Gulmezoglu 2007.

Methods	Study design: CRCT Data collection: data were collected from 1000 consecutive deliveries in each hospital, at baseline and at end of study, or for 6 months, whichever came first. "Each computer has a log file to indicate how many times the program is accessed. We shall analyse the log files as a proxy indicator of RHL use in the intervention hospitals, acknowledging that these may not relate directly to change in behaviour." See method's paper (Gülmezoglu 2004) Unit of analysis issues: number of hospital was unit of analysis
Participants	Participants: hospital obstetric/gynecologic/maternity units. Physicians, nurses, midwives (any practitioner participating in obstetrics and gynecology or maternity care could participate Physicians predominated in Mexico (median number of doctors per hospital was 20 (range 7‐102) in IG and 14 (range 8‐53) in CG hospitals; in Thailand, median number of doctors was 6 (range 3‐18) in IG and 5 (range 2‐11) in CG hospitals; nurses were 15 (10‐28) in IG and 13 (7‐20) in CG hospitals(pp. 19‐20) Total hospitals randomized: 40. IG: 22; CG: 18 Stratification: hospitals were stratified "based on country, type of hospital and number of births per year (>5000 or <=5000)" (p. 17, col 1, para 4) Thailand stratification: 18. Large: IG 3; CG 3. Small: IG 6; CG 6 Mexico stratification: 22. Large social security hospital: IG 3; CG 2; large public hospital: IG 2; CG 1. Small social security hospital: IG 2; CG 1. Small public hospital: IG 6; CG 5 Setting: 40 hospitals with maternity units with > 1000 deliveries per year. 22 in Mexico City; 18 in the northeast region of Thailand Countries: Mexico and Thailand Inclusion criteria: maternity units of hospitals with > 1000 deliveries per year, not associated directly with a university or other academic/research department
Interventions	Description: The first part of the intervention, which might be defined as 'organizational', consisted in meeting with hospital directors/heads of obstetrics and gynecology departments with the goal of ensuring organizational buy in The intervention consisted of 3 interactive workshops using the WHO RHL. The focus of the workshops was to educate users on the content of the RHL and how to use RHL (e.g. "The focus of the workshops was to provide access to knowledge and enable its use" (p. 16, abstract)) A computer and support for using both the computer and RHL were provided at each hospital Workshop 1: focused on giving information about the project, WHO's role, the principles of evidence‐based decision making and presenting RHL Workshop 2: focused on the content of RHL Workshop 3: how to implement change Type of intervention: Educational sessions (e.g. workshop) Distribution of educational material (e.g. RHL database, posters, brochures about RHL) Organizational (e.g. provision of computer hardware) Organizational (e.g. study authors met with hospital leaders to achieve buy‐in/cooperation prior to educational component of intervention) Study period: October 2001 to October 2002 Intervention delivery periods: over a period of 6 months, at time 0, after 6 weeks and after 6 months. Note: time delays were experienced in Thailand resulting in up to 6 months between second and third workshop Duration of intervention: 3 workshops (length of workshop not stated) Follow‐up: 10‐12 months from the time of the first workshop; or 4‐6 months after the third workshop Data collection time: 4‐6 months following the third workshop Comparison: no workshops; but access to WHO RHL
Outcomes	Changes in 10 selected clinical practices as recommended in RHL: social support during labor; MgSO4 for eclampsia; corticosteroids to women with preterm birth; selective episiotomy; uterotonic use after birth; breastfeeding on demand; external cephalic version; iron‐folate supplementation; antibiotic use at cesarean section; vacuum extraction for assisted birth Proportion of staff using RHL once a month measured from baseline to post intervention
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	"We used a stratified cluster randomised design, with the hospitals as the units of randomisation. The stratified allocation was based on country, type of hospital and number of births per year (>5000 or 5000 or less) (Figure 1). The random allocation sequence was produced centrally by WHO in Geneva, assigning hospitals at random in each stratum to intervention or control. Country investigators were informed of the allocation status of the hospitals after collection of baseline data was completed and when the first workshop had to be organised as required in the protocol." Note: additional contact with authors provided further information that this was done using PROC PLAN of SAS software
Allocation concealment (selection bias)	Low risk	"Country investigators were informed of the allocation status of the hospitals after collection of baseline data was completed and when the first workshop had to be organized as required by the protocol." Note: additional contact with authors provided further information that the randomization occurred at 1 time point, and allocation was concealed until after this time point
Blinding (performance bias and detection bias) All outcomes	Low risk	"Hospital staff were unaware of the primary outcome practices." "Field workers not involved in the implementation of the trial collected outcome data" (p. 19)
Incomplete outcome data (attrition bias) All outcomes	High risk	The main outcome measures as designated in the study protocol (e.g. changes to selected clinical practices) were measured for CG and IG in all hospitals Process outcomes were measured in 2 ways ‐ by self report (questionnaires), and objectively (e.g. use of the RHL measured by computer log files) Information on Thai CG was lost Although the aim was to include all hospital staff in the workshops (IG), attendance rates at workshops varied and there was a large staff turnover, as well as participation from students and staff from other departments. "We aimed to include all staff (doctors, midwives, interns and students) at all three workshops. The highest attendance was at the first workshops, but the other two workshops were also well attended. It was not possible to measure attendance with high precision due to staff turn over and participation of students and staff from other departments (anaesthetists, neonatologists)" (p. 17)
Selective reporting (reporting bias)	Low risk	Outcomes delineated in protocol (http://www.isrctn.com/ISRCTN14055385) were reported in study