Quality assessment of perinatal regionalization by multivariate analysis: Illinois, 1991-1993.

OBJECTIVE: To identify (1) those elements in the infrastructure of a regionalized perinatal network that have independent effects on the variation in perinatal mortality among nontertiary units (member level I and II hospitals) and (2) shortcomings, if any, in a traditional perinatal data base that impede quality assessment of contemporary regionalized care. METHODS: We analyzed perinatal surveillance data for 3 years, from 1991 to 1993, in the state of Illinois, representing more than 190,000 annual births. Fetal death and neonatal mortality rates for the 97 nontertiary hospitals studied were the dependent variables of interest. Two sets of independent variables were studied, those assessing the maternal sociobehavioral risk of populations served and those assessing the network infrastructure (defined as the facilities of member hospitals and their function within the regionalized network). We used multivariate analysis to partition the variation in hospital rates of perinatal mortality into two components, one attributable to maternal sociobehavioral risk and the other to the network infrastructure. RESULTS: Maternal sociobehavioral risk alone explained 73% of the variation in hospital fetal death rates and 38% of that in hospital neonatal mortality rates. When controlling for maternal sociobehavioral risk, rates of inborn very low birth weight (VLBW) deliveries (P < .001) and neonatal transport (P = .01) had independent effects on the variation in hospital fetal death rate; rates of inborn VLBW deliveries (P < .001), neonatal transport (P < .001), and proportion of VLBW infants transported out (P = .029) had independent effects on the variation in hospital neonatal mortality rate. CONCLUSIONS: In this mature statewide network, the rate of inborn VLBW deliveries exerted the strongest independent effect on variation in level I and II hospital rates of both fetal death and neonatal mortality. However, that there was such a large effect from maternal sociobehavioral risk alone has important public health implications. Additions and modifications to traditional perinatal surveillance are suggested better to assess the quality of regionalization in a contemporary health care environment.