OBJECTIVE: To assess the completeness of registration of maternal mortality by comparing various data sources and to review the quality of information pertaining to individual cases. DATA SOURCE AND METHODS: Female deaths aged 10-50 years in the Greater Accra region in Ghana from January 1, 2000 to December 31, 2000 were recorded. Identified maternal deaths (International Classification of Diseases, 9th revision) in four major hospitals in the region were compared with those registered in the civil registers. RESULTS: During the study period, 148 maternal deaths were identified and a total of 21,183 live births were recorded in the study hospitals, yielding a maternal mortality ratio of 699/105 live births. This ratio is more than six times the officially reported maternal mortality ratio for the whole region in the same period. Ninety-two percent of the deaths were due to direct obstetric causes and 8\% to indirect causes. One hundred and twenty-one of the cases were correctly classified and registered. By using the multiple source approach, 27 new cases were identified, an under-registration of 18\%. The sources of error were misclassification (n=25) and error in registration (n=2). CONCLUSION: The process of registration of deaths in this region is incomplete. Enhanced registration systems (including routine linkage of death and birth certificates) on a regular basis could be used in Ghana.