Bias is unavoidable in research, but efforts to minimize it are essential if we want to draw valid conclusions. Mortality reviews can be useful tools for managing population health but are fraught with difficulties that are often overlooked.1-4 Bias occurs in the creation of pathology reports when different pathologists use different diagnostic criteria, have differing levels of experience and confidence, and conduct investigations of varying depth. The only way to minimize these types of bias is to have pathologist(s) review and standardize each report. Bias is introduced in the extraction of data from reports when the investigator subjectively interprets a finding in a way that gives the desired diagnosis or outcome, when interpretation of one finding influences the interpretation of another, and when the investigator only relies on the most readily available data rather than the most reliable. These types of bias can be minimized by developing detailed case definitions, protocols for record review, and procedures for data interpretation in advance. Bias in the final interpretation can occur if missing data are not handled properly, when causal inferences are drawn without appropriate controls, when a single disease process is misinterpreted as multiple processes and vice versa (often due to confusion over diagnostic terminology), and when the population to which the results apply (the population at risk) is misidentified. Avoiding these types of bias requires careful study design, addressing the other types of bias described above, and if necessary, recruiting additional expertise for the study.
1. Croskerry P. 2003. The importance of cognitive errors in diagnosis and strategies to minimize them. Acad. Med. 78:775–80.
2. Delgado-Rodríguez, M., and J. Llorca. 2004. Bias. J. Epidemiol. Community Health. 58:635–41.
3. Graber, M., R. Gordon, and N. Franklin. 2002. Reducing diagnostic errors in medicine: what's the goal? Acad. Med. 77:981–92.
4. Worster, A., and T. Haines. 2004. Advanced statistics: understanding medical record review (MRR) studies. Acad. Emerg. Med. 11:187–92.