The processes of sick-leave inception and work resumption are governed by many interrelated factors. Hence conventional univariate models of these processes do not provide an adequate basis for explanation and prediction. In this study, multivariate models were used to fit absenteeism data for a large government department. The variables considered were the age, sex, and grade of the employees concerned; the day, month, and year of inception of sick leave or resumption of work; the section in which the employees worked and the size of that section. The models were used to predict the rates of sick-leave inception and work resumption in a test set of employees. The predicted inception rates appear to agree well with the observed values.