With the proliferation of new health programs, such as Health Maintenance Organizations (HMO's) and Professional Service Review Organizations (PSRO's), the task of evaluating the impact of such programs on the health delivery systems and on the health of the American people becomes more urgent. Thus far no experimental or quasi-experimental designs have been found that are both feasible and satisfactory. Some quasi-experimental designs, particularly interrupted time series, have been suggested as a possible solution to the problem. The strengths and weaknesses of this and other designs, as well as the statistical problems associated with them, are discussed. (Author)