Aim: Several studies have shown that preterm delivery, a primary cause of perinatal mortality and morbidity, is more frequent in HIV-positive women. This study aimed to determine factors associated with prematurity in HIV-infected women and identify risks for which specific interventions could be targeted. Methods: Data were prospectively collected in a clinical trial assessing the efficacy of different zidovudine prophylaxis durations for the prevention of perinatal HIV transmission in Thailand. Characteristics associated with prematurity - delivery before 37 weeks - were assessed using univariate and multivariate logistic regression and were subsequently used to identify subgroups of women at risk. Results: Among 979 women, independent prematurity risk factors were: viral load <3.5 or >4.5 log copies/mL; hemoglobin >11.5 g/dL; weight gain <0.25 kg/week; and body mass index <20 kg/m2. These factors allowed us to define four subgroups with an expected probability of prematurity increasing from 3% to 30%. The two subgroups with the highest expected probability of prematurity were considered to be 'at risk' as opposed to the two lowest (odds ratio = 2.6, 95% confidence interval: 1.7-4.0) and the sensitivity and specificity of the prediction were 51% and 71%, respectively. Conclusion: In this study, four risk factors of preterm delivery were identified allowing the identification of subgroups at increasing risk of prematurity. Adequate nutrition and the provision of highly active antiretroviral therapy during pregnancy as recommended by the World Health Organization for the prevention of perinatal transmission for immunocompromised women in resource-constrained countries may reduce the risk of premature delivery. © 2008 The Authors.