Random and systematic errors in case-control studies calculating the injury risk of driving under the influence of psychoactive substances

Abstract

Between 2006 and 2010, six population based case–control studies were conducted as part of the European research-project DRUID (DRiving Under the Influence of Drugs, alcohol and medicines). The aim of these case–control studies was to calculate odds ratios indicating the relative risk of serious injury in car crashes. The calculated odds ratios in these studies showed large variations, despite the use of uniform guidelines for the study designs. The main objective of the present article is to provide insight into the presence of random and systematic errors in the six DRUID case–control studies. Relevant information was gathered from the DRUID-reports for eleven indicators for errors. The results showed that differences between the odds ratios in the DRUID case–control studies may indeed be (partially) explained by random and systematic errors. Selection bias and errors due to small sample sizes and cell counts were the most frequently observed errors in the six DRUID case–control studies. Therefore, it is recommended that epidemiological studies that assess the risk of psychoactive substances in traffic pay specific attention to avoid these potential sources of random and systematic errors. The list of indicators that was identified in this study is useful both as guidance for systematic reviews and meta-analyses and for future epidemiological studies in the field of driving under the influence to minimize sources of errors already at the start of the study.