Wu 2019.

Study characteristics
Methods	Study design: Cluster‐RCT; Unit of randomisation: community based organisations (CBOs) 150 CBOs assessed for eligibility 22 CBOs eligible and 22 CBOs randomised; 20 operating once trial began Sample size/power calculation: none identified Country of study: Baltimore, Marylands, US Setting: urban privately funded academic health care system (Johns Hopkins Health System (JHHS)). JHHS comprises Johns Hopkins University School of Medicine; The Johns Hopkins Hospital (1177 patient beds; 2230+ full‐time attending physicians) Johns Hopkins Bayview Medical Centre (448 patient beds; 710+ attending physicians). Study duration: 2014 ‐ 2016 Broader consumer involvement: two community leaders were paid co‐principal investigator and co‐investigator on the study team. Both reviewed and approved all stages of proposal and helped identify and recruit CBOs for participation, and recruited members to be interviewed. Both were involved in the partnership intervention co‐development process, implementation and evaluation. The community based co‐principal investigator led meetings with CBO leaders. Both contributed to paper writing and submission.
Participants	Consumer partnership participants: (n = unknown), community leaders from not‐for‐profit CBOs that served adults and addressed one or more social determinants of health, located within the zip code of East or Southeast Baltimore Age: not described Sex: not described Education level: not described Socioeconomic status (SES): not described Diagnosis: not described Other: not described Provider partnership participants: (n = unknown), Johns Hopkins staff members as part of study team Age: not described Sex: not described Education level: not described SES: not described Years practicing: not described Other: not described Health service user trial participants: (n = 6491 patients assessed for eligibility; 5255 eligible patients), high‐risk Medicare and Medicaid Johns Hopkins Community Health Partnership (J‐CHiP) outpatient patients (adults, at least one chronic condition, at least one visit to J‐CHiP site, high risk for future hospitalisation). Patients with monthly healthcare utilization data, including emergency department visits and days hospitalised during both pre and post intervention and known home address were included. Age (mean): intervention 62 years; control 62 years Sex (female): intervention 65%; control 63% Education level: not described SES (Insurance type): intervention: Medicare 64%; Priority Partners MCO 36%; control: Medicare 65%, Priority Partners MCO 35%; Other (race): intervention: African American 59%; Asian 1%; White 35%; Hispanic 1%; Native American 0%; Unknown/other 5%; control: African American 52%; Asian 1%; White 42%; Hispanic 0%; Native American 0%; Unknown /other 4%. Health service provider trial participants: (n = unknown), outpatient staff (case managers, community health workers, health educators and behavioural health specialists) and inpatient staff (social workers, case managers, hospitalists and nurse who help discharge patients to home) from JHHS. Sex: not described Education level: not described SES: not described Number of years practising: not described Other: not described
Interventions	Randomised to intervention: 11 CBOs; 1997 patients (analysed: 10 CBOs; 1864 patients) Nature of intervention: After randomisation the partnership Baltimore CONNECT (Community‐based Organisations Neighbourhood Network: Enhancing Capacity Together) co‐developed an intervention/toolkit (Healthify). Intervention aim: to "enhance the capacity of both CBO staff and frontline hospital workers to address client needs by strengthening the bidirectional flow of information about health and social services and building networks that span both entities." (p e32, Wu 2019) Context of partnering: Partnership based on the African Partnerships for Patient Safety Engagement (ACE) Framework which is underpinned by Community Based Participatory Research and Participatory Action Research approaches. Decision‐making activity: "To begin, iCBO (intervention CBO) leaders completed a needs assessment to identify commonly faced challenges. The most salient issues identified were: referring clients to organisations for support, developing a stronger relationship with other CBOs to better serve clients, and interfacing with JHHS. Results of needs assessment were directly linked to formation of strategies to enhance co‐ordination of health and social services." (p 302, Wu 2018) Meeting format, duration, frequency and location: face‐to‐face (plus email and phone calls); each meeting 1.5 to 2 hours in duration; conducted monthly (for 6 months of co‐development and 12 months of the trial), with location rotated among iCBOs. Partnership duration: 6 months Training/support: no training described; each iCBO assigned a student research assistant Decision‐making process, attempts to resolve conflict: not described Diversity and ratio of consumer and provider participants: not described Attempts to address intrinsic power imbalances: not described Theoretical basis for partnering: Draws from the following 8 concepts of the ACE Framework: establish a community engagement advisory board; know the community; establish an enabling community engagement environment; raise patient quality and safety awareness locally and nationally; collect community knowledge and experiences; ensure robust communication mechanisms; feed into monitoring and evaluation; and develop a community ripple effect. Tailoring/modification/adapting: reverse innovation of the ACE Framework by adapting the ACE approach to partner with iCBO leaders to co‐develop and implement a set of interventions, or toolkit Fidelity/integrity: No assessment of fidelity, but there was survey data on trust, partnership etc (secondary outcomes). Some contamination effects possible, as intervention and control CBOs may have shared some clients and provided services to the J‐CHiP cohort of patients. Healthify also listed both intervention and control CBOs. Randomised to control: 11 CBOs, 3258 patients (analysed: 10 CBOs, 3053 patients) Nature of comparison: usual practice plus J‐CHIP ‐ no other description Co‐intervention: Johns Hopkins Community Health Partnership (not described)
Outcomes	Outcomes measured at Baseline (T0) and follow‐up (12 months after intervention had ended ‐ T1): Outcome measures relevant to the review (reporting available comparative data for longest time point measured): Health service alterations data (changes to services resulting from partnership decisions): CBO staff reported receiving 1+ referral from healthcare staff (T1)** Health service user (trial participant) health service utilisation patterns: Client reported referred from the healthcare system to a CBO (T1) Health service provider (trial participant) reported outcomes: Healthcare staff barriers to referring patients to CBOs, capacity for CBOs and healthcare organizations to work together, confidence in knowledge about community resources, confidence in the capacity of CBOs to meet their clients’ needs, and number of referrals to iCBOs and cCBOs (T1). Data reported as pre‐post separately for inpatient and outpatient staff (but not comparatively for intervention vs. control group) Adverse events: “no harm was done to any of the participants in either group” (p e36, Wu 2019) Consumer (partnership participant) reported behaviours/attitudes outcomes: CBO staff ratings of working together with health service moderately or extremely well (T1)** Measures of partnership among provider and consumer partnership participants: iCBO partner satisfaction with the partnership self‐assessment tool (T1). Data reported for intervention consumer partners only. Other study outcome measures (extracted but not reported in review): CBO client healthcare utilisation (health system billing (claims) data) (Mean and Standard Error*) CBO client reported (survey in‐person/by phone) attitudes and behaviours (%)** CBO staff reported (online/paper survey) attitudes and behaviours (%)** Following outcomes were not comparative (reported pooled data only): Healthcare staff reported (online/paper survey) attitudes (J‐CHiP outpatient staff) (%) Healthcare staff report (online/paper survey) attitudes (hospital staff) (%) Following outcomes measured at T1 for intervention arm only: iCBO partner (online/paper survey) satisfaction with the partnership (partnership self‐assessment tool) iCBO staff use of Healthify (Healthify analytics usage data) In/out patient staff and iCBO staff (post event survey) usefulness of meet and greet sessions (% agree or strongly agree)
Notes	Funding: Patient‐Centred Outcomes Research Institute project grant (CD‐12‐11‐4948). Viva Dadwal supported by Fulbright Canada and the U.S Department of State's Fulbright Scholarship Programme. Conflict of interest: No financial disclosures reported by authors Other: *calculated standard deviation from standard error; ** calculated events from % data.For outcomes reported by the trial but not included in this review, authors identified through consensus that such outcomes were less relevant to the outcomes of the review (where multiple measures of an outcome were reported).
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	High risk	Cluster‐RCT "a restricted randomization process was conducted that constrained the allocation of organisations based on their ZIP code, client population (small, medium or large) and the type of service offered. There was purposeful balance in the number of churches, neighbourhood associations, and Hispanic serving organisation in each group as well as the primary type of services provided (e.g. food, housing, clothing)" (p e33, Wu 2019). Rated as at high risk ‐ even if balance was managed on these factors there might be other potential confounders which were not considered.
Allocation concealment (selection bias)	High risk	"The assignment of CBOs could not be blinded which introduced potential for bias." (p 305, Wu 2018)
Blinding of participants and personnel (performance bias) All outcomes	Unclear risk	CBO clients were not likely to be aware of an individual CBO's group assignment; blinding of personnel not described. CBOs were randomised and therefore in the intervention group the leadership (at least) were aware of the intervention; control group may not have been aware of assignment.
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Not described in manuscript. Author response indicates that primary outcomes were all electronic data, with statisticians blinded to assignment.
Incomplete outcome data (attrition bias) All outcomes	Low risk	A total of 205 and 133 patients were excluded from the comparison and interventions arms respectively who did not have any utilisation data available either in the pre‐or post intervention period; these losses were about 7% in both groups; reasons were similar, and numbers remaining for analysis still large.
Selective reporting (reporting bias)	Unclear risk	All outcomes measured reported in published methods are reported and trial protocol is registered NCT02222909 but protocol only lists primary outcomes measured i.e. Emergency Department visits + hospital days [Time Frame: Up to 1 year]; Number of Emergency Department Visits + Number of Hospital Days in the past 6 months for patients enrolled in the Johns Hopkins Community Partnership; and Change in number of clients seen per month [Time Frame: September 2014 ‐ September 2015] Change in median number of clients seen by Community Based Organizations per month; additional outcomes are reported in publications.
Other bias	Low risk	None.
Selective recruitment of participants	Unclear risk	"For allocation of J‐CHiP patients to a treatment arm, those who lived closer to an iCBO were allocated to the intervention group, and those who lived closer to a cCBO were allocated to the control group. Google application programming interfaces were used to determine the distance between each patient's residence and each of the participating CBOs." (p e33, Wu 2019); timing not mentioned.

ACE: African Partnerships for Patient Safety Engagement Framework; ANC: Antenatal Care; APSQ: Attitudes to Patient Safety Questionnaire; CBO: COmmunity Based Organisation; CHC: Communal Health Centres; CFG: Critical Friends Group; CHDR: Community Health Data Review; CHW: Community Health Worker; CONNECT: Community‐based Organisations Neighbourhood Network: Enhancing Capacity Together; IPQ: Improving Practice Questionnaire; J‐CHiP: Johns Hopkins Community Health Partnership; JHHS: Johns Hopkins Health System; MNHG: Maternal and Newborn Health Groups; NMR: Neonatal Mortality Rate; VHW: Village Health Worker.