To evaluate quality of care, a two-step approach seems appropriate. First, highly-structured explicit criteria, based on patient outcomes such as mortality, readmissions, or unusually long lengths of hospital stay, might help identify adverse events using routinely-collected discharge data. Then, process criteria might be used for subsequent medical record reviews to determine whether a quality problem exists. Large administrative data bases suggest the possibility of developing an epidemiology of quality of care; understanding how quality problems are distributed across the hospitals in a province seems feasible. Population-wide data are essential for comprehensive follow-up and for effective studies of medical practices. Hospital-based follow-up can miss important events; we found the relative percentage of short-term readmissions to hospitals other than the hospital of surgery startling. However, hospital-based data can sometimes be used in place of the more costly and harder-to-generate population data for quality monitoring. For example, in examining correlations among various outcome indicators following five common surgical procedures, we found the ranking of hospitals according to inhospital mortality to be highly correlated with their ranking according to 30-day post-surgical mortality.