Objective: To calculate whether the introduction for systematic use of an epicrisis form for deaths improved the quality of mortality records in a primary care centre (PCC)., Design: Before and after study., Setting: Urban, with a population of 31 000. Patients and other participants. All the deaths recorded at the PCC before 1996 (N=396) and in the period from January 1999 to August 2000 (N=160)., Main Measurements and Results: The number of deaths recorded at the PCC represents only 32.5% of the official data obtained from the municipal registry in the 1996 evaluation, and 51.5% in the 2000 evaluation. In 2000 the epicrisis forms were filled out in 76.8% (95% CI, 65.0-88.6%) of all deaths. In the 1996 evaluation the place of death was not recorded in 65.5% (95% CI, 59.0-71.8%) of cases, but in 2000 this figure was 22.5% (95% CI, 19-26%). Undefined causes of death dropped from 37% (95% CI, 33.4-40.6%) in 1996 to 30.5% (95% CI, 25.8-35.2%) in 2000. Recording of cause of death rose from 64% (95% CI, 57.8-70.4%) in 1996 to 73% (95% CI, 61.8-84.4%) in 2000., Conclusions: Official data reveal under-recording of mortality at the PCC. The introduction of an epicrisis form for deaths has improved the quality of the centres mortality records.