Prediction of preterm birth at St. Mary’s Hospital Lacor, Northern Uganda: a prospective cohort study

Abstract

Background: Preterm birth causes over 2% of perinatal mortality in Africa. Screening in prenatal clinics, may be used to identify women at risk. This study developed and validated second-trimester prediction models of preterm birth, using maternal socio-demographic characteristics, sonographic findings, and laboratory parameters in Northern Uganda. Methods: This prospective cohort study recruited 1,000 pregnant mothers at 16 - 24 weeks, and assessed their socio-demographic and clinical characteristics. Preterm birth (delivery after 28 and before 37 weeks) was the primary study outcome. Multi-variable analyses were performed, built models in RStudio, and cross-vaidated them using K (10)-fold cross-validation. Results: The Incidence of preterm birth was 11.9% (90 out of 774). The predictors of preterm birth were multiple pregnancies, personal history of preeclampsia, history of previous preterm birth, diastolic hypertension, serum ALP<98IU, white blood cell count >11000 cells/μl, platelet lymphocyte ratio >71.38, serum urea of 11-45 IU. These predicted preterm birth by 69.5% AUC, with 62.4% accuracy, 77.2% sensitivity, and 47.1% specificity. Conclusion: Despite low specificity, these models predict up to 77.2% of those destined to have a preterm birth, and may be used for second-trimester preterm birth screening in low-resource clinics. Keywords: Prediction; second-trimester; preterm-birth; Uganda; Africa.