Dumont 2013.

Methods	Cluster‐randomised trial QUARITE (QUAlity of care, RIsk management, and TEchnology) trial
Participants	Mothers who delivered in 46 first‐level and second‐level public referral hospitals in Mali and Senegal, with > 800 deliveries per year that had a functional operating room and had not previously conducted maternal death reviews.
Interventions	Intervention: a complex intervention to develop local leadership and empower obstetric teams. 6‐day training for 1 doctor and 1 midwife from each hospital: using the ALARM (Advances in Labour and Risk Management) international course. Consisted of 3 days of training in best practices in emergency obstetric care; 1 day of training in maternal death reviews; 1 day of awareness training related to economic, sociocultural and ethical barriers (including sexual and reproductive rights); and 1 day of training in adult education methods. The trainees were recertified on an annual basis. Following the initial training, a multidisciplinary audit committee was formed in each site and trained in the process of undertaking maternal death reviews. The audit cycle and onsite training were then launched with the support of external facilitators (certified instructors) during their quarterly educational outreach visits. Control group: no intervention. Administrators of these hospitals were informed that the 6‐day training workshop would be provided at the end of the trial.
Outcomes	Primary outcome Change in inpatient maternal mortality rate. Secondary outcomes Change in inpatient perinatal mortality rate (divided into stillbirths, neonatal deaths before 24 hours and neonatal deaths after 24 hours but prior to discharge). Resource availability, quantified by the hospital complexity index. Medical practice for emergency obstetric care, assessed through the following essential obstetric interventions, considered effective in reducing maternal and perinatal mortality: assisted delivery (forceps and vacuum extraction), caesarean section, transfusion and hysterectomy, or transfer to another, more specialised health facility. Proportion of women receiving good quality of care (defined as a score of 70% or more on a criterion‐based clinical audit).
Notes	Funded by Canadian Institutes of Health Research.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	The participating hospitals were in 6 strata corresponding to the combination of 2 countries (Mali and Senegal) and 3 hospital types: hospitals in the capital, regional hospitals and district hospitals outside the capital. Investigators attempted to ensure optimal balance between the hospitals assigned to the intervention and the control groups in terms of their number and size (number of deliveries per year). Therefore, within each stratum, they first ranked the hospitals with respect to size, and then used blocked randomisation, with each block of size 2 containing 2 hospitals with adjacent ranks, i.e. of similar size.
Allocation concealment (selection bias)	Low risk	All participating hospitals were randomised simultaneously, after their list was provided, which eliminated any risk of allocation bias.
Blinding of participants and personnel (performance bias) All outcomes	High risk	It was not possible to blind participants and personnel.
Blinding of outcome assessment (detection bias) All outcomes	Low risk	The data collection and the implementation of the intervention were undertaken by different and independent organisations in each country. The organisations were not blinded with respect to randomisation but they were not involved in the assessment of the outcome. The statisticians assessing outcome were blinded to randomisation of each group (Dumont, personal communication).
Incomplete outcome data (attrition bias) All outcomes	Low risk	No hospitals were lost to follow‐up. To include all eligible women in the intention‐to‐treat analyses, missing data for individual characteristics were imputed based on their distributions in the study population. In sensitivity analysis, women who died before labour were excluded because they usually sought care only after developing severe complications at home.
Selective reporting (reporting bias)	Low risk	Protocol was available. Predefined outcomes measurements were reported.
Contamination	Low risk	Randomisation by hospital, and hospitals were in different areas.