PubMed1991
This study employs two risk-adjustment strategies to model the cost-effectiveness of obstetric services for eight hospitals in an urban health maintenance organization. Costs are adjusted by an index based on the expected length of a mother's stay, derived from a two stage regression analysis. Logistic regression of the probability of a cesarean-section on a set of clinical indicators constitutes the first stage. The second stage, an ordinary least squares regression, accounts for 30% of the variation in the logarithm of hours of stay but generates unbiased estimates for various subsets of cases. Adjusted costs per delivery range from roughly 22% below to 31% above the mean. Perinatal mortality rates--adjusted for differences in birthweight, sex, plurality, and race--serve as the outcome indicators. Risk-adjusted costs and risk-adjusted mortality rates are positively correlated with one another (r = .69, P = .06); in particular, the lowest cost hospital generated excellent outcomes. Adjusted cesarean-section rates, however, are not correlated with either adjusted costs (r = -.03, P = .95) or adjusted perinatal mortality rates (r = -.13, P = .75). These results suggest that cost management should focus on staff levels and mix more than on practice patterns and that care management should focus on practice patterns in relation to their influences on outcomes.