Maternofetal infection occurs in 1 to 10 of every 1,000 newborns. Prognosis is poor and an extremely rapid the clinical course is sometimes observed. The situation suggests that widespread use of antibiotics and more than 90% of the newborns receiving antibiotics are not infected. A prospective study based on simple, recognized criteria predictive of maternal-fetal infection was conducted in 3,392 deliveries to evaluate the effect of an antibiotic prophylaxy protocol. Specificity, sensitivity, positive and negative predictive value were evaluated for each of the eleven criteria retained with the goal of increasing sensitivity and decrease the use of unnecessary antibiotics. Among the 3,392 infants delivered from January 1989 to December 1990, 286 (8.4%) newborns entered the study and were given mezlocillin (150 mg/kg/12 h). This treatment was stopped at 48 hours of life if the infant was not infected. RESULTS. Infection was confirmed in 48 of 3,392 infants (1.4%). All were in the risk group: 48/286 (16.7%). The germs the most often found were group B Streptococcus (n = 16), Escherichia coli (n = 8) and Listeria monocytogenes (n = 3). Nine criteria were well correlated with maternal-fetal infection. The two most important criteria were maternal pyrexia above 39 degrees C and Apgar score below 7 to 5 min (poor neonatal adaptation), with a 99% and 90% and a positive predictive value of 80% and 37% respectively. Inversely, Two criteria were poorly correlated with maternal-fetal infection: labour duration above 12 hours and instrumental extraction (positive predictive value from 10% to 17%).(ABSTRACT TRUNCATED AT 250 WORDS)