OBJECTIVES: To use routine data to study outcomes of all knee and hip joint replacements in an unselected population with universal coverage of healthcare. METHODS: The study population was all resident in Tayside, Scotland from 1989-1995. Joint replacements, and subsequent hospitalizations were identified from a hospital discharge database. Death was recorded from a community health database. The electronic records were validated against primary case records for a random sample. RESULTS: We identified 1,649 people who had first primary hip replacement and 989 first primary knee replacements. Case note validation was performed for 274 patients, of whom 143 had no previous record of hip or knee surgery and 131 had previous surgery. The proportion of valid cases was higher in the patients with no previous surgery (98% vs 57%). The most common mis-classification was revision arthroplasty as primary arthroplasty because of incomplete recording of previous joint replacement. There were 353 deaths in the hip implant cohort (21.4%) and 187 deaths in the knee implant cohort (18.9%); 6.4% of hip recipients and 5.5% of knee recipients died within the first year of implant. Four percent of hip replacements were for management of hip fracture and they had significantly reduced survival compared to the remaining patients (deaths at 60 days, 1.6% vs. 0.8%). Significantly more knee recipients than hip recipients required revision of the joint replacement (4.4% vs 2.4%, p=0.007). Although time to hip implant revision varied with the year of first implant (with 1990 recipients showing the shortest time to revision), this effect did not occur among knee implant recipients. CONCLUSIONS: This study demonstrates the suitability of the Scottish Morbidity record for postmarket surveillance of medical device implants. The electronic data come from a stable population with universal medical coverage and outcomes can be validated by review of primary case records.