This study compares data regarding alcohol involvement from police records and from chemical analyses of body fluids taken prior to or after death of 121 traffic fatalities in Washtenaw County, Michigan. Differences regarding the choice of who will or will not be screened for alcohol were found. The police and emergency room personnel were more likely to focus on males and drivers, while medical examiners were less biased. The police documented whether or not drinking took place in only 36% of the cases and suspected drinking in only half of these cases. Males and at-fault drivers were most likely to be investigated. Blood alcohol level was measured before death in 11 of 29 emergency room treated victims, with 10 (91%) positive samples. All but two of those tested before death were drivers, and all but one were males. After death, blood alcohol was measured in 47% of the 121 cases, with a higher proportion of males and motor-vehicle occupants tested, compared to females and pedestrians. Alcohol was detected in 63% of the samples. A lower mean blood alcohol concentration was found in victims who received intravenous treatment, and a higher proportion of positive samples was found in victims who died immediately in the crash. Thus, the frequency of alcohol-related traffic fatalities varied between the different data sources. The police records revealed 51%, the emergency records 91%, and the medical examiner records 63% with alcohol involvement. This wide discrepancy has the potential of leading to erroneous results here and possibly in studies done elsewhere.(ABSTRACT TRUNCATED AT 250 WORDS)