The importance of unknows in Epidemiologic studies

1. Epidemiologic study data often include omitted/unobtainable responses (unknowns). In most cases, unknowns are eliminated during data-reduction to facilitate analysis. We examined the effect that elimination of unknowns would have on mortality calculations using data on newborns admitted to a newborn intensive care unit (NICU). 2. Data were divided into unknown (omitted/unobtainable), abnormal and normal response groups. Forty-three factors with at least four unknown responses recorded during 1984 were selected for analysis. NICU mortality (number of NICU deaths per 100 NICU admissions) was selected as the outcome indicator. In order to quantify the effects of eliminating unknowns, NICU mortality rate ratios (NMRR) were calculated, by factor, for abnormals, normals and knowns (abnormals and normals) both without and with knowns.