On the incidence of errors in large epidemiology studies with comments on evaluation, interpretation, and future...

Large epidemiological studies with over 5000 cases have the advantage of providing sufficient case numbers for the statistical evaluation of sub-groups. They are associated with a higher error rate in data collection and the influence of differences in judgment on part of the collaborators. In order to quantitatively assess these problems two studies were performed with data material from the research project "Perinatology" in the German Democratic Republic. This project investigated questions of epidemiology of newborns at risk in the years 1970 to 1972 in eight medical institutions with a total of 6780 cases. 1) By comparing the original hospital records of 310 cases the errors for 33 criteria in the computer printout were determined. 2) For 7 subjective clinical criteria in the newborn the difference in opinions was demonstrated by determining the frequency of abnormal findings in the various institutions.