Fylkesnes 2004.

Methods	Cluster randomised study conducted in Chelston, a residential area of Lusaka, Zambia. Using the Zambian census of population mapping system, 24 standard enumeration areas (SEA) in Chelston (2786 households) was established. Ten clusters or SEAs (44% of all households) were selected by probability proportional to the measure of size using the number of households in each area (Fylkesnes 1999; Fylkesnes 2004). In the sampled SEAs, all households and members aged ≥15 years were listed and contacted. Two thousand four hundred forty‐five subjects from 10 clusters (SEAs), aged ≥15 years, were randomised. Later, 849 subjects expressed willingness to participate in the study (i.e. to be tested for HIV). The study obtained ethical approval from the National AIDS Research Committee of Zambia. Participants gave informed consent.
Participants	Men and women household members aged ≥15 years who lived in 10 clusters or enumeration areas in Chelston, found at home during the time of the survey and who expressed willingness to be tested for HIV. No exclusion criteria reported.
Interventions	Intervention group (VCT at optional location): A counsellor conducted home visits and offered subjects a choice of counselling at home, clinic or other location. The majority (84%) elected home counselling. The following day, a counsellor visited subjects at home and obtained consent, offered pre‐test counselling and collected blood for HIV testing. A counsellor revisited subjects the next day for post‐test counselling Control group (VCT at local clinic): Gave blood for HIV testing, received VCT and results at local clinic on the same day
Outcomes	Acceptability was defined as using services and was measured at three stages: Pre‐test counselled only, pre‐test counselled and HIV tested, and post‐test counselled and received the test result.
Notes	The study was conducted in 1999. Counsellors were recruited from different geographical areas.
Risk of bias
Bias	Authors' judgement	Support for judgement
Adequate sequence generation?	Unclear risk	Randomisation method not clearly reported. Authors reported as "The sampling design to select 10 clusters (SEAs) was probability proportional to the measure of size using the number of households in each area." No detail on how participants were recruited after the 10 clusters were selected.
Allocation concealment?	Unclear risk	Not described
Blinding? All outcomes	Unclear risk	Blood specimen from subjects was collected by counsellors and tested by technicians at a local facility. No detail if participants, counsellors or technicians were blinded.
Incomplete outcome data addressed? All outcomes	High risk	Randomisation included 2445 subjects, but analysis excluded 1596 (65.3% of total subjects randomised) because they did not agree to HIV testing (60% for the local clinic and 70% for the optional location).
Free of selective reporting?	Low risk	We had no access to the study protocol, but the study stated that acceptability of VCT was the primary outcome.
Free of other bias?	High risk	Authors did not account for clustering in the analysis. Cluster sizes or the ICC were not reported. The unit of analysis was at subject level and not at cluster level (enumeration areas).