BACKGROUND: The aim of this cluster-randomized, crossover trial was to compare the efficacy of plain soap and water with an alcohol-based handrub for surgical hand preparation and prevention of surgical-site infection (SSI) in a Kenyan rural hospital. METHODS: A total of 3317 patients undergoing clean and clean-contaminated surgery were included. Follow-up data 30 days after discharge were available for 3133 patients (94.5 per cent). RESULTS: SSI occurred in 255 patients (8.1 per cent), with similar rates for both study arms: 8.3 per cent for alcohol-based handrub versus 8.0 per cent for plain soap and water (odds ratio 1.03, 95 per cent confidence interval 0.80 to 1.33). After adjustment for imbalances between study arms and clustering effects, the main outcome measure remained unchanged (adjusted odds ratio 1.06, 0.81 to 1.38). The duration of surgery and wound contamination class independently predicted SSI. The cost difference between the methods was small (euro4.60 per week for alcohol-based handrub compared with euro3.30 for soap and water). CONCLUSION: There was no statistically or clinically significant difference in SSI rates, probably because more important factors contribute to SSI development. However, this study demonstrated the feasibility and affordability of alcohol-based handrubs for hand preparation before surgery in settings without continuous, clean water. Registration number: NCT00987402 (http://www.clinicaltrials.gov).