Loevinsohn 1995.

Methods	Study design: CBA study Duration of study: June 1991 to January 1992 Unit of allocation: Province (4 intervention, 6 control) Method of allocation: Non‐random; allocation by judgement of programme managers Unit of analysis: Health facilities BHS and RHU (54 intervention, 70 control) Sampling: Intervention and control provinces purposively selected from the same regions. 123 BHS and RHU selected randomly proportional to the population size. 5 mothers selected in each catchment area using WHO cluster sampling.
Participants	Country: Philippines Income classification: Low‐income Geographical Scope: Four remote provinces with poor health status purposively selected in the north, central and south Philippines Rural/urban: Not stated Setting: Health facilities Supervisees: Midwives educated to high school plus two‐year course and training seminars Supervisors: Public health nurses, municipal health officers, district and provincial supervisors Patients/clients: Pregnant women and general population
Interventions	Stated purpose: Improve health worker performance using trained supervisors and systematic checklist Description intervention: Supervisory visits using an ISC. Supervisors all received structured standard training on supervision. ISC contained scores based on percentage of assistance activity correctly performed. 20 items x 3 maximum points = 60. Control: Routine supervision Training: Training in intervention group: 2 days on the use of ISC and theory of supervision and how to give effective feedback Frequency of supervisory visits: Average of 4.5 supervisory visits over 6 months in intervention and control groups Co‐interventions: None
Outcomes	Primary outcomes: Health worker performance at facility‐level, as measured by combined scores on 20 indicators Secondary outcomes: Individual indicators, e.g. prenatal visits, record keeping, drugs supply, immunisation coverage, new acceptors for family planning and health worker knowledge Time points when outcomes measured: 6 months after implementation
Notes	Funding: Philippine Department of Health and USAID.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	High risk	"Four remote provinces with poor health were chosen by the public health programme managers in the DOH [not spelled out in the text] central office to serve as the experimental area"
Allocation concealment (selection bias)	High risk	This is a CBA study
Blinding (performance bias and detection bias) All outcomes	Low risk	“In June 1991, a baseline evaluation was conducted by central office staff in the selected facilities. The assessors were blind as to whether the province they visited would be in the control or experimental group”
Incomplete outcome data (attrition bias) All outcomes	High risk	“Of the 124 health facilities that were evaluated during the baseline survey, 12 (7 experimental and 5 control) facilities could not be followed up 6 months later, due to peace and order problems.  The 12 facilities had an average baseline performance of 23.2 points (95% CI = 15.8 to 30.6) which was slightly lower than the facilities where follow‐up was possible (average. = 26.7, 95% CI = 25.2 to 28.0)”.
Selective reporting (reporting bias)	High risk	Only 6 out of 20 specific indicator scores reported
Other bias	High risk	Clustering not taken into account in the analysis.
Baseline outcome measurements similar?	Low risk	"At baseline the control and experimental health facilities were similar...including their total scores on 20 indicators contained in the supervisory checklist"
Baseline characteristics similar?	Low risk	"At baseline the control and experimental health facilities were similar in all measured characteristics...."
Protection against contamination?	Low risk	“Supervisors of the 4 experimental provinces were given two days of training on the use of the ISC...” “These 4 provinces then implemented the ISC in all their health facilities’. “In the control provinces, no training or copies of the supervisory checklist were provided”