Road selection for hazardous materials transportation relies heavily on risk analysis. With risk being generally expressed as a product of the probability of occurrence and the expected consequence, one will understand that risk analysis is data intensive. However, various authors have noticed the lack of statistical reliability of hazmat accident databases due to the systematic underreporting of such events. Also, official accident databases alone are not always providing all the information required (economical impact, road conditions, etc.). In this paper, we attempt to integrate many data sources to analyze hazmat accidents in the province of Quebec, Canada. Databases on dangerous goods accidents, road accidents and work accidents were cross-analyzed. Results show that accidents can hardly be matched and that these databases suffer from underreporting. Police records seem to have better coverage than official records maintained by hazmat authorities. Serious accidents are missing from government's official databases (some involving deaths or major spills) even though their declaration is mandatory.