The purpose of this multicenter prospective cohort study was to assess personal and job-related factors as determinants of incidence of reported back injuries among nurses. The study population included 5,649 nurses who were surveyed by questionnaire and then observed for a 12-month study period. An annual injury rate of 4.9% was observed. Four factors were found to be significant (P less than .01) predictors of back injury. All four factors--service area, lifting, job category, and previously reported back injury--maintain significance when a forward stepping model of logistic regression is applied. The adjusted odds ratios observed are 4.26 for service areas where lifting occurs most as compared with areas where lifting occurs least; 2.19 for daily lifters v light, occasional, and nonlifters; 1.77 for nursing aides v registered nurses and supervisory personnel; and 1.73 for individuals who have previously reported back injury v those who have not reported previous injury. These findings strongly suggest that job-related rather than personal characteristics are the major predictors of back injury in nurses.