The investigation of deaths from ill-defined causes (DIDC) has been a strategy of health services to reduce the proportion of these events. This study aimed to estimate the adherence of municipalities to the use of recommended forms in the investigation of DIDC and the impact of these investigations on the reduction of these deaths in the Mortality Information System. The use of the Investigation of Death from Ill-defined Cause (IOCMD) and Verbal Autopsy (VA)forms and the proportion of reclassified underlying cause of death following investigations were analyzed in a probabilistic sample of 27 municipalities of Bahia state, and its capital Salvador, in 2010. Of the 27 municipalities, approximately 50% used the recommended forms to investigate DIDCs. Of the 1,092DIDCs in the sample, 53.1% were investigated: in 40.5% of the cases, only the IOCMD form was used; in 15.3%, only the VA form was used; and both forms were used in 14.3% of the cases. The investigation of DIDCs reduced the percentages of these deaths from 16.5% to 9.9% and proved to be more effective when performed using the recommended forms.