A computer model has been developed to determine cost-effectiveness of screening and treatment for diabetic retinopathy from a societal viewpoint. This model was used to evaluate biannual and annual screening programs using ophthalmoscopy, fundus photography with a "nonmydriatic camera," and photography with a "mydriatic camera." Computations were performed for three subpopulations formed by patients with younger onset diabetes (age at diagnosis less than 30 years) of 5 years or more duration, with older onset diabetes (age at diagnosis greater than or equal to 30 years) who are taking insulin, and with older onset diabetes not taking insulin. Population characteristics are from a well-described southern Wisconsin population where data are available, but the computer model may be specialized to other population. Generally costs of screening programs appear to be recovered by avoided costs of blindness in the population subgroups taking insulin; however, the cost of screening programs generally are not recovered by avoiding costs of blindness in the older onset population subgroup not taking insulin. It was estimated that supplying annual examination with mydriatic fundus photography as a screening program to a cohort of 1,000 diabetics from the younger onset population who have been diagnosed at least 5 years and who are currently not receiving care might save 319 sight years over the lifetime of the cohort. This program will save 62 sight years in an older onset cohort who are taking insulin, and 21 sight years in the older onset population not taking insulin (all benefits are presented as present values computed with an annual discount rate of 5%). Other programs achieve slightly lower savings in each subpopulation.