Bias on withdrawing lost subjects from the analysis at the time of loss, in cohort mortality studies, and in follow-up methods.
Cohort mortality studies may differ from morbidity and other follow-up studies in that, in many of the methods of follow-up employed, the identification or follow-up process and the process of determination of outcome (death) are essentially the same. Furthermore, some of the methods may involve preferential identification of deaths over live persons, or vice versa. This latter area of difference is particularly problematic in situations in which the comparison group (the general population) has not been subjected to tracing, and in which certain persons who are being traced are not covered by the universe defined by the follow-up method. Because of these peculiarities, it is shown that, during the analysis, the usual practice of withdrawing persons lost to follow-up at the time of loss can lead to seriously biased results when follow-up rate is not very high.