Abstract Objective To develop a model for predicting premature delivery before 37 weeks’ gestation based on maternal factors, obstetric history and biomarkers in the first trimester of pregnancy. Study design Cohort study based on data collected prospectively between 1 January 2000 and 30 November 2011. Multivariate logistic regression was used to construct a model of the risk of premature delivery. Results 31,834 pregnancies were included, of which 1188 cases were spontaneous premature deliveries before 37 weeks (3.7%). We built a predictive model based on maternal age, body mass index, smoking status and previous obstetric history. This could identify 23.3% of premature deliveries in our study population, with a false positive rate of 10%. In the group of patients who had already had at least one pregnancy at or beyond 16 weeks, the detection level increased to 29.7%. The positive predictive value was 7.4 and 7.3% respectively, while negative predictive value was 97.2 and 97.9%. Conclusions Predicting preterm delivery on the basis of maternal characteristics and obstetric history needs to be further improved. PAPP-A levels and ultrasonographic measurement of cervical length could not be integrated in the model but require further investigations.