Comparison of two methods for the discrimination of avoidable perinatal deaths.

ABSTRACT

PIP: Two methods for the identification of avoidable perinatal deaths (the first based on a Perinatal Mortality Committee [COMOPer] audit using the Ninth Revision of the International Disease Classification and the second on a computerized list developed by Herman et al.] were compared through use of data from the National Institute of Perinatology in Mexico City, Mexico. While the Herman method entails an a priori determination about the avoidability of deaths by cause, the other is derived from a discussion of individual cases by a committee of experts. Analyzed were the 1337 neonatal deaths occurring at the institution during 1987-94; fetal and infant deaths were excluded because of their lack of fit with the Herman model. Of the 749 neonatal deaths (56%) in which Herman's list could be applied, 173 were classified as avoidable, 541 as possibly avoidable, and 35 as nonavoidable. In contrast, COMOPer designated 230 of these same deaths as avoidable, 514 as nonavoidable, and 5 as undeterminable. Only 38% of neonatal deaths assessed by COMOPer as avoidable and 6% of those judged unavoidable were so classified by the Herman method. This extremely low concordance (Cohen's Kappa, 0.30) indicates the importance of a case-by-case analysis that takes into account the factors underlying the most frequent local pathologies as well as local health policies (e.g., the legality of abortion). Although an a priori classification system can enable the assessment of rough indicators of the quality of medical care at a regional level, a careful analysis of individual deaths by a multidisciplinary group of experts is important to identify deficiencies in the quality of care at the local hospital level.