A risk scoring system designed to predict spontaneous preterm birth was implemented in a large, indigent population as part of a multicenter trial of preterm birth prevention. A total of 7478 women with singleton gestations were screened and followed up prospectively at the Birmingham project center. Patients who had an indicated preterm delivery or a fetal anomaly were excluded from the study population. Analysis by assigned risk score and parity showed that, whereas the sensitivity and positive predictive value were better in multiparous women than in nulliparous women, overall the values were low. Logistic regression analyses of the multiparous and nulliparous populations showed independent sets of significant (p <- 0.05) risk variables. A history of preterm delivery and a low prepregnancy weight were the most predictive risk factors in the multiparous and nulliparous models, respectively. We conclude that the clinical usefulness of a risk scoring system to predict spontaneous preterm birth in an indigent population is limited. © 1990.