Participating hospitals had to change both the structure and process of care to deliver the COMPASS intervention. To avoid contamination of study arms, randomisation was at the hospital level.

Patients having a stroke may not have a choice of hospitals. Therefore, at the point of initial clinical treatment, they would effectively be prerandomised (without choice) based on the facility to which they are taken.

To ensure respect for persons, the IRB required the study team to inform patients in writing of the following:

1. That the hospital is participating in a study.

2. Which study arm the hospital is randomised (thus the patient is randomised).

3. That all patients at this hospital will receive high-quality care.

4. That patients can opt out of sharing their data for research purposes at any time.

5. That patients will have the choice to respond to the 90-day survey at a later date.

The patient handout containing this information was iteratively revised by patient-stakeholders until it satisfied their concerns and met all IRB requirements.

Initiating the consent process in the hospital setting was impractical.

Ensuring that adequately trained study personnel were available at 41 hospitals at all times for patient discharge was considered not feasible and cost prohibitive.

COMPASS patient-stakeholders counselled against consenting at discharge due to the high burden already on patients at that time, who are receiving instructions, paperwork, medications and prescriptions.

Consent was not needed to receive the COMPASS intervention because participating sites had agreed to implement the COMPASS care model as their new standard for postacute clinical care.

Consent was needed to collect data for research purposes.

To ensure respect for persons, consent was sought to collect data for research purposes at two separate occasions:

1. Written consent at the follow-up clinic visit within 2 weeks of discharge, to be allowed to use data collected during routine treatment for research purposes.

2. Verbal consent 90 days after discharge over the telephone, to be able to collect survey data.

Some smaller hospitals had little or no prior research experience and no physicians or high-level administrators were appropriate choices for leading the study.

All study staff needed an appropriate level of research training without overburdening clinical staff with comprehensive CITI training (required for those conducting experimental research or site PIs).

The central IRB assessed the qualifications of local site principal investigators by expertise and role in the management of poststroke patients.

An RN or other mid-level provider was sometimes selected by the local team and approved as PI by the IRB when their role within the local study team was the most appropriate for leading the study.

Site PIs were required to take CITI training for oversight of research conduct and management for the site. Other clinicians involved in the study were offered an abbreviated and tailored research training.

A central IRB was new to some sites.

Some small hospitals had never conducted research and required an FWA application.

The innovative consent design was new to some local IRBs.

The WFUHS IRB spoke with each site to determine whether the site had prior experience in research, had a local IRB, and if so would they rely on central IRB review.

Multiple follow-up calls and written correspondence occurred with sites, as needed.