Objectives To compare the predictive value of the shock index (SI) with conventional vital signs in postpartum haemorrhage (PPH), and to establish 'alert' thresholds for use in low-resource settings.Design Retrospective cohort study.Setting UK tertiary centre.Population Women with PPH ≥1500 ml (n = 233).Methods Systolic blood pressure (BP), diastolic BP, mean arterial pressure, pulse pressure, heart rate (HR) and SI (HR/systolic BP) were measured within the first hour following PPH. Values measured at the time of highest SI were selected for analysis. The area under the receiver operating characteristic curve (AUROC) for each parameter, used to predict admission to an intensive care unit and other adverse outcomes, was calculated. Sensitivity, specificity and negative/positive predictive values determined thresholds of the best predictor.Main outcome measures Intensive care unit (ICU) admission, blood transfusion ≥4 iu, haemoglobin level <7 g/dl, and invasive surgical procedures.Results Shock index has the highest AUROC to predict ICU admissions (0.75 for SI [95% CI 0.63-0.87] compared with 0.64 [95% CI 0.44-0.83] for systolic BP). SI compared favourably for other outcomes: SI ≥0.9 had 100% sensitivity (95% CI 73.5-100) and 43.4% specificity (95% CI 36.8-50.3), and SI ≥1.7 had 25.0% sensitivity (95% CI 5.5-57.2) and 97.7% specificity 3), for predicting ICU admission.Conclusions Shock index compared favourably with conventional vital signs in predicting ICU admission and other outcomes in PPH, even after adjusting for confounding; SI <0.9 provides reassurance, whereas SI ≥1.7 indicates a need for urgent attention. In low-resource settings this simple parameter could improve outcomes. It was not possible to adjust for resuscitative measures administered following vital sign measurement that may have influenced the outcome.