 6 7 212 consent taking procedures. Since our study enrolled participants from multi-ethnic 8 213 setting of Singapore without any language related restriction on recruitment, our 9 10 214 research assistants were proficient in different languages of data collection (e.g., 11 12 215 English, Mandarin, Malay or Tamil), and conducted interviews in the language 13 14 216 preferred by the participants. Before embarking on the data collection, we pilot tested 15 217 our survey on 40 participants from two of the five sites and finalized the survey forms 16 17 218 after inclusion of necessary amendments. Further details about the Singapore Stroke 18 For peer review only 19 219 Study are reported elsewhere.(7, 23) The Singapore Stroke Study was approved by 20 21 220 the National University of Singapore Institutional Review Board, SingHealth 22 221 Centralized Institutional Review Board and the National Health Group Domain 23 24 222 Specific Review Board. Written informed consent was obtained from both the 25 26 223 patients and the caregivers in their preferred language by trained researchers. 27 28 29 224 Independent variables (caregiver) 30 31 32 225 With a primary focus on caregiver factors, following caregiver variables were 33 226 considered for current analysis: socio-demographic characteristics, marital status, 34 35 227 relationship with caregiver (caregiver being spouse, adult-child, sibling or others 36 http://bmjopen.bmj.com/ 37 228 including distant relatives and friends), number of chronic ailments, co-residing 38 39 229 status, caregiver burden, family conflict, social support, caregiver reported stroke 40 230 survivor behavioural issues and adopted caregiver management approaches. Under 41 42 231 the caregiving burden, we incorporated measures of both objective and subjective 43 44 232 burden measured by the Oberst Caregiving Burden Scale(24) and the Zarit’s Burden on September 28, 2021 by guest. Protected copyright. 45 233 Interview(25, 26) respectively. The family caregiving conflict scale recommended by 46 47 234 Pearlin and colleagues was used to capture family conflict.(27) Adapting Pearlin and 48 49 235 colleagues’ description of the construct of social support, we tried to incorporate both 50 51 236 “instrumental” and “expressive” dimensions of social support, with former captured 52 237 by presence of paid help or foreign domestic worker (FDW) for general household 53 54 238 tasks or specifically for stroke survivor and latter captured by Pearlin’s 8-item 55 56 239 perceived social support instrument.(27) The caregiver reported occurrence of 57 58 240 problematic behaviour by stroke survivors was recorded using the Revised Memory 59 241 and Behavioural Problem Checklist, which has been previously used in stroke 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 11 of 40 BMJ Open

9 1 2 3 242 survivors.(28-31) Caregivers were asked whether any of the 21 problematic 4 BMJ Open: first published as 10.1136/bmjopen-2019-036631 on 23 April 2020. Downloaded from 5 243 behaviours (e.g., “asking the same question over and over”, “destroying property”, 6 7 244 “crying and tearfulness” etc.) have occurred during the previous week. Responses 8 245 were recorded on a 5-point Likert scale: 0 = never, 1 = not in the past week, 2 = 1 to 2 9 10 246 times per week, 3 = 3 to 6 times per week and 4 = daily or more often.(32) Separate 11 12 247 summated scores were calculated across the 3 domains of disruptive, depressive and 13 14 248 memory related behavioural problems with Cronbach’s alpha for each domain being 15 249 0.73, 0.87 and 0.90 respectively. To capture the care management or coping 16 17 250 approaches by stroke survivor’s caregivers, we used the revised dementia 18 For peer review only 19 251 management strategies scale (rDMSS). Previously validated in Singapore,(33) the 20 21 252 scale has two sub-components of positive and negative types of management 22 253 strategies with good reported internal consistency (Cronbach’s alpha 0.89 and 0.87 23 24 254 respectively). 25 26 27 255 Independent variables (stroke survivor) 28 29 30 256 Following baseline stroke survivor variables were considered: socio-demographic 31 32 257 characteristics, marital status, ward class as a proxy of socio-economic status, 33 258 Charlson Comorbidity Index, type of stroke (ischemic or non-ischemic), recurrent or 34 35 259 first stroke, stroke severity measured on National Institute of Health Scale or NIHSS, 36 http://bmjopen.bmj.com/ 37 260 functional status measured on modified Rankin scale or mRS, impairment in 38 39 261 cognition measured on the Mini-Mental State examination or MMSE, discharge status 40 262 and depression measured on the 11-item version of the Centre for Epidemiological 41 42 263 Studies Depression scale (CES-D). Ward class captured the category of the ward in 43 44 264 which the stroke survivor stayed during the index hospitalization. To make healthcare on September 28, 2021 by guest. Protected copyright. 45 265 affordable for all, the Singaporean government subsidizes inpatient stay in the tertiary 46 47 266 care setting in a tiered manner. Based on the financial or means testing, the patients 48 49 267 can be eligible to stay at A, B1, B2 or C ward types, being entitled to increasing level 50 51 268 of subsidies. With quality of care remaining constant across all ward types, they 52 269 usually differ in the type of amenities provided to the warded patients. For the current 53 54 270 analysis, we categorized ward class into subsidized and non-subsidized categories.(7) 55 56 271 For scales with more than 10 missing cases (NIHSS, MMSE, Revised memory and 57 58 272 behaviour checklist), we used the person mean substitution approach to impute for 59 273 missing values for cases with less than half constituting items missing.(34) 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 12 of 40

10 1 2 3 274 Outcome variables 4 BMJ Open: first published as 10.1136/bmjopen-2019-036631 on 23 April 2020. Downloaded from 5 6 275 The outcome of interest was SCR participation, which comprised of participation at 7 8 276 any of the following: rehabilitation at home, DRC or nursing homes. This information 9 277 was captured in the survey at 3 monthly intervals by asking the caregiver, “Has the 10 11 278 stroke patient at any time during the last 3 months received rehabilitation? Please 12 13 279 include any rehabilitation at home, DRC and nursing homes”. For the subsequent 3 to 14 15 280 12 months, we created a variable for capturing the SCR participation information 16 281 collected at 6, 9 and 12-month interviews. It was coded as yes if the caregiver 17 18 282 reported any participationFor peer at either of review the time points and only no if the caregiver reported 19 20 283 no participation across all time points. 21 22 23 284 Data analysis 24 25 26 285 Univariate analysis was conducted to describe the caregiver and stroke survivor 27 286 characteristics. We conducted bivariate analysis to examine the associations between 28 29 287 independent variables (caregiver and stroke survivor factors) and the SCR 30 31 288 participation. The independent variables with p values < 0.1 on bivariate analysis 32 33 289 were chosen as potential predictors for the multivariable regression model. With these 34 290 potential predictor variables, we built the most parsimonious model using a backward 35 36 291 variable selection approach. At each model building stage, the most insignificant http://bmjopen.bmj.com/ 37 38 292 variable was removed until we were left with the variables having a p-value < 0.05, 39 40 293 except for age, gender, ethnicity and ward class of stroke survivors which were kept 41 294 in the model. Logistic regression was used at both bivariate and model building stages 42 43 295 and we reported the unadjusted and adjusted odds ratio (OR) estimates with 95% 44 on September 28, 2021 by guest. Protected copyright. 45 296 confidence intervals (CI). We ran separate models for the SCR participation across 46 297 two time periods of the first 3 months post-stroke and subsequent 3 to 12 months as 47 48 298 researchers have previously reported variations in the determinants of stroke 49 50 299 survivor’s outcomes over these two periods.(7, 35) The significance level was set at 51 52 300 5%. With the most parsimonious model, we performed diagnostics for the model fit 53 301 using Hosmer and Lemeshow’s goodness-of-fit test, checked for model 54 55 302 misspecifications, multi-collinearity and influential observations. All analysis was 56 57 303 performed in Stata version 14.1.(36) 58 59 304 60 305 Patient and Public Involvement

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 13 of 40 BMJ Open

11 1 2 3 306 This research was done without patient involvement. Patients were not invited to 4 BMJ Open: first published as 10.1136/bmjopen-2019-036631 on 23 April 2020. Downloaded from 5 307 comment on the study design and were not consulted to develop patient relevant 6 7 308 outcomes or interpret the results. Patients were not invited to contribute to the writing 8 309 or editing of this document for readability or accuracy. 9 10 11 12 310 RESULTS 13 14 311 Out of the 661 caregivers assessed for recruitment at baseline, 399 caregivers were 15 16 312 recruited after exclusion of 190 caregivers and 72 stroke survivors not having a 17 18 313 caregiver. Two Forhundred fifty-onepeer stroke review survivor-caregiver only dyads were available for 19 20 21 314 the current analysis after exclusion of stroke survivors with deaths within the follow 22 23 315 up period and limiting to complete cases. (Please refer figure 1 for study flowchart) 24 25 316 The follow-up rates for caregivers were 87.2% and 73.4% at 3 and 12 months. The 26 27 317 prevalence of any SCR participation was 49% over 0-3 month and 25% over 3-12 28 29 30 318 month time periods. The mean age of caregivers was 50.1 years, with the majority 31 32 319 being female, married and co-residing with the stroke survivor. There were 61%, 33 34 320 28%, 4% and 7% of spousal, adult-child, sibling and others as caregivers. The mean 35 36 http://bmjopen.bmj.com/ 37 321 scores for summated values for memory related, depressive and disruptive behaviour 38 39 322 problems of stroke survivors were 5.01, 3.15 and 2.67 respectively. 34.27 (10.85) and 40 41 323 11.07 (4.60) were the mean (SD) scores for caregiver reported positive and negative 42 43 44 324 care management strategies respectively. The stroke survivors had a mean age of 61.8 on September 28, 2021 by guest. Protected copyright. 45 46 325 years, with the majority being male (65%), of Chinese ethnicity (59%) and married 47 48 326 (81%). About 89% had ischemic index stroke and for 17%, the index stroke was a 49 50 327 recurrent one. Out of all stroke survivors, 57%, 38% and 5% had a mild, moderately 51 52 53 328 severe and severe type of index stroke as measured on NIHSS. More than half (59%) 54 55 329 had moderate to severe disability (3-5) on MRS. (Please refer table 1 and table 2 for 56 57 330 baseline characteristics) The findings from the diagnostics for model fit using Hosmer 58 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 14 of 40

12 1 2 3 331 and Lemeshow’s goodness-of-fit test are as follows: for 0-3 months Model: p-value= 4 BMJ Open: first published as 10.1136/bmjopen-2019-036631 on 23 April 2020. Downloaded from 5 6 332 0.663; and for 3-12 months Model: p-value= 0.778. 7 8 333 9 10 334 11 335 [Figure 1 near here] 12 13 336 14 15 337 Table 1. Descriptive characteristics of participants (caregiver factors) 16 17 All stroke Participated in Participated 18 For peer review only 19 * * survivors , SCR , in SCR †, 20 21 No. (%) 3 months 22 3-12 months 23 No. (%) 24 No. (%) 25 26 27 Variable Category 28 29 CAREGIVER FACTORS 30 31 Age of caregiver (in years) 50.13 (13.09) 49.03 (12.97) 51.16 (10.76) 32 33 mean (SD) 34 35 Gender of caregiver Male 61 (23) 29 (24) 9 (16) 36 http://bmjopen.bmj.com/ 37 Female 199 (77) 93 (76) 49 (84) 38 39 Ethnicity of caregiver Chinese 151 (58) 66 (54) 35 (60) 40 41 Non-Chinese 109 (42) 56 (46) 23 (40) 42 43 Marital Status of caregiver Married 205 (79) 93 (76) 47 (81) 44 on September 28, 2021 by guest. Protected copyright. 45 46 Single 55 (21) 29 (24) 11 (19) 47 48 Caregiver identity Spouse 159 (61) 71 (58) 32 (55) 49 50 Adult-child 74 (28) 36 (30) 17 (29) 51 52 Sibling 10 (4) 6 (5) 4 (7) 53 54 Others 17 (7) 9 (7) 5 (9) 55 56 Co-residing with patient Yes 231 (89) 109 (89) 49 (84) 57 58 No 29 (11) 13 (11) 9 (16) 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 15 of 40 BMJ Open

13 1 2 3 Caring for multiple care Yes 108 (42) 49 (40) 22 (38) 4 BMJ Open: first published as 10.1136/bmjopen-2019-036631 on 23 April 2020. Downloaded from 5 recipients 6 7 No 152 (58) 73 (60) 36 (62) 8 9 10 Revised memory and behaviour checklist 11 12 Memory problems mean (SD) 5.01 (5.93) 4.95 (6.30) 5.09 (6.06) 13 14 Depressive behaviour mean (SD) 3.15 (4.82) 2.72 (4.21) 3.61 (5.21) 15 16 problems 17 18 Disruptive behaviourFor peermean (SD) review2.67 (3.62) only1.97 (2.86) 2.10 (2.83) 19 20 problems 21 22 Caregiver Burden 23 24 Oberst Caregiving Burden mean (SD) 31.71 (12.63) 32.94 (12.02) 35.18 (13.20) 25 26 Scale 27 28 29 Zarit Burden Interview mean (SD) 8.73 (7.86) 8.25 (6.83) 8.74 (8.13) 30 31 Family conflict 32 33 Attitude towards patient mean (SD) 11.42 (4.49) 11.58 (4.46) 12.00 (4.12) 34 35 Attitude towards caregiver mean (SD) 11.63 (4.37) 11.68 (4.40) 12.26 (3.91) 36 http://bmjopen.bmj.com/ 37 Social Support (instrumental) 38 39 FDW for general help Yes 212 (82) 98 (80) 42 (72) 40 41 No 48 (18) 24 (20) 16 (28) 42 43 FDW for stroke survivor Yes 33 (13) 17 (14) 11 (19) 44 on September 28, 2021 by guest. Protected copyright. 45 No 227 (87) 105 (86) 47 (81) 46 47 48 Social Support (perceived) mean (SD) 26.33 (4.90) 26.49 (5.03) 25.95 (4.61) 49 50 Care Management Strategies 51 52 Positive strategies mean (SD) 34.27 (10.85) 36.52 (10.19) 35.88 (10.92) 53 54 Negative strategies mean (SD) 11.07 (4.60) 10.56 (4.34) 11.36 (4.46) 55 56 57 58 *N = 251, †N=238 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 16 of 40

14 1 2 3 Abbreviations: SCR: supervised community rehabilitation; No.: number; SD: standard deviation; FDW: foreign 4 BMJ Open: first published as 10.1136/bmjopen-2019-036631 on 23 April 2020. Downloaded from 5 domestic worker 6 7 338 8 9 10 339 Table 2. Descriptive characteristics of participants (stroke survivor factors) 11 12 All stroke Participated in Participated 13 * * 14 survivors , SCR , in SCR †, 15 16 No. (%) 3 months 17 3-12 months 18 For peer review onlyNo. (%) 19 No. (%) 20 21 Variable Category 22 23 PATIENT FACTORS 24 25 Age of patient mean (SD) 61.77 (10.42) 60.86 (10.63) 62.29 (10.53) 26 27 28 (in years) 29 30 Gender of patient Male 169 (65) 77 (63) 37 (64) 31 32 Female 91 (35) 45 (37) 21 (36) 33 34 Ethnicity of patient Chinese 153 (59) 65 (53) 35 (60) 35 36 Non-Chinese 107 (41) 57 (47) 23 (40) http://bmjopen.bmj.com/ 37 38 Marital Status of patient Married 210 (81) 97 (80) 45 (78) 39 40 Single 50 (19) 25 (20) 13 (22) 41 42 Ward Class Unsubsidized 21 (8) 11 (9) 7 (12) 43 44 Subsidized 235 (92) 110 (91) 51 (88) on September 28, 2021 by guest. Protected copyright. 45 46 47 Charlson Comorbidity 1 - 3 52 (20) 22 (18) 13 (22) 48 49 Index 50 51 4 - 6 165 (63) 79 (65) 31 (53) 52 53 >= 7 43 (17) 21 (17) 14 (24) 54 55 Stroke type Ischemic 231 (89) 106 (87) 48 (83) 56 57 Non-ischemic 29 (11) 16 (13) 10 (17) 58 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 17 of 40 BMJ Open

15 1 2 3 Recurrent stroke Yes 43 (17) 16 (13) 8 (14) 4 BMJ Open: first published as 10.1136/bmjopen-2019-036631 on 23 April 2020. Downloaded from 5 No 217 (83) 106 (87) 50 (86) 6 7 National Institute of Health Mild (0-4) 149 (57) 60 (49) 23 (40) 8 9 10 Scale 11 12 Moderately 97 (38) 55 (45) 28 (48) 13 14 severe (5-14) 15 16 Severe (15-24) 14 (5) 7 (6) 7 (12) 17 18 Modified RankinFor Scale peerNo or slight review 106 (41) only38 (31) 11 (19) 19 20 disability (0-2) 21 22 Moderate or 154 (59) 84 (69) 47 (81) 23 24 severe disability 25 26 (3-5) 27 28 29 Mini-Mental State No cognitive 150 (58) 70 (57) 30 (52) 30 31 Examination impairment (24- 32 33 30) 34 35 Mild cognitive 65 (25) 33 (27) 15 (26) 36 http://bmjopen.bmj.com/ 37 impairment (18- 38 39 23) 40 41 Severe cognitive 45 (17) 19 (16) 13 (22) 42 43 impairment (1- 44 on September 28, 2021 by guest. Protected copyright. 45 17) 46 47 48 Discharge to Step-down Yes 66 (25) 33 (27) 19 (33) 49 50 facility (Community 51 52 Hospital) 53 54 No 194 (75) 89 (73) 39 (67) 55 56 Centre for Epidemiological mean (SD) 6.31 (5.61) 6.59 (5.51) 6.60 (5.64) 57 58 Studies Depression Scale 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 18 of 40

16 1 2 3

4 BMJ Open: first published as 10.1136/bmjopen-2019-036631 on 23 April 2020. Downloaded from 5 *N = 251, †N=238 6 7 8 Abbreviations: SCR: supervised community rehabilitation; No.: number; SD: standard deviation 9 10 340 11 12 13 341 14 15 16 342 Supervised community rehabilitation participation (0-3 months post- 17 18 343 stroke) For peer review only 19 20 344 Supplement table 1 depicts the results of the association of caregiver and patient 21 22 345 characteristics with odds of SCR participation across 3 months post-stroke. The 23 24 346 bivariate association of caregiver reported patient disruptive behaviour and positive 25 26 347 management strategy with SCR participation was statistically significant. Among the 27 348 patient factors, the bivariate association of the following variables with odds of SCR 28 29 349 participation was statistically significant: stroke severity and functional status. The 30 31 350 variables that entered the final adjusted model of odds of SCR participation in the first 32 351 3 months post-stroke were caregiver reported disruptive behaviour of the patient, 33 34 352 positive management strategy by the caregiver and patient’s functional status (Please 35 36 353 refer table 3). For every 1-unit increase in the caregiver reported stroke survivor’s http://bmjopen.bmj.com/ 37 38 354 disruptive behaviour score, the odds of SCR participation decreased by about 15% 39 355 (OR: 0.845; 95% CI: 0.769, 0.929). For every 1-unit increase in the caregiver reported 40 41 356 positive care management strategy score, the odds of SCR participation increased by 42 43 357 about 4% (OR: 1.039; 95% CI: 1.011, 1.068). Compared to stroke survivors with no 44 on September 28, 2021 by guest. Protected copyright. 45 358 or mild functional disability, those with moderate to severe functional disability had 46 359 2.76 times the odds of SCR participation over the first 3 months post-stroke 47 48 360 (p=0.001). 49 50 361 51 52 362 Table 3. Multi-variable regression analysis results of supervised community 53 54 363 rehabilitation (any use) across 3 months post-stroke. 55 Any Rehabilitation (3 months) 56 57 58 Variable Reference category aOR (95% CI) P-value 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 19 of 40 BMJ Open

17 1 2 3 Revised memory and behaviour checklist 4 BMJ Open: first published as 10.1136/bmjopen-2019-036631 on 23 April 2020. Downloaded from 5 Disruptive behaviour problems 0.845 (0.769, 0.929) <0.001 6 7 Care management strategies 8 9 10 Positive strategies 1.039 (1.011, 1.068) 0.006 11 12 Modified Rankin Scale No or slight 0.001 13 14 disability (0-2) 15 16 Moderate or severe disability 2.759 (1.532, 4.969) 17 18 (3-5) For peer review only 19 20 Age (in years)* 0.975 (0.949, 1.002) 0.074 21 22 Gender* Male 0.875 (0.480, 1.594) 0.662 23 24 Ethnicity* Non-Chinese 0.723 (0.409, 1.277) 0.264 25 26 Ward Class* Unsubsidized 0.806 (0.298, 2.180) 0.671 27 28 Abbreviations: aOR: adjusted odds ratio; CI: confidence interval 29 30 *Model adjusted for patient’s age, gender, ethnicity and ward class 31 32 33 364 34 35 36 365 Supervised community rehabilitation participation (3-12 months post- http://bmjopen.bmj.com/ 37 38 366 stroke) 39 40 41 367 Supplement table 2 depicts the results of the association of caregiver and patient 42 368 characteristics with odds of SCR participation across 3-12 months post-stroke. The 43 44 369 bivariate association of caregiver burden (measured on Oberst Burden Scale), FDW on September 28, 2021 by guest. Protected copyright. 45 46 370 for general help and FDW for stroke patient with SCR participation were statistically 47 48 371 significant. Among the patient factors, the bivariate association of the following 49 372 variables with SCR participation was statistically significant: stroke severity and 50 51 373 functional status. The only variable that entered the final adjusted model of odds of 52 53 374 SCR participation 3-12 months post-stroke was the functional status, with odds of 54 55 375 SCR participation being 4.19 times in those with moderate to severe functional 56 376 disability when compared to those with none to mild disability (OR: 4.234; 95% CI: 57 58 377 2.034, 8.812) (Please refer table 4). 59 60 378

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 20 of 40

18 1 2 3 379 Table 4. Multi-variable regression analysis results of supervised community 4 BMJ Open: first published as 10.1136/bmjopen-2019-036631 on 23 April 2020. Downloaded from 5 380 rehabilitation (any use) across 3 – 12 months post-stroke. 6 7 Any Rehabilitation (3-12 months) 8 9 Variable Reference category aOR* (95% CI) P-value 10 11 Modified Rankin No or slight disability (0-2) <0.001 12 13 Scale 14 15 Moderate or severe 4.234 (2.034, 8.812) 16 17 disability (3-5) 18 For peer review only 19 Age (in years)* 0.999 (0.968, 1.032) 0.973 20 21 Gender* 0.875 (0.448, 1.709) 0.695 22 23 24 Ethnicity* 1.146 (0.603, 2.176) 0.678 25 26 Ward Class* 0.432 (0.146, 1.279) 0.130 27 28 Abbreviations: aOR: adjusted odds ratio; CI: confidence interval 29 30 *Model adjusted for patient’s age, gender, ethnicity and ward class 31 32 33 34 382 35 36 383 http://bmjopen.bmj.com/ 37 38 39 384 40 41 385 DISCUSSION 42 43 386 We are among the first to establish that both the caregiver and stroke survivor factors 44 on September 28, 2021 by guest. Protected copyright. 45 387 jointly determine the participation in SCR after stroke. We also demonstrated that the 46 388 SCR participation determinants vary across different periods post-stroke, with the 47 48 389 caregiver determinants being significant over the early post-stroke period (0-3 49 50 390 months) and the stroke survivor determinants being significant over both the early and 51 52 391 late post-stroke period (3-12 months). 53 392 Past literature has acquainted us fairly well with the role of caregiver factors, such as 54 55 393 caregiver availability,(17) support,(15, 18) and psychosocial health(14) in the 56 57 394 functional recovery or gains post-stroke. Clark and colleagues reported stronger 58 395 chance beliefs of the caregivers to be associated with a decreased likelihood of their 59 60 396 stroke survivor attending outpatient medicine and rehabilitation therapy

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 21 of 40 BMJ Open

19 1 2 3 397 appointments.(37) Researchers have reported caregiver factors to be associated with 4 BMJ Open: first published as 10.1136/bmjopen-2019-036631 on 23 April 2020. Downloaded from 5 398 delayed discharge from inpatient settings.(38, 39) Another study exploring the 6 7 399 caregiver determinants of post-stroke inpatient rehabilitation reported co-residing with 8 400 caregivers to be associated with decreased utilization of inpatient rehabilitation.(40) 9 10 401 We did not find any significant association between co-residing caregiver status and 11 12 402 SCR participation. While authors in these studies demonstrated the role of caregivers 13 14 403 in an inpatient rehabilitation setting, our study adds new knowledge on the role of 15 404 caregivers in SCR participation once they are discharged back home. 16 17 405 We are the first to study the role of caregiver factors in SCR participation over early 18 For peer review only 19 406 and late post-stroke periods and have demonstrated that the association of caregiver 20 21 407 factors with SCR participation varied over these two periods. Specifically, the 22 408 caregiver factors played a significant role during the early post-stroke period, with the 23 24 409 stroke survivor’s functional status being the only significant factor in the late post- 25 26 410 stroke period. A possible explanation could be related to the transition to the 27 411 community and adaptation related challenges in the early post-stroke period with 28 29 412 caregivers playing a crucial role in the rehabilitation journey during this phase. The 30 31 413 importance of this finding is further emphasized considering higher participation in 32 33 414 supervised rehabilitation in the early post-stroke period is reported to be associated 34 415 with better functional outcomes at 1 year post-stroke.(41) 35 36 416 We found that the odds of SCR participation increased with an increase in the positive http://bmjopen.bmj.com/ 37 38 417 care management strategy score. A possible explanation could be the caregivers 39 40 418 adopting positive care management strategies adapt better to their new role with lower 41 419 psychological issues like anxiety and are better able to care for the stroke survivors 42 43 420 including the facilitation of SCR participation. Along the same lines, a cross-sectional 44 on September 28, 2021 by guest. Protected copyright. 45 421 study on patients suffering cerebrovascular accidents reported poorer functional 46 422 outcomes in patients of caregivers suffering from anxiety.(14) We found that the odds 47 48 423 of SCR participation decreased with an increase in the caregiver reported stroke 49 50 424 survivor’s disruptive behaviour score. Researchers have reported that managing 51 52 425 problematic behaviours post-stroke can be difficult for the caregivers resulting in 53 426 caregiver strain which might make them unable to continue providing care(42) and, in 54 55 427 some circumstances, resulting in the institutionalization of stroke survivors.(43) Our 56 57 428 finding of stroke survivor’s disruptive behaviour being a predictor of decreased SCR 58 429 participation may be explained by caregivers being stressed and unable to comply 59 60 430 with their caregiving obligations, including facilitation of SCR participation.

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 22 of 40

20 1 2 3 431 Following are the practical implications of our work. Efforts should be directed 4 BMJ Open: first published as 10.1136/bmjopen-2019-036631 on 23 April 2020. Downloaded from 5 432 towards promoting positive care management strategies among caregivers by 6 7 433 optimizing their efficacy in caregiving related tasks so that they adapt well. Currently, 8 434 caregiver competency training is mainly focused on physical assistance. However, the 9 10 435 scope should also include mastering skills to manage behavioural issues post-stroke. 11 12 436 Management of such behavioural issues can be viewed under the three broad domains 13 14 437 of memory related, disruptive and depressive behavioural problems. Examples of the 15 438 memory related behavioural problems are asking the same questions over and over, 16 17 439 trouble remembering events, losing or misplacing things, difficulty concentrating on a 18 For peer review only 19 440 task and so forth. Alternatively, stroke survivor may present with disruptive 20 21 441 behavioural problems, such as talking loudly and rapidly, verbally aggressive to 22 442 others, arguing, irritability and so forth. The stroke survivor may also exhibit 23 24 443 depressive behavioural problems like being anxious or worried, crying, appear sad or 25 26 444 express feelings of hopelessness about future. Considering the variability in the stroke 27 445 survivors’ behavioural issues post-stroke, caregiver training should be supplemented 28 29 446 with assessment of such behavioural problems post-stroke, which can enable the 30 31 447 caregiving training to be tailored and specific towards caregivers’ management needs. 32 33 448 Our results support the adoption of a family-centred approach to post-stroke 34 449 rehabilitation providing due recognition to the family caregivers. A review on family- 35 36 450 centred approach towards post-stroke rehabilitation recommended keeping the http://bmjopen.bmj.com/ 37 38 451 caregivers informed, involving them in setting rehabilitation goals, teaching coping 39 40 452 skills and improving self-efficacy.(44) Another practical recommendation would be 41 453 moving from a reactive to a pro-active approach towards ensuring the stroke survivor- 42 43 454 caregiver dyads adjust well in the community. Caregiver readiness can be improved 44 on September 28, 2021 by guest. Protected copyright. 45 455 by conducting an assessment of the stroke survivor’s care needs and the caregiver’s 46 456 capacity and commitment to care at different time points across the post-stroke 47 48 457 recovery journey, considering stroke survivor-caregiver dyads’ needs are dynamic and 49 50 458 change with time. One such crucial point for conducting readiness assessment is 51 52 459 before being discharged home to ensure the dyad is prepared to transition into the 53 460 community setting, and the caregiver can successfully engage in caregiving 54 55 461 responsibilities.(6) 56 57 58 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 23 of 40 BMJ Open

21 1 2 3 462 Study limitations 4 BMJ Open: first published as 10.1136/bmjopen-2019-036631 on 23 April 2020. Downloaded from 5 463 Following are the limitations. There is a possibility of information bias related to the 6 7 464 limited recall of SCR participation. To address this, we kept our recall period to 3 8 9 465 months as past literature recommends shorter recall periods to ensure greater accuracy 10 466 of reporting utilization.(45) We limited the current sample to patients with stroke 11 12 467 surviving the first post-stroke year, excluding deaths within the follow-up period 13 14 468 (<5%). Since those that survived the first post-stroke year versus those who died 15 16 469 during this period might be systematically different, our findings would be 17 470 generalizable to those stroke survivors who are alive at the end of first year post- 18 For peer review only 19 471 stroke. With respect to caregiver related exclusions, we limited the current sample to 20 21 472 stroke survivors with available caregivers at baseline, excluding those without any 22 23 473 caregivers (11%). Considering the scope of the current study, which was examining 24 474 the determinants of SCR participation adopting a stroke survivor-caregiver dyadic 25 26 475 approach, along with the exclusion of stroke survivors without a caregiver, the 27 28 476 generalizability of our findings is limited to those stroke survivors who have a 29 477 caregiver post-stroke. To further comment on the representativeness of our sample, 30 31 478 we compared the demographic characteristics of current sample with the estimates 32 33 479 from the Singapore Stroke Registry for the year of 2013. With a mean age of 61.77 34 35 480 years, our cohort was on average younger than the national cohort by about 6 years. 36 481 Both the cohorts were similar with respect to having higher proportion of male stroke http://bmjopen.bmj.com/ 37 38 482 survivors and those of Chinese ethnicity. Refusal to participate by both the stroke 39 40 483 survivors and their caregivers could be one of the factors that can potentially 41 484 introduce selection bias. This is especially relevant if those who refused to participate 42 43 485 are systematically different from those who didn’t, in factors of direct relevance to 44 on September 28, 2021 by guest. Protected copyright. 45 486 this study. We were not able to capture reasons for refusal to participate. However, 46 47 487 the proportion of caregivers excluded due to refusal to participate either by the 48 488 caregivers (3%) or their stroke survivors (1.4%) was low, so the possibility of refusals 49 50 489 biasing our findings is unlikely. Another limitation is related to the temporality across 51 52 490 caregiver characteristics and SCR participation over the first three months after stroke 53 54 491 as both were determined simultaneously at the end of 3 months. However, we did 55 492 have the temporality across caregiver characteristics and SCR participation over 3-12 56 57 493 months after stroke. We did not include environment as one of the factors in the 58 59 494 current analysis as our scope was limited to stroke survivor-caregiver dyadic level 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 24 of 40

22 1 2 3 495 which did not include macro-level variables. While it is recommended to consider 4 BMJ Open: first published as 10.1136/bmjopen-2019-036631 on 23 April 2020. Downloaded from 5 496 environment or person-environment interaction in the context of participation post- 6 7 497 stroke(46, 47), the inclusion of environment as an independent variable is complicated 8 498 by the inherent challenges in the conceptualization, measurement and analysis of this 9 10 499 construct, which is reported to be generic and broad.(47, 48) Moreover, environment 11 12 500 as a factor is reported to be significant in studying the ADL participation level(49) as 13 14 501 compared to healthcare service utilization. Future studies can build on the current 15 502 findings to explore the influence of environment as a factor on the association of 16 17 503 stroke survivor-caregiver factors and the SCR participation. 18 For peer review only 19 504 20 21 505 22 506 Study strengths 23 24 507 Our study has some strengths. We are among the first, to the best of our knowledge, 25 26 508 to demonstrate the role caregivers play in stroke survivor’s SCR participation. Our 27 509 results have substantiated to some extent the rationale for the adoption of a stroke 28 29 510 survivor-caregiver dyadic approach(7) to studying post-stroke SCR utilization. We 30 31 511 reported the relative importance of caregiver factors in early as compared to late post- 32 33 512 stroke period. Another strength was the comprehensiveness of caregiver variables 34 513 considered which enabled us to explore the role of caregivers in-depth. Being a multi- 35 36 514 centre study (i.e., covering all the tertiary hospitals in Singapore at the time of the http://bmjopen.bmj.com/ 37 38 515 study), enhances the representativeness of the recruited sample. In addition, we did 39 40 516 not have any language barriers to recruitment, enrolling multi-ethnic participants into 41 517 the study, which further increases the generalizability of our findings. 42 43 44 518 CONCLUSION on September 28, 2021 by guest. Protected copyright. 45 46 519 With the aim to study the caregiver determinants of SCR participation after stroke, 47 48 520 our study demonstrated that the decision of community rehabilitation participation is 49 50 521 not singularly dependent on the stroke survivor’s clinical or functional characteristics 51 52 522 but rather influenced by both the caregiver and the stroke survivor’s characteristics. 53 523 We found that the caregiver’s positive care management strategies increased the odds 54 55 524 of SCR participation and the caregiver reported stroke survivor’s disruptive behaviour 56 57 525 decreased the odds of SCR participation over 3 months post-stroke. Moving forward, 58 59 526 our results make a case for the adoption of a stroke survivor-caregiver dyadic 60 527 approach for studying post-stroke utilization of community rehabilitation services,

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 25 of 40 BMJ Open

23 1 2 3 528 giving due consideration to both the stroke survivors and their caregivers. Adopting a 4 BMJ Open: first published as 10.1136/bmjopen-2019-036631 on 23 April 2020. Downloaded from 5 529 stroke survivor-caregiver dyadic approach in practice settings should include 6 7 530 promotion of positive care management strategies, comprehensive caregiving training 8 531 including both physical and behavioural dimensions, active engagement of caregivers 9 10 532 in rehabilitation journey and conducting regular caregiver needs assessments in the 11 12 533 community. 13 14 15 534 Acknowledgements 16 17 535 We would like to thank the medical staff at the public tertiary hospitals for 18 For peer review only 19 536 assisting with the recruitment of patients and their caregivers. We would 20 21 537 also like to thank all the participants in our study for their participation and 22 538 cooperation. 23 24 25 26 539 Declaration of interest statement 27 540 The authors report no conflicts of interest. 28 29 30 31 541 Funding 32 33 542 This work was supported by Health Services Research Competitive Research Grant 34 543 from the National Medical Research Council, Singapore. 35 36 http://bmjopen.bmj.com/ 37 38 544 Data availability 39 40 545 The dataset used and analysed during the current study is available from the 41 546 corresponding author on reasonable request. 42 43 44 on September 28, 2021 by guest. Protected copyright. 45 547 Authors’ contribution 46 47 548 ST was involved in conceptualization and design of the study, analysis and 48 549 interpretation of data, original draft preparation and incorporating revisions in 49 50 550 manuscript based on critical inputs from other co-authors. GCHK was involved in 51 52 551 conceptualization and design of the study, acquisition of data, drafting of the 53 552 manuscript and providing critical inputs to revision of manuscript along with 54 55 553 supervision of the study. NL was involved in conceptualization and design of the 56 57 554 study, acquisition of data, drafting of the manuscript and providing critical inputs to 58 59 555 revision of manuscript. KBT was involved in conceptualization and design of the 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 26 of 40

24 1 2 3 556 study, acquisition of data, drafting of the manuscript and providing critical inputs to 4 BMJ Open: first published as 10.1136/bmjopen-2019-036631 on 23 April 2020. Downloaded from 5 557 revision of manuscript. HH made substantial contributions to conception and design 6 7 558 of the study specifically with provision of expertise in medical domain and was 8 559 involved in revising the manuscript critically for intellectual content. DBM made 9 10 560 substantial contributions to conception and design of the study specifically, with 11 12 561 provision of expertise in medical domain and was involved in revising the manuscript 13 14 562 critically for intellectual content. JY made substantial contributions to conception and 15 563 design of the study specifically with provision of expertise in financial domain and 16 17 564 was involved in revising the manuscript critically for intellectual content. AC made 18 For peer review only 19 565 substantial contributions to conception and design of the study specifically with 20 21 566 provision of expertise in social domain and was involved in revising the manuscript 22 567 critically for intellectual content. KEL was involved in acquisition of data and in 23 24 568 revising the manuscript critically for intellectual content. NV was involved in 25 26 569 acquisition of data and in revising the manuscript critically for intellectual content. 27 570 EM was involved in acquisition of data and in revising the manuscript critically for 28 29 571 intellectual content. KMC was involved in acquisition of data and in revising the 30 31 572 manuscript critically for intellectual content. DADS was involved in acquisition of 32 33 573 data and in revising the manuscript critically for intellectual content. PY was involved 34 574 in acquisition of data and in revising the manuscript critically for intellectual content. 35 36 575 BYT was involved in acquisition of data and in revising the manuscript critically for http://bmjopen.bmj.com/ 37 38 576 intellectual content. EC was involved in acquisition of data and in revising the 39 40 577 manuscript critically for intellectual content. SHY was involved in acquisition of data 41 578 and in revising the manuscript critically for intellectual content. YSN was involved in 42 43 579 acquisition of data and in revising the manuscript critically for intellectual content. 44 on September 28, 2021 by guest. Protected copyright. 45 580 TMT was involved in acquisition of data and in revising the manuscript critically for 46 581 intellectual content. YHA was involved in acquisition of data and in revising the 47 48 582 manuscript critically for intellectual content. KHK was involved in acquisition of data 49 50 583 and in revising the manuscript critically for intellectual content. RS was involved in 51 52 584 acquisition of data and in revising the manuscript critically for intellectual content. 53 585 RAM was involved in acquisition of data and in revising the manuscript critically for 54 55 586 intellectual content. HMC was involved in acquisition of data and in revising the 56 57 587 manuscript critically for intellectual content. TTY was involved in acquisition of data 58 588 and in revising the manuscript critically for intellectual content. CN was involved in 59 60 589 acquisition of data and in revising the manuscript critically for intellectual content.

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 27 of 40 BMJ Open

25 1 2 3 590 AC was involved in acquisition of data and in revising the manuscript critically for 4 BMJ Open: first published as 10.1136/bmjopen-2019-036631 on 23 April 2020. Downloaded from 5 591 intellectual content. CST was involved in conceptualization and design of the study, 6 7 592 analysis and interpretation of data, drafting of the manuscript and providing critical 8 593 inputs to revision of manuscript. All the authors have read and approved the final 9 10 594 version of the manuscript to be published and are agreeable to take accountability of 11 12 595 all aspects of the work involved in the manuscript. 13 14 596 15 597 Ethics approval 16 17 598 The Singapore Stroke Study was approved by the National University of Singapore 18 For peer review only 19 599 Institutional Review Board, SingHealth Centralized Institutional Review Board and 20 21 600 the National Health Group Domain Specific Review Board. Written informed consent 22 601 was obtained from both the patients and the caregivers in their preferred language by 23 24 602 trained researchers. 25 26 603 27 28 29 604 REFERENCES 30 31 605 1. Strong K, Mathers C, Bonita R. Preventing stroke: saving lives around the 32 33 606 world. The Lancet Neurology. 2007;6(2):182-7. 34 607 2. Singapore Stroke Registry. Trends in Stroke in Singapore 2005-2012. 35 36 608 Singapore: National Registry of Diseases Office, Ministry of Health. http://bmjopen.bmj.com/ 37 38 609 3. Cameron JI, Tsoi C, Marsella A. Optimizing stroke systems of care by 39 40 610 enhancing transitions across care environments. Stroke. 2008;39(9):2637-43. 41 611 4. Lutz BJ, Young ME, Cox KJ et al. The crisis of stroke: experiences of patients 42 43 612 and their family caregivers. Topics in stroke rehabilitation. 2011;18(6):786-97. 44 on September 28, 2021 by guest. Protected copyright. 45 613 5. Thom T, Haase N, Rosamond W et al. Heart disease and stroke statistics-- 46 614 2006 update: a report from the American Heart Association Statistics Committee and 47 48 615 Stroke Statistics Subcommittee. Circulation. 2006;113(6):e85-151. 49 50 616 6. Lutz BJ, Young ME, Creasy KR et al. Improving stroke caregiver readiness 51 52 617 for transition from inpatient rehabilitation to home. The Gerontologist. 53 618 2016;57(5):880-9. 54 55 619 7. Tyagi S, Koh GC, Luo N et al. Dyadic approach to post-stroke 56 57 620 hospitalizations: role of caregiver and patient characteristics. BMC neurology. 58 59 621 2019;19(1):267. 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 28 of 40

26 1 2 3 622 8. Bakas T, Clark PC, Kelly-Hayes M et al. Evidence for stroke family caregiver 4 BMJ Open: first published as 10.1136/bmjopen-2019-036631 on 23 April 2020. Downloaded from 5 623 and dyad interventions: a statement for healthcare professionals from the American 6 7 624 Heart Association and American Stroke Association. Stroke. 2014;45(9):2836-52. 8 625 9. McCarthy MJ, Lyons KS, Powers LE. Expanding poststroke depression 9 10 626 research: movement toward a dyadic perspective. Topics in stroke rehabilitation. 11 12 627 2011;18(5):450-60. 13 14 628 10. Tyagi S, Tan CS, Koh GC-H. Dyadic approach to outpatient healthcare 15 629 utilization by stroke patients: can caregivers make a difference? Archives of Physical 16 17 630 Medicine and Rehabilitation. 2018;99(10):e45. 18 For peer review only 19 631 11. Lyons KS, Lee CS. The theory of dyadic illness management. Journal of 20 21 632 Family Nursing. 2018;24(1):8-28. 22 633 12. Lyons KS, Vellone E, Lee CS et al. A dyadic approach to managing heart 23 24 634 failure with confidence. Journal of Cardiovascular Nursing. 2015;30(4S):S64-S71. 25 26 635 13. Moon H, Adams KB. The effectiveness of dyadic interventions for people 27 636 with dementia and their caregivers. Dementia. 2013;12(6):821-39. 28 29 637 14. Em S, Bozkurt M, Caglayan M et al. Psychological health of caregivers and 30 31 638 association with functional status of stroke patients. Topics in stroke rehabilitation. 32 33 639 2017;24(5):323-9. 34 640 15. Harris JE, Eng JJ, Miller WC et al. The role of caregiver involvement in 35 36 641 upper-limb treatment in individuals with subacute stroke. Physical therapy. http://bmjopen.bmj.com/ 37 38 642 2010;90(9):1302-10. 39 40 643 16. Koh GC, Chen C, Cheong A et al. Trade-offs between effectiveness and 41 644 efficiency in stroke rehabilitation. International journal of stroke : official journal of 42 43 645 the International Stroke Society. 2012;7(8):606-14. 44 on September 28, 2021 by guest. Protected copyright. 45 646 17. Koh GCH, Wee LE, Chen C et al. Caregivers and their impact on inpatient 46 647 rehabilitation efficiency and effectiveness amongst recent stroke survivors in an 47 48 648 urbanised Asian society. Am Heart Assoc; 2012. 49 50 649 18. Tsouna-Hadjis E, Vemmos KN, Zakopoulos N et al. First-stroke recovery 51 52 650 process: the role of family social support. Arch Phys Med Rehabil. 2000;81(7):881-7. 53 651 19. Baker AC. The spouse's positive effect on the stroke patient's recovery. 54 55 652 Rehabilitation nursing : the official journal of the Association of Rehabilitation 56 57 653 Nurses. 1993;18(1):30-3. 58 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 29 of 40 BMJ Open

27 1 2 3 654 20. Fang Y, Tao Q, Zhou X et al. Patient and Family Member Factors Influencing 4 BMJ Open: first published as 10.1136/bmjopen-2019-036631 on 23 April 2020. Downloaded from 5 655 Outcomes of Poststroke Inpatient Rehabilitation. Arch Phys Med Rehabil. 6 7 656 2017;98(2):249-55 e2. 8 657 21. Chen C, Koh GC, Naidoo N et al. Trends in length of stay, functional 9 10 658 outcomes, and discharge destination stratified by disease type for inpatient 11 12 659 rehabilitation in Singapore community hospitals from 1996 to 2005. Arch Phys Med 13 14 660 Rehabil. 2013;94(7):1342-51 e4. 15 661 22. Intermediate and long-term care (ILTC) services. Ministry of Health (MOH), 16 17 662 Singapore. [cited 2018 November 5, 2018]. Available from: 18 For peer review only 19 663 https://http://www.moh.gov.sg/our-healthcare-system/healthcare-services-and- 20 21 664 facilities/intermediate-and-long-term-care-(iltc)-services 22 665 23. Tyagi S, Koh GC-H, Nan L et al. Healthcare utilization and cost trajectories 23 24 666 post-stroke: role of caregiver and stroke factors. BMC health services research. 25 26 667 2018;18(1):881. 27 668 24. Bakas T, Austin JK, Jessup SL et al. Time and difficulty of tasks provided by 28 29 669 family caregivers of stroke survivors. Journal of Neuroscience Nursing. 30 31 670 2004;36(2):95. 32 33 671 25. Bédard M, Molloy DW, Squire L et al. The Zarit Burden Interview: a new 34 672 short version and screening version. The gerontologist. 2001;41(5):652-7. 35 36 673 26. Seng BK, Luo N, Ng WY et al. Validity and reliability of the Zarit Burden http://bmjopen.bmj.com/ 37 38 674 Interview in assessing caregiving burden. Ann Acad Med Singapore. 2010;39:758-63. 39 40 675 27. Pearlin LI, Mullan JT, Semple SJ et al. Caregiving and the stress process: An 41 676 overview of concepts and their measures. The gerontologist. 1990;30(5):583-94. 42 43 677 28. Bakas T, Kroenke K, Plue LD et al. Outcomes among family caregivers of 44 on September 28, 2021 by guest. Protected copyright. 45 678 aphasic versus nonaphasic stroke survivors. Rehabilitation Nursing. 2006;31(1):33- 46 679 42. 47 48 680 29. Clark PC, Dunbar SB, Aycock DM et al. Caregiver perspectives of memory 49 50 681 and behavior changes in stroke survivors. Rehabilitation Nursing. 2006;31(1):26-32. 51 52 682 53 683 30. Gonzalez C, Bakas T. Factors associated with stroke survivor behaviors as 54 55 684 identified by family caregivers. Rehabilitation Nursing. 2013;38(4):202-11. 56 57 685 31. Haley WE, Allen JY, Grant JS et al. Problems and benefits reported by stroke 58 686 family caregivers: results from a prospective epidemiological study. Stroke. 59 60 687 2009;40(6):2129-33.

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 30 of 40

28 1 2 3 688 32. Teri L, Truax P, Logsdon R et al. Assessment of behavioral problems in 4 BMJ Open: first published as 10.1136/bmjopen-2019-036631 on 23 April 2020. Downloaded from 5 689 dementia: the revised memory and behavior problems checklist. Psychology and 6 7 690 aging. 1992;7(4):622. 8 691 33. Tan L, Yap P, Ng WY et al. Exploring the use of the Dementia Management 9 10 692 Strategies Scale in caregivers of persons with dementia in Singapore. Aging & mental 11 12 693 health. 2013;17(8):935-41. 13 14 694 34. Downey RG, King CV. Missing data in Likert ratings: A comparison of 15 695 replacement methods. The Journal of general psychology. 1998;125(2):175-91. 16 17 696 35. Bjerkreim AT, Thomassen L, Brøgger J et al. Causes and predictors for 18 For peer review only 19 697 hospital readmission after ischemic stroke. Journal of Stroke and Cerebrovascular 20 21 698 Diseases. 2015;24(9):2095-101. 22 699 36. StataCorp. 2015. Stata Statistical Software: Release 14. College Station, TX: 23 24 700 StataCorp LP. 25 26 701 37. Clark AN, Sander AM, Pappadis MR et al. Caregiver characteristics and their 27 702 relationship to health service utilization in minority patients with first episode stroke. 28 29 703 NeuroRehabilitation. 2010;27(1):95-104. 30 31 704 38. Lai W, Buttineau M, Harvey JK et al. Clinical and psychosocial predictors of 32 33 705 exceeding target length of stay during inpatient stroke rehabilitation. Topics in stroke 34 706 rehabilitation. 2017;24(7):510-6. 35 36 707 39. Tan WS, Chong WF, Chua KS et al. Factors associated with delayed http://bmjopen.bmj.com/ 37 38 708 discharges after inpatient stroke rehabilitation in Singapore. Ann Acad Med 39 40 709 Singapore. 2010;39(6):435-41. 41 710 40. Hinojosa MS, Rittman M, Hinojosa R. Informal caregivers and racial/ethnic 42 43 711 variation in health service use of stroke survivors. Journal of rehabilitation research 44 on September 28, 2021 by guest. Protected copyright. 45 712 and development. 2009;46(2):233-41. 46 713 41. Koh GC, Saxena SK, Ng TP et al. Effect of duration, participation rate, and 47 48 714 supervision during community rehabilitation on functional outcomes in the first 49 50 715 poststroke year in Singapore. Arch Phys Med Rehabil. 2012;93(2):279-86. 51 52 716 42. Cameron JI, Cheung AM, Streiner DL et al. Stroke survivors' behavioral and 53 717 psychologic symptoms are associated with informal caregivers' experiences of 54 55 718 depression. Arch Phys Med Rehabil. 2006;87(2):177-83. 56 57 719 43. Stephens S. Who's There?: When stroke or Alzheimer's changes a person's 58 720 behavior, caregiving can become extreme. Here, experienced caregivers, patients, and 59 60 721 experts share their stories and advice. Neurology Now. 2009;5(4):26-9.

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 31 of 40 BMJ Open

29 1 2 3 722 44. Visser-Meily A, Post M, Gorter JW et al. Rehabilitation of stroke patients 4 BMJ Open: first published as 10.1136/bmjopen-2019-036631 on 23 April 2020. Downloaded from 5 723 needs a family-centred approach. Disability and rehabilitation. 2006;28(24):1557-61. 6 7 724 45. Bhandari A, Wagner T. Self-reported utilization of health care services: 8 725 improving measurement and accuracy. Medical Care Research and Review. 9 10 726 2006;63(2):217-35. 11 12 727 46. Organization WH. International Classification of Functioning, Disability, and 13 14 728 Health: Children & Youth Version: ICF-CY: World Health Organization; 2007. 15 729 47. Whiteneck G, Dijkers MP. Difficult to measure constructs: conceptual and 16 17 730 methodological issues concerning participation and environmental factors. Archives of 18 For peer review only 19 731 physical medicine and rehabilitation. 2009;90(11):S22-S35. 20 21 732 48. Keysor J, Jette A, Haley S. Development of the home and community 22 733 environment (HACE) instrument. Journal of rehabilitation medicine. 2005;37(1):37- 23 24 734 44. 25 26 735 49. Keysor JJ, Jette AM, Coster W, Bettger JP, Haley SM. Association of 27 736 environmental factors with levels of home and community participation in an adult 28 29 737 rehabilitation cohort. Archives of physical medicine and rehabilitation. 30 31 738 2006;87(12):1566-75. 32 33 739 34 35 36 740 Figure 1. Study flowchart http://bmjopen.bmj.com/ 37 38 741 39 40 41 42 43 44 on September 28, 2021 by guest. Protected copyright. 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 32 of 40 BMJ Open: first published as 10.1136/bmjopen-2019-036631 on 23 April 2020. Downloaded from 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 For peer review only 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31

32 http://bmjopen.bmj.com/ 33 34 35 36 37 38 39

40 on September 28, 2021 by guest. Protected copyright. 41 42 43 44 45 Figure 1. Study Flowchart 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 33 of 40 BMJ Open

1 2 3 Supplement Table 1. Bivariate association of caregiver and patient characteristics with 4 BMJ Open: first published as 10.1136/bmjopen-2019-036631 on 23 April 2020. Downloaded from 5 supervised community rehabilitation (any use) across 3 months post-stroke. 6 Any Rehabilitation (3 months) 7 8 9 Variable Reference category OR (95% CI) P-value 10 11 CAREGIVER FACTORS 12 13 Age (in years) 0.990 (0.971, 1.009) 0.303 14 15 16 Gender Male 0.889 (0.492, 1.605) 0.696 17 18 Ethnicity For peerNon- Chinesereview only0.631 (0.380, 1.049) 0.076 19 20 Marital Status Single 0.810 (0.445, 1.474) 0.489 21 22 23 Caregiver identity Spouse 0.848 24 25 Adult-child 1.096 (0.627, 1.918) 26 27 Sibling 1.690 (0.458, 6.232) 28

29 Others 1.268 (0.464, 3.461) 30 31 32 Comorbid Conditions None 0.076 33 34 1 2.136 (1.131, 4.035) 35 36 2 0.697 (0.276, 1.761) http://bmjopen.bmj.com/ 37 38 39 3 and more 1.119 (0.465, 2.694) 40 41 Co-residing with patient No 1.187 (0.545, 2.584) 0.665 42 43 Caring for multiple care No 0.875 (0.530, 1.446) 0.602 44 on September 28, 2021 by guest. Protected copyright. 45 46 recipients 47 48 Caregiver relationship 1.131 (0.976, 1.309) 0.101 49 50 with patient 51 52 Revised memory and behaviour checklist 53 54 55 Memory problems 1.004 (0.963, 1.047) 0.856 56 57 58 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 34 of 40

1 2 3 Depressive behaviour 0.961 (0.911, 1.013) 0.136 4 BMJ Open: first published as 10.1136/bmjopen-2019-036631 on 23 April 2020. Downloaded from 5 6 problems 7 8 Disruptive behaviour problems 0.889 (0.819, 0.964) 0.005 9 10 Caregiver burden 11 12 13 Oberst Caregiving Burden 1.013 (0.993, 1.033) 0.202 14 15 Scale 16 17 Zarit Burden Interview 0.984 (0.953, 1.016) 0.318 18 For peer review only 19 Family conflict 20 21 22 Attitude towards patient 1.029 (0.973, 1.087) 0.202 23 24 Attitude towards caregiver 1.018 (0.962, 1.078) 0.529 25 26 Social support (instrumental) 27 28 29 FDW for general help No 0.933 (0.497, 1.751) 0.830 30 31 FDW for stroke patient No 1.143 (0.550, 2.379) 0.720 32 33 Social Support (perceived) 1.020 (0.970, 1.073) 0.436 34 35 36 Care management strategies http://bmjopen.bmj.com/ 37 38 Positive strategies 1.043 (1.018, 1.068) 0.001 39 40 Negative strategies 0.947 (0.896, 1.002) 0.057 41 42 PATIENT FACTORS 43 44 on September 28, 2021 by guest. Protected copyright. 45 Age (in years) 0.982 (0.959, 1.006) 0.148 46 47 Gender Male 1.169 (0.696, 1.964) 0.556 48 49 Ethnicity Non-Chinese 0.551 (0.330, 0.919) 0.022 50 51 52 Marital Status Single 0.933 (0.502, 1.733) 0.826 53 54 Ward Class Unsubsidized 0.769 (0.307, 1.928) 0.576 55 56 Charlson Comorbidity Index 1 - 3 0.673 57 58 59 4 - 6 1.335 (0.707, 2.523) 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 35 of 40 BMJ Open

1 2 3 >= 7 1.258 (0.557, 2.843) 4 BMJ Open: first published as 10.1136/bmjopen-2019-036631 on 23 April 2020. Downloaded from 5 6 Stroke type Non-ischemic 0.679 (0.307, 1.502) 0.340 7 8 Recurrent stroke No 0.628 (0.317, 1.244) 0.182 9 10 National Institute of Health Mild (0-4) 0.038 11 12 13 Scale 14 15 Moderately severe (5-14) 1.974 (1.165, 3.345) 16 17 Severe (15-24) 1.633 (0.523, 5.105) 18 For peer review only 19 Modified Rankin Scale No or slight disability 0.003 20 21 22 (0-2) 23 24 Moderate or severe disability 2.177 (1.300, 3.645) 25 26 (3-5) 27 28 29 Mini-Mental State No cognitive 0.480 30 31 Examination impairment 32 33 (24-30) 34 35 36 Mild cognitive impairment 1.263 (0.693, 2.300) http://bmjopen.bmj.com/ 37 38 (18-23) 39 40 Severe cognitive impairment 0.783 (0.399, 1.538) 41 42 (1-17) 43 44 on September 28, 2021 by guest. Protected copyright. 45 Discharge to Step-down No 1.172 (0.664, 2.069) 0.584 46 47 facility (Community 48 49 Hospital) 50 51 52 Centre for Epidemiological 1.014 (0.969, 1.060) 0.555 53 54 Studies Depression Scale 55 56 Abbreviations: OR: odds ratio; CI: confidence interval; FDW: foreign domestic worker 57 58 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 36 of 40

1 2 3 4 BMJ Open: first published as 10.1136/bmjopen-2019-036631 on 23 April 2020. Downloaded from 5 6 7 8 9 Supplement Table 2. Bivariate association of caregiver and patient characteristics with 10 supervised community rehabilitation (any use) across 3-12 months post-stroke. 11 12 Any Rehabilitation 13 14 15 (3 - 12 months) 16 17 Variable Reference category OR (95% CI) P-value 18 For peer review only 19 CAREGIVER FACTORS 20 21 Age (in years) 1.008 (0.985, 1.031) 0.500 22 23 24 Gender Male 1.828 (0.832, 4.016) 0.133 25 26 Ethnicity Non-Chinese 1.097 (0.600, 2.008) 0.763 27 28 Marital Status Single 1.230 (0.584, 2.589) 0.587 29 30 31 Caregiver identity Spouse 0.485 32 33 Adult-child 1.236 (0.628, 2.432) 34 35 Sibling 2.375 (0.632, 8.931) 36 http://bmjopen.bmj.com/ 37 38 Others 1.781 (0.568, 5.586) 39 40 Comorbid Conditions None 0.349 41 42 1 1.036 (0.505, 2.124) 43 44 2 0.432 (0.121, 1.543) on September 28, 2021 by guest. Protected copyright. 45 46 47 3 and more 0.432 (0.121, 1.543) 48 49 Co-residing with patient No 0.534 (0.222, 1.284) 0.161 50 51 Caring for multiple care No 0.828 (0.451, 1.521) 0.543 52 53 54 recipients 55 56 Caregiver relationship 1.002 (0.846, 1.187) 0.979 57 58 with patient 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 37 of 40 BMJ Open

1 2 3 Revised memory and behaviour checklist 4 BMJ Open: first published as 10.1136/bmjopen-2019-036631 on 23 April 2020. Downloaded from 5 6 Memory problems 1.004 (0.954, 1.056) 0.888 7 8 Depressive behaviour 1.027 (0.968, 1.089) 0.377 9 10 problems 11 12 13 Disruptive behaviour problems 0.941 (0.855, 1.035) 0.209 14 15 Caregiver burden 16 17 Oberst Caregiving Burden 1.026 (1.003, 1.049) 0.025 18 For peer review only 19 Scale 20 21 22 Zarit Burden Interview 1.000 (0.963, 1.038) 0.997 23 24 Family conflict 25 26 Attitude towards patient 1.030 (0.962, 1.102) 0.399 27 28 29 Attitude towards caregiver 10.33 (0.963, 1.108) 0.369 30 31 Social support (instrumental) 32 33 FDW for general help No 0.406 (0.198, 0.834) 0.014 34 35 36 FDW for stroke patient No 2.758 (1.175, 6.477) 0.020 http://bmjopen.bmj.com/ 37 38 Social Support (perceived) 0.971 (0.914, 1.033) 0.356 39 40 Care management strategies 41 42 Positive strategies 1.016 (0.988, 1.044) 0.278 43 44 on September 28, 2021 by guest. Protected copyright. 45 Negative strategies 1.009 (0.948, 1.074) 0.778 46 47 PATIENT FACTORS 48 49 Age (in years) 1.010 (0.981, 1.040) 0.491 50 51 52 Gender Male 1.126 (0.606, 2.090) 0.708 53 54 Ethnicity Non-Chinese 1.072 (0.586, 1.963) 0.821 55 56 Marital Status Single 0.725 (0.350, 1.502) 0.387 57 58 59 Ward Class Unsubsidized 0.439 (0.159, 1.212) 0.112 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 38 of 40

1 2 3 Charlson Comorbidity Index 1 - 3 0.186 4 BMJ Open: first published as 10.1136/bmjopen-2019-036631 on 23 April 2020. Downloaded from 5 6 4 - 6 0.687 (0.324, 1.457) 7 8 >= 7 1.356 (0.547, 3.364) 9 10 Stroke type Non-ischemic 0.471 (0.201, 1.106) 0.084 11 12 13 Recurrent stroke No 0.735 (0.318, 1.700) 0.472 14 15 National Institute of Health Mild (0-4) 0.003 16 17 Scale 18 For peer review only 19 Moderately severe (5-14) 2.235 (1.186, 4.212) 20 21 22 Severe (15-24) 5.681 (1.747, 18.473) 23 24 Modified Rankin Scale No or slight <0.001 25 26 disability (0-2) 27 28 29 Moderate or severe disability 4.132 (2.013, 8.479) 30 31 (3-5) 32 33 Mini-Mental State No cognitive 0.303 34 35 36 Examination impairment http://bmjopen.bmj.com/ 37 38 (24-30) 39 40 Mild cognitive impairment 1.279 (0.627, 2.609) 41 42 (18-23) 43 44 on September 28, 2021 by guest. Protected copyright. 45 Severe cognitive impairment 1.833 (0.842, 3.994) 46 47 (1-17) 48 49 Discharge to Community No 1.808 (0.939, 3.480) 0.076 50 51 52 Hospital 53 54 Centre for Epidemiological 1.009 (0.957, 1.065) 0.730 55 56 Studies Depression Scale 57 58

59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 39 of 40 BMJ Open

1 2 3 Abbreviations: OR: odds ratio; CI: confidence interval; FDW: foreign domestic worker 4 BMJ Open: first published as 10.1136/bmjopen-2019-036631 on 23 April 2020. Downloaded from 5 6 7 8 9 10 11 12 13 14 15 16 17 18 For peer review only 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 http://bmjopen.bmj.com/ 37 38 39 40 41 42 43 44 on September 28, 2021 by guest. Protected copyright. 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 40 of 40

1 STROBE Statement—Checklist of items that should be included in reports of cohort studies 2 3

4 Item Page No BMJ Open: first published as 10.1136/bmjopen-2019-036631 on 23 April 2020. Downloaded from 5 No Recommendation 6 Title and abstract 1 (a) Indicate the study’s design with a commonly used term in the title or Title page, Page-1 7 the abstract 8 (b) Provide in the abstract an informative and balanced summary of what Abstract, 9 Page-3 10 was done and what was found 11 Introduction 12 Introduction, 13 Background/rationale 2 Explain the scientific background and rationale for the investigation Page-5-6 14 being reported 15 Objectives 3 State specific objectives, including any prespecified hypotheses Introduction, 16 Page-6 17 Methods 18 For peer review only Methods, 19 Study design 4 Present key elements of study design early in the paper Page-6 20 Setting 5 Describe the setting, locations, and relevant dates, including periods of Methods, 21 Page-6 22 recruitment, exposure, follow-up, and data collection 23 Participants 6 (a) Give the eligibility criteria, and the sources and methods of selection Methods, 24 Page-6-8 of participants. Describe methods of follow-up 25 NA 26 (b) For matched studies, give matching criteria and number of exposed 27 and unexposed 28 Variables 7 Clearly define all outcomes, exposures, predictors, potential confounders, Methods, 29 Page 8-10 and effect modifiers. Give diagnostic criteria, if applicable 30 31 Data sources/ 8* For each variable of interest, give sources of data and details of methods Methods, Page 8-9 32 measurement of assessment (measurement). Describe comparability of assessment 33 methods if there is more than one group 34 Bias 9 Describe any efforts to address potential sources of bias Discussion, 35 Page-20-21

36 http://bmjopen.bmj.com/ Study size 10 Explain how the study size was arrived at Based on 37 availability 38 of data. 39 Quantitative variables 11 Explain how quantitative variables were handled in the analyses. If Methods 40 (analysis), applicable, describe which groupings were chosen and why 41 Page-10 42 Statistical methods 12 (a) Describe all statistical methods, including those used to control for Methods 43 (analysis), 44 confounding Page-10 on September 28, 2021 by guest. Protected copyright. 45 (b) Describe any methods used to examine subgroups and interactions NA 46 NA 47 (c) Explain how missing data were addressed 48 (d) If applicable, explain how loss to follow-up was addressed Follow-up 49 rates are 50 given under Results, 51 Page-11. 52 Not done. 53 (e) Describe any sensitivity analyses 54 Results 55 Participants 13* (a) Report numbers of individuals at each stage of study—eg numbers Results, 56 Page-11 & potentially eligible, examined for eligibility, confirmed eligible, included 57 figure 1 for 58 in the study, completing follow-up, and analysed study 59 flowchart. 60 (b) Give reasons for non-participation at each stage Figure 1 (c) Consider use of a flow diagram Figure 1

1 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 41 of 40 BMJ Open

1 Descriptive data 14* (a) Give characteristics of study participants (eg demographic, clinical, Results, 2 Page 11 & social) and information on exposures and potential confounders 3 Table 1, 2

4 (b) Indicate number of participants with missing data for each variable of Table 1, 2 BMJ Open: first published as 10.1136/bmjopen-2019-036631 on 23 April 2020. Downloaded from 5 interest 6 (c) Summarise follow-up time (eg, average and total amount) Follow-up 7 of 1 year 8 after index 9 stroke. 10 Outcome data 15* Report numbers of outcome events or summary measures over time Results, 11 Page-11 12 13 14 15 16 17 18 For peer review only 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35

36 http://bmjopen.bmj.com/ 37 38 39 40 41 42 43 44 on September 28, 2021 by guest. Protected copyright. 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

2 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 42 of 40

1 Main results 16 (a) Give unadjusted estimates and, if applicable, confounder-adjusted estimates and Results, 2 Table 3 & their precision (eg, 95% confidence interval). Make clear which confounders were 3 Table 4.

4 adjusted for and why they were included BMJ Open: first published as 10.1136/bmjopen-2019-036631 on 23 April 2020. Downloaded from 5 (b) Report category boundaries when continuous variables were categorized Table 1, 2. 6 (c) If relevant, consider translating estimates of relative risk into absolute risk for a NA 7 meaningful time period 8 9 Other analyses 17 Report other analyses done—eg analyses of subgroups and interactions, and NA 10 sensitivity analyses 11 12 Discussion 13 Key results 18 Summarise key results with reference to study objectives Discussion, 14 Page-18 15 Limitations 19 Discuss limitations of the study, taking into account sources of potential bias or Discussion, Page-20-22 16 imprecision. Discuss both direction and magnitude of any potential bias 17 Interpretation 20 Give a cautious overall interpretation of results considering objectives, limitations, Conclusion, 18 For peer review only Page-22-23 19 multiplicity of analyses, results from similar studies, and other relevant evidence 20 Generalisability 21 Discuss the generalisability (external validity) of the study results Discussion, 21 under study 22 limitations Page-20-21 23 24 Other information 25 Funding 22 Give the source of funding and the role of the funders for the present study and, if Funding 26 Page - 23 applicable, for the original study on which the present article is based 27 28 29 *Give information separately for exposed and unexposed groups. 30 31 Note: An Explanation and Elaboration article discusses each checklist item and gives methodological background and 32 33 published examples of transparent reporting. The STROBE checklist is best used in conjunction with this article (freely 34 available on the Web sites of PLoS Medicine at http://www.plosmedicine.org/, Annals of Internal Medicine at 35 http://www.annals.org/, and Epidemiology at http://www.epidem.com/). Information on the STROBE Initiative is

36 http://bmjopen.bmj.com/ available at http://www.strobe-statement.org. 37 38 39 40 41 42 43 44 on September 28, 2021 by guest. Protected copyright. 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

3 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open BMJ Open: first published as 10.1136/bmjopen-2019-036631 on 23 April 2020. Downloaded from

Dyadic approach to supervised community rehabilitation participation in an Asian setting post-stroke: exploring the role of caregiver and patient characteristics in a prospective cohort study

ForJournal: peerBMJ Open review only Manuscript ID bmjopen-2019-036631.R2

Article Type: Original research

Date Submitted by the 04-Apr-2020 Author:

32 http://bmjopen.bmj.com/ 33 34 35 36 37 38 39

1 2 3

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 3 of 40 BMJ Open

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 4 of 40

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 5 of 40 BMJ Open

3 1 2 3 49 4 BMJ Open: first published as 10.1136/bmjopen-2019-036631 on 23 April 2020. Downloaded from 5 6 50 ABSTRACT 7 8 9 51 Objective: To study the association of caregiver factors and stroke survivor 10 52 factors with supervised community rehabilitation (SCR) participation over the 11 12 53 first 3 months and subsequent 3-12 months post-stroke in an Asian setting. 13 14 15 54 Design: prospective cohort study 16 17 18 55 Setting: communityFor setting peer review only 19 20 56 Participants: We recruited stroke survivors and their caregivers into our yearlong 21 22 57 cohort. Caregiver and stroke survivor variables were collected over 3-monthly 23 58 intervals. We performed logistic regression with the outcome variable being SCR 24 25 59 participation post-stroke. 26 27 60 Outcome measures: SCR participation over the first 3 months and subsequent 3-12 28 29 61 months post-stroke 30 62 Results: 251 stroke survivor-caregiver dyads were available for the current analysis. 31 32 63 The mean age of caregivers was 50.1 years, with the majority being female, married, 33 34 64 and co-residing with the stroke survivor. There were 61%, 28%, 4% and 7% of 35 65 spousal, adult-child, sibling and other caregivers. The odds of SCR participation 36 http://bmjopen.bmj.com/ 37 66 decreased by about 15% for every unit increase in caregiver-reported stroke 38 39 67 survivor’s disruptive behaviour score (OR: 0.845; 95% CI: 0.769, 0.929). For every 1- 40 41 68 unit increase in the caregiver’s positive management strategy score, the odds of using 42 69 SCR service increased by about 4% (OR: 1.039; 95% CI: 1.011, 1.068). 43 44 on September 28, 2021 by guest. Protected copyright. 45 70 Conclusion: We established that SCR participation is jointly determined by 46 47 71 both caregiver and stroke survivor factors, with factors varying over the early 48 49 72 and late post-stroke period. Our results support the adoption of a dyadic or more 50 73 inclusive approach for studying the utilization of community rehabilitation 51 52 74 services, giving due consideration to both the stroke survivors and their 53 54 75 caregivers. Adopting a stroke survivor-caregiver dyadic approach in practice 55 76 settings should include promotion of positive care management strategies, 56 57 77 comprehensive caregiving training including both physical and behavioural 58 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 6 of 40

101 supervised community rehabilitation participation by stroke survivors and on September 28, 2021 by guest. Protected copyright. 45 46 102 their caregivers, which was addressed by keeping a relatively shorter recall 47 103 period. 48 49 104 50 51 105 52 53 106 List of abbreviations 54 107 CES-D: Centre for Epidemiological Studies Depression scale; DRCs: day 55 56 108 rehabilitation centres; FDW: foreign domestic worker; mRS: modified Rankin scale; 57 58 109 NIHSS: National Institute of Health Scale; rDMSS: Revised Dementia Management 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 7 of 40 BMJ Open

5 1 2 3 110 Strategies scale; SCR: supervised community rehabilitation; S3: Singapore Stroke 4 BMJ Open: first published as 10.1136/bmjopen-2019-036631 on 23 April 2020. Downloaded from 5 111 Study. 6 7 112 8 9 10 113 INTRODUCTION 11 12 114 Stroke is associated with a significant mortality burden globally.(1) However, recent 13 14 115 epidemiological trends with an increased incidence in the younger population and 15 116 decreasing mortality rates over the years highlight the importance of functional 16 17 117 recovery post-stroke.(2) While rehabilitation is essential for functional recovery and 18 For peer review only 19 118 re-integration back in the community, it is conditional on patients with stroke taking 20 21 119 the initiative to attend such rehabilitation services post-stroke. The transition from 22 120 inpatient settings into the community is challenging for the stroke survivor-caregiver 23 24 121 dyads.(3) They move from a well-supported setting with a multi-disciplinary team 25 26 122 providing care and facilitating rehabilitation to a setting where they are on their own 27 123 trying to maintain the care continuum and seeking community services including 28 29 124 supervised community rehabilitation (SCR). This challenging transition is further 30 31 125 described by the concept of the duality of stroke crisis with first crisis coinciding with 32 33 126 the stroke and the second occurring during the discharge from an inpatient setting to 34 127 home.(4) During this second crisis, stroke survivors often feel unprepared and 35 36 128 overwhelmed to navigate post-stroke recovery journey. Many stroke survivors rely on http://bmjopen.bmj.com/ 37 38 129 family caregivers’ assistance to continue their recovery journey. 39 40 130 41 131 The role of family caregivers becomes highly relevant post-stroke, considering more 42 43 132 than half of the stroke survivors are discharged home with differing degrees of 44 on September 28, 2021 by guest. Protected copyright. 45 133 residual physical impairments.(5) However, neither is this caregiving role explicitly 46 134 acknowledged nor is the caregiver’s capacity and commitment to providing care 47 48 135 assessed.(6) This could result in a mismatch between the expected caregiving 49 50 136 responsibilities and the caregiver’s ability to fulfil these, potentially leading to 51 52 137 adverse consequences for the stroke survivor-caregiver dyad. High reliance on 53 138 caregivers to assist the stroke survivors in the community implies that caregiver 54 55 139 factors like coping or perceived stress can in turn, influence the stroke survivors’ 56 57 140 outcomes like participation in SCR. This highlights the importance of adopting a 58 59 141 stroke survivor-caregiver dyadic approach to studying and implementing SCR, which 60 142 includes provision of physical rehabilitation services by licensed physiotherapists or

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 8 of 40

6 1 2 3 143 occupational therapists in the community settings, such as day rehabilitation centres 4 BMJ Open: first published as 10.1136/bmjopen-2019-036631 on 23 April 2020. Downloaded from 5 144 (DRCs) or patient’s home. Dyadic approach is described as a “holistic approach to 6 7 145 post-stroke care provision by healthcare practitioners, giving due importance to both 8 146 patients with stroke and their caregivers, integrating caregivers in the healthcare 9 10 147 system to extend the care continuum to include informal care in the community and 11 12 148 provision of timely support for caregivers.”(7) Prior accounts in both stroke(7-10) and 13 14 149 non-stroke(11-13) populations have included such dyadic approach in their narratives. 15 150 In addition, giving due consideration to both the stroke survivors and their caregivers 16 17 151 in psycho-educational, skill-building and support interventions is reported to improve 18 For peer review only 19 152 stroke survivors’ outcomes.(8) 20 21 153 Recognizing the relevance of caregivers in stroke survivor’s recovery process, 22 154 researchers have attempted to study the association of caregiver availability and some 23 24 155 socio-demographic characteristics with functional outcomes post-stroke across 25 26 156 inpatient rehabilitation services.(14-18) While a study in the US(19) reported positive 27 157 role of the spouse in the recovery of stroke survivors, another study in Canada(15) 28 29 158 reported caregiver support being associated with a higher functional gain as compared 30 31 159 to those without caregiver support. A recent study in China explored the role of 32 33 160 family member’s positive and negative attitudes in the functional and cognitive 34 161 recovery of stroke survivors, with higher positive attitudes being associated with 35 36 162 higher cognitive gains after rehabilitation.(20) Existing literature supports caregivers http://bmjopen.bmj.com/ 37 38 163 playing an important role in stroke survivors’ functional and cognitive outcomes post- 39 40 164 rehabilitation across different settings and contexts. However, none of the studies so 41 165 far have focused on the association of caregiver characteristics with SCR 42 43 166 participation. 44 on September 28, 2021 by guest. Protected copyright. 45 167 Addressing the above mentioned gaps, we aimed to study the association of caregiver 46 168 factors along with stroke survivor factors with SCR participation over the first 3 47 48 169 months and subsequent 3-12 months post-stroke in an Asian setting. 49 50 51 170 MATERIALS AND METHODS 52 53 171 Study setting 54 55 172 In Singapore, after stabilization in a tertiary hospital, stroke survivors are assessed for 56 57 173 rehabilitation eligibility, and based on this assessment, they may undergo intensive 58 59 174 rehabilitation in an inpatient setting.(21) Another option either in succession to above 60 175 or as an alternative is SCR, often delivered at the DRCs. DRCs are run either within

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 9 of 40 BMJ Open

7 1 2 3 176 the premises of a step-down facility or as stand-alone centres, mainly providing 4 BMJ Open: first published as 10.1136/bmjopen-2019-036631 on 23 April 2020. Downloaded from 5 177 physiotherapy and occupational therapy.(22) DRCs along with nursing homes (i.e., 6 7 178 long-term residential care settings within the community) and home-based 8 179 rehabilitation fall under the broad umbrella of SCR. 9 10 11 12 180 Participants 13 14 181 Our participants were part of the Singapore Stroke Study (S3), a prospective 15 182 observational study with recruitment of stroke survivors and caregivers over a period 16 17 183 extending from December 2010 to September 2013. Singaporeans or permanent 18 For peer review only 19 184 residents 40 years and above who suffered a stroke or experienced symptoms within 4 20 21 185 weeks of admission to any of the five tertiary hospitals in Singapore during the 22 186 recruitment period with confirmed stroke diagnosis were recruited along with their 23 24 187 caregivers. Caregivers could be an immediate or extended family member or friend 25 26 188 who provided care or assistance of any kind and took the responsibility for the stroke 27 189 survivor and were recognized by the stroke survivor, not fully paid for caregiving. 28 29 190 The on-site research nurses reviewed the list of stroke survivors on a daily basis to 30 31 191 screen for eligible participants and conduct recruitment. All participants were 32 33 192 explained the study purpose and procedures in their preferred language, and written 34 193 informed consent was taken and documented. Participants were informed that they 35 36 194 could withdraw from the study at any point during the follow-up period, if they http://bmjopen.bmj.com/ 37 38 195 wished. 39 40 196 41 197 Data was collected at 3-monthly intervals, via in-person interviews at baseline, 3- 42 43 198 month and 12-month time points, and via telephone interviews at 6-month and 9- 44 on September 28, 2021 by guest. Protected copyright. 45 199 month time points. Trained interviewers conducted interviews covering the health, 46 200 social and financial domains. Several measures were taken to ensure good compliance 47 48 201 and minimize attrition, such as sending reminders prior to scheduled interviews, 49 50 202 scheduling interviews over weekends or evenings during weekdays, multiple contact 51 52 203 attempts (upto 3) before categorizing as lost to follow-up. To ensure the 53 204 standardization and quality of data collection, main investigators trained the research 54 55 205 assistants. The training sessions were video-recorded, and these recordings were used 56 57 206 to train subsequent research assistants covering the content and method of data 58 59 207 collection along with consent taking procedures. We enrolled multi-ethnic 60 208 participants, and all participants were interviewed in their preferred language (e.g.,

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 10 of 40

8 1 2 3 209 English, Mandarin, Malay or Tamil). Before collecting data, we pilot tested our 4 BMJ Open: first published as 10.1136/bmjopen-2019-036631 on 23 April 2020. Downloaded from 5 210 survey on 40 participants from two of the five sites and finalized the survey forms 6 7 211 after inclusion of necessary amendments. Further details about the S3 are reported 8 212 elsewhere.(7, 23) The S3 was approved by the National University of Singapore 9 10 213 Institutional Review Board, SingHealth Centralized Institutional Review Board and 11 12 214 the National Health Group Domain Specific Review Board. 13 14 15 215 Independent variables (caregiver) 16 17 18 216 With a primary Forfocus on peercaregiver factors, review following caregiver only variables were 19 20 217 considered for current analysis: socio-demographic characteristics, marital status, 21 218 relationship with caregiver (caregiver being spouse, adult-child, sibling or others 22 23 219 including distant relatives and friends), number of chronic ailments, co-residing 24 25 220 status, caregiver burden, family conflict, social support, caregiver reported stroke 26 221 survivor behavioural issues and adopted caregiver management approaches. Under 27 28 222 the caregiver burden, we incorporated measures of both objective and subjective 29 30 223 burden measured by the Oberst Caregiving Burden Scale(24) and the Zarit’s Burden 31 32 224 Interview(25, 26) respectively. The family caregiving conflict scale recommended by 33 225 Pearlin and colleagues was used to capture family conflict.(27) Adapting Pearlin and 34 35 226 colleagues’ description of social support, we incorporated both “instrumental” and 36 http://bmjopen.bmj.com/ 37 227 “expressive” dimensions of social support, with former captured by presence of paid 38 39 228 help or foreign domestic worker (FDW) for general household tasks or specifically 40 229 for stroke survivor and latter captured by Pearlin’s 8-item perceived social support 41 42 230 instrument.(27) The caregiver reported occurrence of problematic behaviour by stroke 43 44 231 survivors was recorded using the Revised Memory and Behavioural Problem on September 28, 2021 by guest. Protected copyright. 45 232 Checklist, previously used in stroke survivors.(28-31) Caregivers were asked whether 46 47 233 any of the 21 problematic behaviours (e.g., “asking the same question over and over”, 48 49 234 “destroying property”, “crying and tearfulness” etc.) have occurred during the 50 51 235 previous week. Responses were recorded on a 5-point Likert scale: 0 = never, 1 = not 52 236 in the past week, 2 = 1 to 2 times per week, 3 = 3 to 6 times per week and 4 = daily or 53 54 237 more often.(32) Separate summated scores were calculated across the 3 domains of 55 56 238 disruptive, depressive and memory related behavioural problems with Cronbach’s 57 58 239 alpha for each being 0.73, 0.87 and 0.90 respectively. To capture the care 59 240 management approaches by stroke survivors’ caregivers, we used the revised 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 11 of 40 BMJ Open

9 1 2 3 241 dementia management strategies scale (rDMSS). Previously validated in 4 BMJ Open: first published as 10.1136/bmjopen-2019-036631 on 23 April 2020. Downloaded from 5 242 Singapore,(33) the scale has two sub-components of positive and negative types of 6 7 243 management strategies with good reported internal consistency (Cronbach’s alpha 8 244 0.89 and 0.87 respectively). 9 10 11 12 245 Independent variables (stroke survivor) 13 14 246 Following baseline stroke survivor variables were considered: socio-demographic 15 16 247 characteristics, marital status, ward class as a proxy of socio-economic status, 17 18 248 Charlson ComorbidityFor Index, peer type of reviewstroke (ischemic oronly non-ischemic), recurrent or 19 20 249 first stroke, stroke severity measured on National Institute of Health Scale (NIHSS), 21 250 functional status measured on modified Rankin scale (mRS), impairment in cognition 22 23 251 measured on the Mini-Mental State examination (MMSE), discharge status and 24 25 252 depression measured on the 11-item version of the Centre for Epidemiological Studies 26 253 Depression scale (CES-D). Ward class captured the category of the ward in which the 27 28 254 stroke survivor stayed during the index hospitalization. To make healthcare affordable 29 30 255 for all, the Singaporean government subsidizes inpatient stay in the tertiary care 31 32 256 setting in a tiered manner. Based on financial assessment, the patients can be eligible 33 257 to stay at A, B1, B2 or C ward types, being entitled to increasing level of subsidies. 34 35 258 With quality of care remaining constant, the ward types usually differ in the amenities 36 http://bmjopen.bmj.com/ 37 259 provided to the warded patients. For the current analysis, we categorized ward class 38 39 260 into subsidized and non-subsidized categories.(7) For scales with more than 10 40 261 missing cases (NIHSS, MMSE, Revised memory and behaviour checklist), we used 41 42 262 the person mean substitution approach to impute for missing values for cases with less 43 44 263 than half constituting items missing.(34) on September 28, 2021 by guest. Protected copyright. 45 46 47 264 Outcome variables 48 49 50 265 The outcome of interest was SCR participation, which comprised of participation at 51 266 any of the following: rehabilitation at home, DRC or nursing homes. This information 52 53 267 was captured in the survey at 3 monthly intervals by asking the caregiver, “Has the 54 55 268 stroke survivor at any time during the last 3 months received rehabilitation? Please 56 57 269 include any rehabilitation at home, DRC and nursing homes”. For the subsequent 3 to 58 270 12 months, we created a variable for capturing the SCR participation information 59 60 271 collected at 6, 9 and 12-month interviews. It was coded as yes if the caregiver

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 12 of 40

10 1 2 3 272 reported any participation at either of the time points and no if the caregiver reported 4 BMJ Open: first published as 10.1136/bmjopen-2019-036631 on 23 April 2020. Downloaded from 5 273 no participation across all time points. 6 7 8 274 Data analysis 9 10 11 275 Univariate analysis was conducted to describe the caregiver and stroke survivor 12 13 276 characteristics. We conducted bivariate analysis to examine the associations between 14 277 independent variables (caregiver and stroke survivor factors) and the SCR 15 16 278 participation. The independent variables with p values < 0.1 on bivariate analysis 17 18 279 were chosen as Forpotential peerpredictors for review the multivariable only regression model. With these 19 20 280 potential predictor variables, we built the most parsimonious model using a backward 21 281 variable selection approach. At each model building stage, the most insignificant 22 23 282 variable was removed until we were left with the variables having a p-value < 0.05, 24 25 283 except for age, gender, ethnicity and ward class of stroke survivors which were kept 26 284 in the model. Logistic regression was used at both bivariate and model building stages 27 28 285 and we reported the unadjusted and adjusted odds ratio (OR) estimates with 95% 29 30 286 confidence intervals (CI). We ran separate models for the SCR participation across 31 32 287 the first 3 months post-stroke and subsequent 3 to 12 months as researchers have 33 288 previously reported variations in the determinants of stroke survivors’ outcomes over 34 35 289 these periods.(7, 35) The significance level was set at 5%. With the most 36 http://bmjopen.bmj.com/ 37 290 parsimonious model, we performed diagnostics for the model fit using Hosmer and 38 39 291 Lemeshow’s goodness-of-fit test, checked for model misspecifications, multi- 40 292 collinearity and influential observations. All analysis was performed in Stata version 41 42 293 14.1.(36) 43 44 294 on September 28, 2021 by guest. Protected copyright. 45 295 Patient and Public Involvement 46 47 296 This research was done without patient involvement. Patients were not invited to 48 49 297 comment on the study design and were not consulted to develop patient relevant 50 51 298 outcomes or interpret the results. Patients were not invited to contribute to the writing 52 299 or editing of this document for readability or accuracy. 53 54 55 56 300 RESULTS 57 58 301 Out of the 661 caregivers assessed at baseline, 399 caregivers were recruited after 59 60 302 exclusion of 190 caregivers and 72 stroke survivors not having a caregiver. Two

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 13 of 40 BMJ Open

11 1 2 3 303 hundred fifty-one stroke survivor-caregiver dyads were available for the current 4 BMJ Open: first published as 10.1136/bmjopen-2019-036631 on 23 April 2020. Downloaded from 5 6 304 analysis after exclusion of stroke survivors with deaths within the follow up period 7 8 305 and limiting to complete cases. (Please refer figure 1 for study flowchart) The follow- 9 10 306 up rates for caregivers were 87.2% and 73.4% at 3 and 12 months. The prevalence of 11 12 307 SCR participation was 49% over 0-3 month and 25% over 3-12 month periods. The 13 14 15 308 mean age of caregivers was 50.1 years, with the majority being female, married and 16 17 309 co-residing with the stroke survivor. There were 61%, 28%, 4% and 7% of spousal, 18 For peer review only 19 310 adult-child, sibling and other caregivers. The mean scores for memory-related, 20 21 22 311 depressive and disruptive behaviour problems of stroke survivors were 5.01, 3.15 and 23 24 312 2.67 respectively. 34.27 (10.85) and 11.07 (4.60) were the mean (SD) scores for 25 26 313 caregiver reported positive and negative care management strategies respectively. The 27 28 29 314 stroke survivors had a mean age of 61.8 years, with the majority being male (65%), of 30 31 315 Chinese ethnicity (59%) and married (81%). About 89% had ischemic index stroke 32 33 316 and for 17%, the index stroke was a recurrent one. Out of all stroke survivors, 57%, 34 35 317 38% and 5% had a mild, moderately severe and severe type of index stroke as 36 http://bmjopen.bmj.com/ 37 38 318 measured on NIHSS. More than half (59%) had moderate to severe disability (3-5) on 39 40 319 mRS. (Please refer table 1 and table 2) The findings from the diagnostics for model 41 42 320 fit using Hosmer and Lemeshow’s goodness-of-fit test are as follows: for 0-3 months 43 44 on September 28, 2021 by guest. Protected copyright. 45 321 Model: p-value= 0.663; and for 3-12 months Model: p-value= 0.778. 46 47 322 48 49 323 50 324 [Figure 1 near here] 51 52 325 53 54 326 Table 1. Descriptive characteristics of participants (caregiver factors) 55 56 57 58 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 14 of 40

12 1 2 3 All stroke Participated in Participated 4 BMJ Open: first published as 10.1136/bmjopen-2019-036631 on 23 April 2020. Downloaded from 5 * * survivors , SCR , † 6 in SCR , 7 No. (%) 3 months 8 9 3-12 months 10 No. (%) 11 No. (%) 12 13 Variable Category 14 15 CAREGIVER FACTORS 16 17 Age of caregiver (in years) 50.13 (13.09) 49.03 (12.97) 51.16 (10.76) 18 For peer review only 19 mean (SD) 20 21 Gender of caregiver Male 61 (23) 29 (24) 9 (16) 22 23 Female 199 (77) 93 (76) 49 (84) 24 25 Ethnicity of caregiver Chinese 151 (58) 66 (54) 35 (60) 26 27 28 Non-Chinese 109 (42) 56 (46) 23 (40) 29 30 Marital Status of caregiver Married 205 (79) 93 (76) 47 (81) 31 32 Single 55 (21) 29 (24) 11 (19) 33 34 Caregiver identity Spouse 159 (61) 71 (58) 32 (55) 35 36

Adult-child 74 (28) 36 (30) 17 (29) http://bmjopen.bmj.com/ 37 38 Sibling 10 (4) 6 (5) 4 (7) 39 40 Others 17 (7) 9 (7) 5 (9) 41 42 Co-residing with patient Yes 231 (89) 109 (89) 49 (84) 43 44 No 29 (11) 13 (11) 9 (16) on September 28, 2021 by guest. Protected copyright. 45 46 Caring for multiple care Yes 108 (42) 49 (40) 22 (38) 47 48 49 recipients 50 51 No 152 (58) 73 (60) 36 (62) 52 53 Revised memory and behaviour checklist 54 55 Memory problems mean (SD) 5.01 (5.93) 4.95 (6.30) 5.09 (6.06) 56 57 Depressive behaviour mean (SD) 3.15 (4.82) 2.72 (4.21) 3.61 (5.21) 58 59 problems 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 15 of 40 BMJ Open

13 1 2 3 Disruptive behaviour mean (SD) 2.67 (3.62) 1.97 (2.86) 2.10 (2.83) 4 BMJ Open: first published as 10.1136/bmjopen-2019-036631 on 23 April 2020. Downloaded from 5 problems 6 7 Caregiver Burden 8 9 10 Oberst Caregiving Burden mean (SD) 31.71 (12.63) 32.94 (12.02) 35.18 (13.20) 11 12 Scale 13 14 Zarit Burden Interview mean (SD) 8.73 (7.86) 8.25 (6.83) 8.74 (8.13) 15 16 Family conflict 17 18 Attitude towardsFor patient peermean (SD) review11.42 (4.49) only11.58 (4.46) 12.00 (4.12) 19 20 Attitude towards caregiver mean (SD) 11.63 (4.37) 11.68 (4.40) 12.26 (3.91) 21 22 Social Support (instrumental) 23 24 FDW for general help Yes 212 (82) 98 (80) 42 (72) 25 26 No 48 (18) 24 (20) 16 (28) 27 28 29 FDW for stroke survivor Yes 33 (13) 17 (14) 11 (19) 30 31 No 227 (87) 105 (86) 47 (81) 32 33 Social Support (perceived) mean (SD) 26.33 (4.90) 26.49 (5.03) 25.95 (4.61) 34 35 Care Management Strategies 36 http://bmjopen.bmj.com/ 37 Positive strategies mean (SD) 34.27 (10.85) 36.52 (10.19) 35.88 (10.92) 38 39 Negative strategies mean (SD) 11.07 (4.60) 10.56 (4.34) 11.36 (4.46) 40 41 42 43 *N = 251, †N=238 44 on September 28, 2021 by guest. Protected copyright. 45 46 Abbreviations: SCR: supervised community rehabilitation; No.: number; SD: standard deviation; FDW: foreign 47 48 domestic worker 49 50 327 51 52 53 328 Table 2. Descriptive characteristics of participants (stroke survivor factors) 54 55 All stroke Participated in Participated 56 * * 57 survivors , SCR , in SCR †, 58 59 No. (%) 3 months 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 16 of 40

14 1 2 3 No. (%) 3-12 months 4 BMJ Open: first published as 10.1136/bmjopen-2019-036631 on 23 April 2020. Downloaded from 5 No. (%) 6 7 8 Variable Category 9 10 PATIENT FACTORS 11 12 Age of patient mean (SD) 61.77 (10.42) 60.86 (10.63) 62.29 (10.53) 13 14 (in years) 15 16 Gender of patient Male 169 (65) 77 (63) 37 (64) 17 18 For peer review only 19 Female 91 (35) 45 (37) 21 (36) 20 21 Ethnicity of patient Chinese 153 (59) 65 (53) 35 (60) 22 23 Non-Chinese 107 (41) 57 (47) 23 (40) 24 25 Marital Status of patient Married 210 (81) 97 (80) 45 (78) 26 27 Single 50 (19) 25 (20) 13 (22) 28 29 Ward Class Unsubsidized 21 (8) 11 (9) 7 (12) 30 31 Subsidized 235 (92) 110 (91) 51 (88) 32 33 Charlson Comorbidity 1 - 3 52 (20) 22 (18) 13 (22) 34 35 Index 36 http://bmjopen.bmj.com/ 37 38 4 - 6 165 (63) 79 (65) 31 (53) 39 40 >= 7 43 (17) 21 (17) 14 (24) 41 42 Stroke type Ischemic 231 (89) 106 (87) 48 (83) 43 44 Non-ischemic 29 (11) 16 (13) 10 (17) on September 28, 2021 by guest. Protected copyright. 45 46 Recurrent stroke Yes 43 (17) 16 (13) 8 (14) 47 48 No 217 (83) 106 (87) 50 (86) 49 50 National Institute of Health Mild (0-4) 149 (57) 60 (49) 23 (40) 51 52 Scale 53 54 Moderately 97 (38) 55 (45) 28 (48) 55 56 severe (5-14) 57 58 59 Severe (15-24) 14 (5) 7 (6) 7 (12) 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 17 of 40 BMJ Open

15 1 2 3 Modified Rankin Scale No or slight 106 (41) 38 (31) 11 (19) 4 BMJ Open: first published as 10.1136/bmjopen-2019-036631 on 23 April 2020. Downloaded from 5 disability (0-2) 6 7 Moderate or 154 (59) 84 (69) 47 (81) 8 9 10 severe disability 11 12 (3-5) 13 14 Mini-Mental State No cognitive 150 (58) 70 (57) 30 (52) 15 16 Examination impairment (24- 17 18 For peer30) review only 19 20 Mild cognitive 65 (25) 33 (27) 15 (26) 21 22 impairment (18- 23 24 23) 25 26 Severe cognitive 45 (17) 19 (16) 13 (22) 27 28 29 impairment (1- 30 31 17) 32 33 Discharge to Step-down Yes 66 (25) 33 (27) 19 (33) 34 35 facility (Community 36 http://bmjopen.bmj.com/ 37 Hospital) 38 39 No 194 (75) 89 (73) 39 (67) 40 41 Centre for Epidemiological mean (SD) 6.31 (5.61) 6.59 (5.51) 6.60 (5.64) 42 43 Studies Depression Scale 44 on September 28, 2021 by guest. Protected copyright. 45 46 47 * † 48 N = 251, N=238 49 50 Abbreviations: SCR: supervised community rehabilitation; No.: number; SD: standard deviation 51 52 53 329 54 55 330 56 57 58 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 18 of 40

16 1 2 3 331 Supervised community rehabilitation participation (0-3 months post- 4 BMJ Open: first published as 10.1136/bmjopen-2019-036631 on 23 April 2020. Downloaded from 5 332 stroke) 6 7 8 333 Supplement table 1 depicts the results of the association of caregiver and stroke 9 334 survivor characteristics with odds of SCR participation across 3 months post-stroke. 10 11 335 The bivariate associations of caregiver reported disruptive behaviour of stroke 12 13 336 survivor and positive management strategy with SCR participation were statistically 14 15 337 significant. Among the stroke survivor factors, the bivariate associations of stroke 16 338 severity and functional status with SCR participation were statistically significant. 17 18 339 The variables thatFor entered peer the final adjusted review model of oddsonly of SCR participation over 19 20 340 first 3 months post-stroke were caregiver reported disruptive behaviour of the stroke 21 22 341 survivor, positive management strategy of the caregiver and stroke survivor’s 23 342 functional status (Please refer table 3). For every 1-unit increase in the caregiver 24 25 343 reported stroke survivor’s disruptive behaviour score, the odds of SCR participation 26 27 344 decreased by about 15% (OR: 0.845; 95% CI: 0.769, 0.929). For every 1-unit increase 28 345 in the caregiver reported positive care management strategy score, the odds of SCR 29 30 346 participation increased by about 4% (OR: 1.039; 95% CI: 1.011, 1.068). Compared to 31 32 347 stroke survivors with no or mild functional disability, those with moderate to severe 33 34 348 functional disability had 2.76 times the odds of SCR participation over the first 3 35 349 months post-stroke (p=0.001). 36 http://bmjopen.bmj.com/ 37 350 38 39 351 Table 3. Multi-variable regression analysis results of supervised community 40 41 352 rehabilitation (any use) across 3 months post-stroke. 42 43 Any Rehabilitation (3 months) 44 on September 28, 2021 by guest. Protected copyright. 45 Variable Reference category aOR (95% CI) P-value 46 47 Revised memory and behaviour checklist 48 49 Disruptive behaviour problems 0.845 (0.769, 0.929) <0.001 50 51 Care management strategies 52 53 Positive strategies 1.039 (1.011, 1.068) 0.006 54 55 Modified Rankin Scale No or slight 0.001 56 57 58 disability (0-2) 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 19 of 40 BMJ Open

17 1 2 3 Moderate or severe disability 2.759 (1.532, 4.969) 4 BMJ Open: first published as 10.1136/bmjopen-2019-036631 on 23 April 2020. Downloaded from 5 (3-5) 6 7 Age (in years)* 0.975 (0.949, 1.002) 0.074 8 9 10 Gender* Male 0.875 (0.480, 1.594) 0.662 11 12 Ethnicity* Non-Chinese 0.723 (0.409, 1.277) 0.264 13 14 Ward Class* Unsubsidized 0.806 (0.298, 2.180) 0.671 15 16 Abbreviations: aOR: adjusted odds ratio; CI: confidence interval 17 18 *Model adjusted Forfor patient’s peer age, gender, ethnicityreview and ward class only 19 20 353 21 22 23 24 354 Supervised community rehabilitation participation (3-12 months post- 25 355 stroke) 26 27 28 356 Supplement table 2 depicts the results of the association of caregiver and stroke 29 30 357 survivor characteristics with odds of SCR participation across 3-12 months post- 31 32 358 stroke. The bivariate associations of caregiver burden (measured on Oberst Burden 33 359 Scale), FDW for general help and FDW for stroke survivors with SCR participation 34 35 360 were statistically significant. Among the stroke survivor factors, the bivariate 36 http://bmjopen.bmj.com/ 37 361 associations of stroke severity and functional status with SCR participation were 38 362 statistically significant. The only variable that entered the final adjusted model of 39 40 363 odds of SCR participation 3-12 months post-stroke was the functional status, with 41 42 364 odds of SCR participation being 4.19 times in those with moderate to severe 43 44 365 functional disability when compared to those with none to mild disability (OR: 4.234; on September 28, 2021 by guest. Protected copyright. 45 366 95% CI: 2.034, 8.812) (Please refer table 4). 46 47 367 48 49 368 Table 4. Multi-variable regression analysis results of supervised community 50 51 369 rehabilitation (any use) across 3 – 12 months post-stroke. 52 53 Any Rehabilitation (3-12 months) 54 55 Variable Reference category aOR* (95% CI) P-value 56 57 Modified Rankin No or slight disability (0-2) <0.001 58 59 Scale 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 20 of 40

18 1 2 3 Moderate or severe 4.234 (2.034, 8.812) 4 BMJ Open: first published as 10.1136/bmjopen-2019-036631 on 23 April 2020. Downloaded from 5 disability (3-5) 6 7 Age (in years)* 0.999 (0.968, 1.032) 0.973 8 9 10 Gender* 0.875 (0.448, 1.709) 0.695 11 12 Ethnicity* 1.146 (0.603, 2.176) 0.678 13 14 Ward Class* 0.432 (0.146, 1.279) 0.130 15 16 Abbreviations: aOR: adjusted odds ratio; CI: confidence interval 17 18 *Model adjusted Forfor patient’s peer age, gender, ethnicityreview and ward class only 19 20 21 22 371 23 24 372 25 26 27 373 28 29 374 DISCUSSION 30 31 375 We are among the first to establish that both the caregiver and stroke survivor factors 32 33 376 jointly determine the participation in SCR post-stroke. We also demonstrated the 34 377 variability in the determinants of SCR participation, with the caregiver determinants 35 36 378 being significant over the early post-stroke period (0-3 months) and the stroke http://bmjopen.bmj.com/ 37 38 379 survivor determinants being significant over both the early and late post-stroke period 39 40 380 (3-12 months). 41 381 Past literature has acquainted us fairly well with the role of caregiver factors, such as 42 43 382 caregiver availability,(17) support,(15, 18) and psychosocial health(14) in the 44 on September 28, 2021 by guest. Protected copyright. 45 383 functional recovery post-stroke. Clark and colleagues reported stronger chance beliefs 46 47 384 of the caregivers to be associated with a decreased likelihood of their stroke survivors 48 385 attending outpatient medicine and rehabilitation therapy appointments.(37) 49 50 386 Researchers have reported caregiver factors to be associated with delayed discharge 51 52 387 from inpatient settings.(38, 39) Another study exploring the caregiver determinants of 53 388 post-stroke inpatient rehabilitation reported co-residing caregivers to be associated 54 55 389 with decreased utilization of inpatient rehabilitation.(40) We did not find any 56 57 390 significant association between co-residing status and SCR participation. While 58 59 391 authors in these studies demonstrated the role of caregivers in inpatient rehabilitation, 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 21 of 40 BMJ Open

19 1 2 3 392 our study adds new knowledge on the role of caregivers in SCR participation once 4 BMJ Open: first published as 10.1136/bmjopen-2019-036631 on 23 April 2020. Downloaded from 5 393 they are discharged home. 6 7 394 We reported that caregiver factors played a significant role during the early post- 8 395 stroke period, with the stroke survivors’ functional status being the only significant 9 10 396 factor in the late post-stroke period. A possible explanation could be related to the 11 12 397 transition to the community and related challenges in the early post-stroke period with 13 14 398 caregivers playing a crucial role in the rehabilitation journey during this phase. 15 399 Another possibility could be related to the improvement in functional status of stroke 16 17 400 survivors over time, making them less reliant on caregivers during the late post-stroke 18 For peer review only 19 401 period. The importance of this finding is further emphasized considering higher 20 21 402 participation in supervised rehabilitation in the early post-stroke period is reported to 22 403 be associated with better functional outcomes at 1 year post-stroke.(41) 23 24 404 We found that the odds of SCR participation increased with an increase in the positive 25 26 405 care management strategy score. A possible explanation could be the caregivers 27 406 adopting positive care management strategies adapt better to their new role with lower 28 29 407 psychological issues like anxiety and are better able to care for the stroke survivors 30 31 408 including the facilitation of SCR participation. Along the same lines, a cross-sectional 32 33 409 study on patients suffering cerebrovascular accidents reported poorer functional 34 410 outcomes in patients of caregivers having anxiety.(14) We found that the odds of SCR 35 36 411 participation decreased with an increase in the caregiver reported stroke survivor’s http://bmjopen.bmj.com/ 37 38 412 disruptive behaviour score. Managing problematic behaviours post-stroke can be 39 40 413 difficult for the caregivers resulting in caregiver strain and termination of care 41 414 provision(42) and, in some circumstances, institutionalization of stroke survivors.(43) 42 43 415 Similarly, in our setting, caregivers may be strained managing stroke survivor’s 44 on September 28, 2021 by guest. Protected copyright. 45 416 disruptive behaviour, limiting their ability to comply with their caregiving obligations, 46 417 including facilitation of SCR participation. 47 48 418 Following are the practical implications of our work. Efforts should be directed 49 50 419 towards promoting positive care management strategies among caregivers by 51 52 420 optimizing their efficacy in caregiving tasks so that they adapt well. Currently, 53 421 caregiver competency training is mainly focused on physical assistance. However, the 54 55 422 scope should also include mastering skills to manage behavioural issues post-stroke. 56 57 423 Management of such behavioural issues can be addressed under the three domains of 58 424 memory-related, disruptive and depressive behavioural problems. Examples of the 59 60 425 memory-related behavioural problems are asking the same questions over and over,

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 22 of 40

20 1 2 3 426 trouble remembering events, difficulty concentrating on a task and so forth. 4 BMJ Open: first published as 10.1136/bmjopen-2019-036631 on 23 April 2020. Downloaded from 5 427 Alternatively, stroke survivor may present with disruptive behavioural problems, such 6 7 428 as talking loudly and rapidly, verbally aggressive to others, arguing and so forth. The 8 429 stroke survivor may also exhibit depressive behavioural problems like being anxious 9 10 430 or worried, crying or appearing sad. Considering the variability in the stroke 11 12 431 survivors’ behavioural issues post-stroke, caregiver training should be supplemented 13 14 432 with assessment of such behavioural problems post-stroke, which can enable the 15 433 caregiving training to be tailored and specific towards caregivers’ management needs. 16 17 434 Our results support the adoption of a family-centred approach to post-stroke 18 For peer review only 19 435 rehabilitation providing due recognition to the family caregivers. A review on family- 20 21 436 centred approach towards post-stroke rehabilitation recommended keeping the 22 437 caregivers informed, involving them in setting rehabilitation goals, teaching coping 23 24 438 skills and improving self-efficacy.(44) Another practical recommendation would be to 25 26 439 pro-actively conduct caregiver readiness assessment to ensure the stroke survivor- 27 440 caregiver dyads adjust well in the community. One crucial point for conducting 28 29 441 readiness assessment is before discharge to home from inpatient rehabilitation 30 31 442 setting.(6) Building on work done by researchers describing stroke survivors’ needs 32 33 443 related to re-integration into community post-discharge,(45) and caregivers’ needs 34 444 related to caregiving and facilitating community transition,(46) readiness assessment 35 36 445 at discharge should focus on stroke survivors’ functional needs, community re- http://bmjopen.bmj.com/ 37 38 446 integration challenges, caregivers’ commitment and capacity to care (i.e., assessing 39 40 447 for pre-existing health issues and self-care strategies), prior caregiving experience, 41 448 available resources and overall impact of stroke.(46) 42 43 44 449 Study limitations on September 28, 2021 by guest. Protected copyright. 45 46 450 Following are the limitations. There is a possibility of information bias related to the 47 48 451 limited recall of SCR participation. To address this, we kept our recall period to 3 49 50 452 months as past literature recommends shorter recall periods to ensure greater accuracy 51 52 453 of reporting utilization.(47) We limited the current sample to patients with stroke 53 454 surviving the first post-stroke year, excluding deaths within the follow-up period 54 55 455 (<5%). Considering the possibility of systematic differences in survivors and those 56 57 456 who died during the follow-up, our findings would be generalizable to stroke 58 59 457 survivors alive at the end of first year post-stroke. With respect to caregiver related 60 458 exclusions, we limited the current sample to stroke survivors with available caregivers

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 23 of 40 BMJ Open

21 1 2 3 459 at baseline, excluding those without any caregivers (11%). Considering the scope of 4 BMJ Open: first published as 10.1136/bmjopen-2019-036631 on 23 April 2020. Downloaded from 5 460 the current study, which was examining the determinants of SCR participation 6 7 461 adopting a stroke survivor-caregiver dyadic approach, along with the exclusion of 8 462 stroke survivors without a caregiver, the generalizability of our findings is limited to 9 10 463 those stroke survivors who have a caregiver post-stroke. To further comment on the 11 12 464 representativeness of our sample, we compared the demographic characteristics of 13 14 465 current sample with the estimates from the Singapore Stroke Registry for the year of 15 466 2013. With a mean age of 61.77 years, our cohort was on average younger than the 16 17 467 national cohort by about 6 years. Both the cohorts were similar with respect to having 18 For peer review only 19 468 higher proportion of male stroke survivors and those of Chinese ethnicity. Refusal to 20 21 469 participate by both the stroke survivors and their caregivers could be one of the 22 470 factors that can potentially introduce selection bias. This is especially relevant if those 23 24 471 who refused to participate are systematically different from those who didn’t, in 25 26 472 factors of direct relevance to this study. We were not able to capture reasons for 27 473 refusal to participate. However, the proportion of caregivers excluded due to refusal to 28 29 474 participate either by the caregivers (3%) or their stroke survivors (1.4%) was low, so 30 31 475 the possibility of refusals biasing our findings is unlikely. Also, as with any 32 33 476 longitudinal study, we encountered a relatively lower response rate over the late post- 34 477 stroke period. Another limitation is related to the temporality across caregiver 35 36 478 characteristics and SCR participation over the first three months post-stroke as both http://bmjopen.bmj.com/ 37 38 479 were determined simultaneously. However, we did have the temporality across 39 40 480 caregiver characteristics and SCR participation over 3-12 months post-stroke. We did 41 481 not include environment as one of the factors in the current analysis as our scope was 42 43 482 limited to stroke survivor-caregiver dyadic level, excluding macro-level variables. 44 on September 28, 2021 by guest. Protected copyright. 45 483 While it is recommended to consider environment or person-environment interaction 46 484 in the context of participation post-stroke(48, 49), the inclusion of environment as an 47 48 485 independent variable is complicated by the inherent challenges in the 49 50 486 conceptualization, measurement and analysis of this construct, which is reported to be 51 52 487 generic and broad.(49, 50) Moreover, environment as a factor is reported to be 53 488 significant in studying the ADL participation level(51) as compared to healthcare 54 55 489 service utilization. Future studies can build on the current findings to explore the 56 57 490 influence of environment on the association of stroke survivor-caregiver factors and 58 491 the SCR participation. 59 60 492

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 24 of 40

22 1 2 3 493 4 BMJ Open: first published as 10.1136/bmjopen-2019-036631 on 23 April 2020. Downloaded from 5 494 Study strengths 6 7 495 Our study has some strengths. We are among the first, to the best of our knowledge, 8 496 to demonstrate the role caregivers play in stroke survivor’s SCR participation. Our 9 10 497 results have substantiated to some extent the rationale for the adoption of a stroke 11 12 498 survivor-caregiver dyadic approach(7) to studying post-stroke SCR utilization. We 13 14 499 reported the relative importance of caregiver factors in early as compared to late post- 15 500 stroke period. Another strength was the comprehensiveness of caregiver variables 16 17 501 considered which enabled us to explore the role of caregivers in-depth. Being a multi- 18 For peer review only 19 502 centre study enhances the representativeness of the recruited sample. In addition, we 20 21 503 did not have any language barriers to recruitment, enrolling multi-ethnic participants 22 504 into the study, which further increases the generalizability of our findings. 23 24 25 505 CONCLUSION 26 27 506 With the aim to study the caregiver determinants of SCR participation after stroke, 28 29 507 our study demonstrated that the SCR participation is determined by both the caregiver 30 31 508 and the stroke survivor characteristics. We found that the caregiver’s positive care 32 33 509 management strategies increased the odds of SCR participation and the caregiver 34 510 reported stroke survivor’s disruptive behaviour decreased the odds of SCR 35 36 511 participation over 3 months post-stroke. Our results support the adoption of a stroke http://bmjopen.bmj.com/ 37 38 512 survivor-caregiver dyadic approach for studying post-stroke utilization of community 39 40 513 rehabilitation services. Within practice settings, dyadic approach should include 41 514 promotion of positive care management strategies, comprehensive caregiving training 42 43 515 including both physical and behavioural dimensions, active engagement of caregivers 44 on September 28, 2021 by guest. Protected copyright. 45 516 in rehabilitation and conducting regular caregiver needs assessments in the 46 517 community. 47 48 49 50 518 Acknowledgements 51 52 519 We would like to thank the medical staff at the public tertiary hospitals for 53 520 assisting with the recruitment of patients and their caregivers. We would 54 55 521 also like to thank all the participants in our study for their participation and 56 57 522 cooperation. 58 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 25 of 40 BMJ Open

23 1 2 3 523 Declaration of interest statement 4 BMJ Open: first published as 10.1136/bmjopen-2019-036631 on 23 April 2020. Downloaded from 5 524 The authors report no conflicts of interest. 6 7 8 525 Funding 9 10 526 This work was supported by MOH Health Services Research Competitive Research 11 12 527 Grant (Project Number: HSRG/0006/2010) from the National Medical Research 13 14 528 Council, Singapore. 15 16 17 529 Data availability 18 For peer review only 19 530 The dataset used and analysed during the current study is available from the 20 21 531 corresponding author on reasonable request. 22 23 24 532 Authors’ contribution 25 26 533 ST was involved in conceptualization and design of the study, analysis and 27 28 534 interpretation of data, original draft preparation and incorporating revisions in 29 535 manuscript based on critical inputs from other co-authors. GCHK was involved in 30 31 536 conceptualization and design of the study, acquisition of data, drafting of the 32 33 537 manuscript and providing critical inputs to revision of manuscript along with 34 35 538 supervision of the study. NL was involved in conceptualization and design of the 36 539 study, acquisition of data, drafting of the manuscript and providing critical inputs to http://bmjopen.bmj.com/ 37 38 540 revision of manuscript. KBT was involved in conceptualization and design of the 39 40 541 study, acquisition of data, drafting of the manuscript and providing critical inputs to 41 42 542 revision of manuscript. HH made substantial contributions to conception and design 43 543 of the study specifically with provision of expertise in medical domain and was 44 on September 28, 2021 by guest. Protected copyright. 45 544 involved in revising the manuscript critically for intellectual content. DBM made 46 47 545 substantial contributions to conception and design of the study specifically, with 48 546 provision of expertise in medical domain and was involved in revising the manuscript 49 50 547 critically for intellectual content. JY made substantial contributions to conception and 51 52 548 design of the study specifically with provision of expertise in financial domain and 53 54 549 was involved in revising the manuscript critically for intellectual content. AC made 55 550 substantial contributions to conception and design of the study specifically with 56 57 551 provision of expertise in social domain and was involved in revising the manuscript 58 59 552 critically for intellectual content. KEL was involved in acquisition of data and in 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 26 of 40

24 1 2 3 553 revising the manuscript critically for intellectual content. NV was involved in 4 BMJ Open: first published as 10.1136/bmjopen-2019-036631 on 23 April 2020. Downloaded from 5 554 acquisition of data and in revising the manuscript critically for intellectual content. 6 7 555 EM was involved in acquisition of data and in revising the manuscript critically for 8 556 intellectual content. KMC was involved in acquisition of data and in revising the 9 10 557 manuscript critically for intellectual content. DADS was involved in acquisition of 11 12 558 data and in revising the manuscript critically for intellectual content. PY was involved 13 14 559 in acquisition of data and in revising the manuscript critically for intellectual content. 15 560 BYT was involved in acquisition of data and in revising the manuscript critically for 16 17 561 intellectual content. EC was involved in acquisition of data and in revising the 18 For peer review only 19 562 manuscript critically for intellectual content. SHY was involved in acquisition of data 20 21 563 and in revising the manuscript critically for intellectual content. YSN was involved in 22 564 acquisition of data and in revising the manuscript critically for intellectual content. 23 24 565 TMT was involved in acquisition of data and in revising the manuscript critically for 25 26 566 intellectual content. YHA was involved in acquisition of data and in revising the 27 567 manuscript critically for intellectual content. KHK was involved in acquisition of data 28 29 568 and in revising the manuscript critically for intellectual content. RS was involved in 30 31 569 acquisition of data and in revising the manuscript critically for intellectual content. 32 33 570 RAM was involved in acquisition of data and in revising the manuscript critically for 34 571 intellectual content. HMC was involved in acquisition of data and in revising the 35 36 572 manuscript critically for intellectual content. TTY was involved in acquisition of data http://bmjopen.bmj.com/ 37 38 573 and in revising the manuscript critically for intellectual content. CN was involved in 39 40 574 acquisition of data and in revising the manuscript critically for intellectual content. 41 575 AC was involved in acquisition of data and in revising the manuscript critically for 42 43 576 intellectual content. CST was involved in conceptualization and design of the study, 44 on September 28, 2021 by guest. Protected copyright. 45 577 analysis and interpretation of data, drafting of the manuscript and providing critical 46 578 inputs to revision of manuscript. All the authors have read and approved the final 47 48 579 version of the manuscript to be published and are agreeable to take accountability of 49 50 580 all aspects of the work involved in the manuscript. 51 52 581 53 582 Ethics approval 54 55 583 The Singapore Stroke Study was approved by the National University of Singapore 56 57 584 Institutional Review Board, SingHealth Centralized Institutional Review Board and 58 585 the National Health Group Domain Specific Review Board. Written informed consent 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 27 of 40 BMJ Open

25 1 2 3 586 was obtained from both the patients and the caregivers in their preferred language by 4 BMJ Open: first published as 10.1136/bmjopen-2019-036631 on 23 April 2020. Downloaded from 5 587 trained researchers. 6 7 588 8 9 10 589 REFERENCES 11 12 590 1. Strong K, Mathers C, Bonita R. Preventing stroke: saving lives around the 13 14 591 world. The Lancet Neurology. 2007;6(2):182-7. 15 592 2. Singapore Stroke Registry. Trends in Stroke in Singapore 2005-2012. 16 17 593 Singapore: National Registry of Diseases Office, Ministry of Health. 18 For peer review only 19 594 3. Cameron JI, Tsoi C, Marsella A. Optimizing stroke systems of care by 20 21 595 enhancing transitions across care environments. Stroke. 2008;39(9):2637-43. 22 596 4. Lutz BJ, Young ME, Cox KJ et al. The crisis of stroke: experiences of patients 23 24 597 and their family caregivers. Topics in stroke rehabilitation. 2011;18(6):786-97. 25 26 598 5. Thom T, Haase N, Rosamond W et al. Heart disease and stroke statistics-- 27 599 2006 update: a report from the American Heart Association Statistics Committee and 28 29 600 Stroke Statistics Subcommittee. Circulation. 2006;113(6):e85-151. 30 31 601 6. Lutz BJ, Young ME, Creasy KR et al. Improving stroke caregiver readiness 32 33 602 for transition from inpatient rehabilitation to home. The Gerontologist. 34 603 2016;57(5):880-9. 35 36 604 7. Tyagi S, Koh GC, Luo N et al. Dyadic approach to post-stroke http://bmjopen.bmj.com/ 37 38 605 hospitalizations: role of caregiver and patient characteristics. BMC neurology. 39 40 606 2019;19(1):267. 41 607 8. Bakas T, Clark PC, Kelly-Hayes M et al. Evidence for stroke family caregiver 42 43 608 and dyad interventions: a statement for healthcare professionals from the American 44 on September 28, 2021 by guest. Protected copyright. 45 609 Heart Association and American Stroke Association. Stroke. 2014;45(9):2836-52. 46 610 9. McCarthy MJ, Lyons KS, Powers LE. Expanding poststroke depression 47 48 611 research: movement toward a dyadic perspective. Topics in stroke rehabilitation. 49 50 612 2011;18(5):450-60. 51 52 613 10. Tyagi S, Tan CS, Koh GC-H. Dyadic approach to outpatient healthcare 53 614 utilization by stroke patients: can caregivers make a difference? Archives of Physical 54 55 615 Medicine and Rehabilitation. 2018;99(10):e45. 56 57 616 11. Lyons KS, Lee CS. The theory of dyadic illness management. Journal of 58 59 617 Family Nursing. 2018;24(1):8-28. 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 28 of 40

26 1 2 3 618 12. Lyons KS, Vellone E, Lee CS et al. A dyadic approach to managing heart 4 BMJ Open: first published as 10.1136/bmjopen-2019-036631 on 23 April 2020. Downloaded from 5 619 failure with confidence. Journal of Cardiovascular Nursing. 2015;30(4S):S64-S71. 6 7 620 13. Moon H, Adams KB. The effectiveness of dyadic interventions for people 8 621 with dementia and their caregivers. Dementia. 2013;12(6):821-39. 9 10 622 14. Em S, Bozkurt M, Caglayan M et al. Psychological health of caregivers and 11 12 623 association with functional status of stroke patients. Topics in stroke rehabilitation. 13 14 624 2017;24(5):323-9. 15 625 15. Harris JE, Eng JJ, Miller WC et al. The role of caregiver involvement in 16 17 626 upper-limb treatment in individuals with subacute stroke. Physical therapy. 18 For peer review only 19 627 2010;90(9):1302-10. 20 21 628 16. Koh GC, Chen C, Cheong A et al. Trade-offs between effectiveness and 22 629 efficiency in stroke rehabilitation. International journal of stroke : official journal of 23 24 630 the International Stroke Society. 2012;7(8):606-14. 25 26 631 17. Koh GCH, Wee LE, Chen C et al. Caregivers and their impact on inpatient 27 632 rehabilitation efficiency and effectiveness amongst recent stroke survivors in an 28 29 633 urbanised Asian society. Am Heart Assoc; 2012. 30 31 634 18. Tsouna-Hadjis E, Vemmos KN, Zakopoulos N et al. First-stroke recovery 32 33 635 process: the role of family social support. Arch Phys Med Rehabil. 2000;81(7):881-7. 34 636 19. Baker AC. The spouse's positive effect on the stroke patient's recovery. 35 36 637 Rehabilitation nursing : the official journal of the Association of Rehabilitation http://bmjopen.bmj.com/ 37 38 638 Nurses. 1993;18(1):30-3. 39 40 639 20. Fang Y, Tao Q, Zhou X et al. Patient and Family Member Factors Influencing 41 640 Outcomes of Poststroke Inpatient Rehabilitation. Arch Phys Med Rehabil. 42 43 641 2017;98(2):249-55 e2. 44 on September 28, 2021 by guest. Protected copyright. 45 642 21. Chen C, Koh GC, Naidoo N et al. Trends in length of stay, functional 46 643 outcomes, and discharge destination stratified by disease type for inpatient 47 48 644 rehabilitation in Singapore community hospitals from 1996 to 2005. Arch Phys Med 49 50 645 Rehabil. 2013;94(7):1342-51 e4. 51 52 646 22. Intermediate and long-term care (ILTC) services. Ministry of Health (MOH), 53 647 Singapore. [cited 2018 November 5, 2018]. Available from: 54 55 648 https://http://www.moh.gov.sg/our-healthcare-system/healthcare-services-and- 56 57 649 facilities/intermediate-and-long-term-care-(iltc)-services 58 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 29 of 40 BMJ Open

27 1 2 3 650 23. Tyagi S, Koh GC-H, Nan L et al. Healthcare utilization and cost trajectories 4 BMJ Open: first published as 10.1136/bmjopen-2019-036631 on 23 April 2020. Downloaded from 5 651 post-stroke: role of caregiver and stroke factors. BMC health services research. 6 7 652 2018;18(1):881. 8 653 24. Bakas T, Austin JK, Jessup SL et al. Time and difficulty of tasks provided by 9 10 654 family caregivers of stroke survivors. Journal of Neuroscience Nursing. 11 12 655 2004;36(2):95. 13 14 656 25. Bédard M, Molloy DW, Squire L et al. The Zarit Burden Interview: a new 15 657 short version and screening version. The gerontologist. 2001;41(5):652-7. 16 17 658 26. Seng BK, Luo N, Ng WY et al. Validity and reliability of the Zarit Burden 18 For peer review only 19 659 Interview in assessing caregiving burden. Ann Acad Med Singapore. 2010;39:758-63. 20 21 660 27. Pearlin LI, Mullan JT, Semple SJ et al. Caregiving and the stress process: An 22 661 overview of concepts and their measures. The gerontologist. 1990;30(5):583-94. 23 24 662 28. Bakas T, Kroenke K, Plue LD et al. Outcomes among family caregivers of 25 26 663 aphasic versus nonaphasic stroke survivors. Rehabilitation Nursing. 2006;31(1):33- 27 664 42. 28 29 665 29. Clark PC, Dunbar SB, Aycock DM et al. Caregiver perspectives of memory 30 31 666 and behavior changes in stroke survivors. Rehabilitation Nursing. 2006;31(1):26-32. 32 33 667 34 668 30. Gonzalez C, Bakas T. Factors associated with stroke survivor behaviors as 35 36 669 identified by family caregivers. Rehabilitation Nursing. 2013;38(4):202-11. http://bmjopen.bmj.com/ 37 38 670 31. Haley WE, Allen JY, Grant JS et al. Problems and benefits reported by stroke 39 40 671 family caregivers: results from a prospective epidemiological study. Stroke. 41 672 2009;40(6):2129-33. 42 43 673 32. Teri L, Truax P, Logsdon R et al. Assessment of behavioral problems in 44 on September 28, 2021 by guest. Protected copyright. 45 674 dementia: the revised memory and behavior problems checklist. Psychology and 46 675 aging. 1992;7(4):622. 47 48 676 33. Tan L, Yap P, Ng WY et al. Exploring the use of the Dementia Management 49 50 677 Strategies Scale in caregivers of persons with dementia in Singapore. Aging & mental 51 52 678 health. 2013;17(8):935-41. 53 679 34. Downey RG, King CV. Missing data in Likert ratings: A comparison of 54 55 680 replacement methods. The Journal of general psychology. 1998;125(2):175-91. 56 57 681 35. Bjerkreim AT, Thomassen L, Brøgger J et al. Causes and predictors for 58 682 hospital readmission after ischemic stroke. Journal of Stroke and Cerebrovascular 59 60 683 Diseases. 2015;24(9):2095-101.

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 30 of 40

28 1 2 3 684 36. StataCorp. 2015. Stata Statistical Software: Release 14. College Station, TX: 4 BMJ Open: first published as 10.1136/bmjopen-2019-036631 on 23 April 2020. Downloaded from 5 685 StataCorp LP. 6 7 686 37. Clark AN, Sander AM, Pappadis MR et al. Caregiver characteristics and their 8 687 relationship to health service utilization in minority patients with first episode stroke. 9 10 688 NeuroRehabilitation. 2010;27(1):95-104. 11 12 689 38. Lai W, Buttineau M, Harvey JK et al. Clinical and psychosocial predictors of 13 14 690 exceeding target length of stay during inpatient stroke rehabilitation. Topics in stroke 15 691 rehabilitation. 2017;24(7):510-6. 16 17 692 39. Tan WS, Chong WF, Chua KS et al. Factors associated with delayed 18 For peer review only 19 693 discharges after inpatient stroke rehabilitation in Singapore. Ann Acad Med 20 21 694 Singapore. 2010;39(6):435-41. 22 695 40. Hinojosa MS, Rittman M, Hinojosa R. Informal caregivers and racial/ethnic 23 24 696 variation in health service use of stroke survivors. Journal of rehabilitation research 25 26 697 and development. 2009;46(2):233-41. 27 698 41. Koh GC, Saxena SK, Ng TP et al. Effect of duration, participation rate, and 28 29 699 supervision during community rehabilitation on functional outcomes in the first 30 31 700 poststroke year in Singapore. Arch Phys Med Rehabil. 2012;93(2):279-86. 32 33 701 42. Cameron JI, Cheung AM, Streiner DL et al. Stroke survivors' behavioral and 34 702 psychologic symptoms are associated with informal caregivers' experiences of 35 36 703 depression. Arch Phys Med Rehabil. 2006;87(2):177-83. http://bmjopen.bmj.com/ 37 38 704 43. Stephens S. Who's There?: When stroke or Alzheimer's changes a person's 39 40 705 behavior, caregiving can become extreme. Here, experienced caregivers, patients, and 41 706 experts share their stories and advice. Neurology Now. 2009;5(4):26-9. 42 43 707 44. Visser-Meily A, Post M, Gorter JW et al. Rehabilitation of stroke patients 44 on September 28, 2021 by guest. Protected copyright. 45 708 needs a family-centred approach. Disability and rehabilitation. 2006;28(24):1557-61. 46 709 45. Glickman LB. Clients with stroke and non-stroke and their guardians’ views 47 48 710 on community reintegration status after in-patient rehabilitation. Malawi Medical 49 50 711 Journal. 2018;30(3):174-9. 51 52 712 46. Young ME, Lutz BJ, Creasy KR et al. A comprehensive assessment of family 53 713 caregivers of stroke survivors during inpatient rehabilitation. Disability and 54 55 714 rehabilitation. 2014 Oct 1;36(22):1892-902. 56 57 715 47. Bhandari A, Wagner T. Self-reported utilization of health care services: 58 716 improving measurement and accuracy. Medical Care Research and Review. 59 60 717 2006;63(2):217-35.

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 31 of 40 BMJ Open

29 1 2 3 718 48. Organization WH. International Classification of Functioning, Disability, and 4 BMJ Open: first published as 10.1136/bmjopen-2019-036631 on 23 April 2020. Downloaded from 5 719 Health: Children & Youth Version: ICF-CY: World Health Organization; 2007. 6 7 720 49. Whiteneck G, Dijkers MP. Difficult to measure constructs: conceptual and 8 721 methodological issues concerning participation and environmental factors. Archives of 9 10 722 physical medicine and rehabilitation. 2009;90(11):S22-S35. 11 12 723 50. Keysor J, Jette A, Haley S. Development of the home and community 13 14 724 environment (HACE) instrument. Journal of rehabilitation medicine. 2005;37(1):37- 15 725 44. 16 17 726 51. Keysor JJ, Jette AM, Coster W, Bettger JP, Haley SM. Association of 18 For peer review only 19 727 environmental factors with levels of home and community participation in an adult 20 21 728 rehabilitation cohort. Archives of physical medicine and rehabilitation. 22 729 2006;87(12):1566-75. 23 24 25 730 26 27 731 Figure 1. Study flowchart 28 29 732 30 31 32 33 34 35 36 http://bmjopen.bmj.com/ 37 38 39 40 41 42 43 44 on September 28, 2021 by guest. Protected copyright. 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

32 http://bmjopen.bmj.com/ 33 34 35 36 37 38 39

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 34 of 40

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 35 of 40 BMJ Open

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 36 of 40

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 37 of 40 BMJ Open

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 38 of 40

59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 39 of 40 BMJ Open

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 40 of 40

1 STROBE Statement—Checklist of items that should be included in reports of cohort studies 2 3

4 Item Page No BMJ Open: first published as 10.1136/bmjopen-2019-036631 on 23 April 2020. Downloaded from 5 No Recommendation 6 Title and abstract 1 (a) Indicate the study’s design with a commonly used term in the title or Title page, Page-1 7 the abstract 8 (b) Provide in the abstract an informative and balanced summary of what Abstract, 9 Page-3 10 was done and what was found 11 Introduction 12 Introduction, 13 Background/rationale 2 Explain the scientific background and rationale for the investigation Page-5-6 14 being reported 15 Objectives 3 State specific objectives, including any prespecified hypotheses Introduction, 16 Page-6 17 Methods 18 For peer review only Methods, 19 Study design 4 Present key elements of study design early in the paper Page-6-7 20 Setting 5 Describe the setting, locations, and relevant dates, including periods of Methods, 21 Page-6-7 22 recruitment, exposure, follow-up, and data collection 23 Participants 6 (a) Give the eligibility criteria, and the sources and methods of selection Methods, 24 Page-7-8 of participants. Describe methods of follow-up 25 NA 26 (b) For matched studies, give matching criteria and number of exposed 27 and unexposed 28 Variables 7 Clearly define all outcomes, exposures, predictors, potential confounders, Methods, 29 Page 8-10 and effect modifiers. Give diagnostic criteria, if applicable 30 31 Data sources/ 8* For each variable of interest, give sources of data and details of methods Methods, Page 8-9 32 measurement of assessment (measurement). Describe comparability of assessment 33 methods if there is more than one group 34 Bias 9 Describe any efforts to address potential sources of bias Discussion, 35 Page-20-21

36 http://bmjopen.bmj.com/ Study size 10 Explain how the study size was arrived at Based on 37 availability 38 of data. 39 Quantitative variables 11 Explain how quantitative variables were handled in the analyses. If Methods 40 (analysis), applicable, describe which groupings were chosen and why 41 Page-10 42 Statistical methods 12 (a) Describe all statistical methods, including those used to control for Methods 43 (analysis), 44 confounding Page-10 on September 28, 2021 by guest. Protected copyright. 45 (b) Describe any methods used to examine subgroups and interactions NA 46 NA 47 (c) Explain how missing data were addressed 48 (d) If applicable, explain how loss to follow-up was addressed Follow-up 49 rates are 50 given under Results, 51 Page-11. 52 Not done. 53 (e) Describe any sensitivity analyses 54 Results 55 Participants 13* (a) Report numbers of individuals at each stage of study—eg numbers Results, 56 Page-10-11 potentially eligible, examined for eligibility, confirmed eligible, included 57 & figure 1 58 in the study, completing follow-up, and analysed for study 59 flowchart. 60 (b) Give reasons for non-participation at each stage Figure 1 (c) Consider use of a flow diagram Figure 1

1 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 41 of 40 BMJ Open

1 Descriptive data 14* (a) Give characteristics of study participants (eg demographic, clinical, Results, 2 Page 11 & social) and information on exposures and potential confounders 3 Table 1, 2

36 http://bmjopen.bmj.com/ 37 38 39 40 41 42 43 44 on September 28, 2021 by guest. Protected copyright. 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

2 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 42 of 40

4 adjusted for and why they were included BMJ Open: first published as 10.1136/bmjopen-2019-036631 on 23 April 2020. Downloaded from 5 (b) Report category boundaries when continuous variables were categorized Table 1, 2. 6 (c) If relevant, consider translating estimates of relative risk into absolute risk for a NA 7 meaningful time period 8 9 Other analyses 17 Report other analyses done—eg analyses of subgroups and interactions, and NA 10 sensitivity analyses 11 12 Discussion 13 Key results 18 Summarise key results with reference to study objectives Discussion, 14 Page-18 15 Limitations 19 Discuss limitations of the study, taking into account sources of potential bias or Discussion, Page-20-21 16 imprecision. Discuss both direction and magnitude of any potential bias 17 Interpretation 20 Give a cautious overall interpretation of results considering objectives, limitations, Conclusion, 18 For peer review only Page-22 19 multiplicity of analyses, results from similar studies, and other relevant evidence 20 Generalisability 21 Discuss the generalisability (external validity) of the study results Discussion, 21 under study 22 limitations Page-20-21 23 24 Other information 25 Funding 22 Give the source of funding and the role of the funders for the present study and, if Funding 26 Page - 23 applicable, for the original study on which the present article is based 27 28 29 *Give information separately for exposed and unexposed groups. 30 31 Note: An Explanation and Elaboration article discusses each checklist item and gives methodological background and 32 33 published examples of transparent reporting. The STROBE checklist is best used in conjunction with this article (freely 34 available on the Web sites of PLoS Medicine at http://www.plosmedicine.org/, Annals of Internal Medicine at 35 http://www.annals.org/, and Epidemiology at http://www.epidem.com/). Information on the STROBE Initiative is

36 http://bmjopen.bmj.com/ available at http://www.strobe-statement.org. 37 38 39 40 41 42 43 44 on September 28, 2021 by guest. Protected copyright. 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

3 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml