Abstract A mathematical model for predicting the risk of acquisition of postoperative wound infection in individuals or groups of patients is described. It is based on data from prevalence surveys of 41 hospitals and includes 1980 wounds. The factors included in the model i.e., age, sex, length of pre-operative stay, type of operation, wound drainage, number of occupied beds in ward, and special factors e.g., diabetes, steroid therapy, were obtained by stepwise regression analysis of the original data. Most of the ward facilities and practices were excluded as non-significant. The model has been modified for use with incidence studies and its accuracy confirmed in a further 1331 patients by comparing predicted and recorded infection rates.