Computation of Indirect-Adjusted Rates in the Presence of Confounding

SUMMARY Mortality and morbidity rates are ordinarily adjusted for effects of age, race, or sex by either the direct or the indirect method. For the direct method one usually employs an external known standard population while for the indirect method one usually employs an external known set of standard rates. Using data on the occurrence of mongolism in children, the authors present a methodology for indirect adjustment in which the set of standard rates are derived internally from the data. Specifically, the authors were concerned with obtaining rates of mongolism for varying birth order of child with adjustment for age of mother and, conversely, rates for varying maternal age with adjustment for birth order. What is required are two sets of adjusted rates, one for birth order and one for maternal age, with a mutual relationship such that each is the set of standard rates on which the other set of adjusted rates is based. These two sets are obtained by a convergent iterative procedure. Without the requirement for mutuality, adjusted rates can be anomalous, particularly when the study variables, like birth order and maternal age, are interrelated. Thus when crude birth-order rates were used as standard rates in obtaining adjusted maternal-age rates, those adjusted rates fell outside the range of the supporting specific rates.