Selection and data extraction

Duplicate screening of studies was done on the basis of title and abstract, and the full text. Potentially relevant studies were first identified by title by two independent reviewers. When the title proved to be inconclusive for assessing potential relevance, abstracts were read to decide whether a specific study should be included. All the identified studies were collated into EndNote for easy management of our references. Duplicates were removed from EndNote and the remaining studies were exported to Covidence software for screening. The screening process was conducted by two independent reviewers and was guided by the inclusion and exclusion criteria. Discrepancies during the screening process were resolved through discussions among the team.

We abstracted data on article characteristics: country where research was conducted; LHW programme characteristics and contextual factors, such as type of intervention, role of LHWs in intervention, frequency and intensity of engagement; barriers and facilitators to programme implementation. In addition, results of any formal assessment or evaluation of HIV specific MCH outcomes, such as MTCT rate, HIV testing, ART initiation among HIV positive pregnant and/or breastfeeding women and their infants, adherence to ART, uptake of EID of HIV and infant HIV status. Data abstraction was conducted using a standardised data extraction form that was developed a priori and pilot tested on a sample of five included papers. Two reviewers carried out data abstraction, and two other reviewers checked quality for consistency and accuracy.