To determine the effect of black belt use on the incidence of low back injury in home attendants, a cluster-randomized trial involving employees of nine home attendant agencies in New York City was conducted. Nine agencies employing 12,772 home attendants between June 1997 and September 1999 were randomized into three groups-one group received back belts with use instruction, one group received lifting advice only, and one group served as a control. Low back injury rates per 100 full-time equivalents and rate ratios adjusted for potential confounders were estimated with random-effects Poisson regression. The back-belt group had a lower rate of low back injury than did those in both the advice-only and control groups, though the differences were marginally significant. Age, body mass index, history of back injury, years worked as a home attendant, and level of exercise were associated with risk of low back injury. The findings suggest that use of back belts is associated with some reduction in risk of low back injury.