Chi-square and logistic stepwise multiple regression analysis of perinatal determinants of infant bacterial infection following prolonged rupture of amniotic membranes for 24 hours or more prior to delivery was applied in 33 infected infants and 66 matched control infants from the NINCDS Collaborative Project. In order of statistical significance, the most important variables were placental inflammation (P = 0.002), gestational age less than 34 weeks (P = 0.008), gestational age 34 to 37 weeks (P = 0.013), male sex (P = 0.015), Apgar score less than 6 at 5 minutes (P = 0.023), and clinical amnionitis (maternal fever, fetal tachycardia, or amniotic or gastric fluid leukocytes or bacteria) (P = 0.044). Duration of labor during PROM, race, and maternal age and parity were insignificant. Using these predictive variables, identification of infected infants for either microbial surveillance (superficial and systemic cultures) or microbial surveillance and anticipatory antibiotic therapy (discontinued after 3 days of negative cultures) was highly significant (P = 0.0001). Incorporating these variables and derived coefficients from multivariate analysis, a mathematical model was used for evaluation and prediction of perinatal bacterial infection with a sensitivity of 82% and specificity of 70%. Analysis of 46 infants prior to and 310 infants after implementation of this process at Harbor-UCLA Medical Center indicated significant improvement in the appropriate management of these infants at risk (from 59% to 87% of the population, P less than 0.05). Inappropriate antibiotic therapy decreased from 35% to 10% (P less than 0.05). In the absence of a shift in the median days of hospitalization of non-PROM infants, determination of the grand median days of PROM infant hospital stay showed a decrease (P less than 0.01) after initiation of this evaluation and management scheme.