The Impact of Hemodynamic and Hematologic Markers on Postpartum Hemorrhage Risk in Hypertensive Pregnancies
Volkan Özgür Akbulut1, Can Ozan Ulusoy1, Aziz Kından1, Ahmet Kurt2, Ayse Gizem Yildiz3, Gözde Başpınar Doğru3, Hasan Altınsoy4, Yaprak Engin-Ustun31Ankara Etlik City Hospital, Department of Perinatology, Ankara/TURKEY2Ankara Etlik City Hospital, Department of Obstetrics and Gynecology, Ankara /TURKEY
3Ankara Etlik Zubeyde Hanim Women’s Health Training and Research Hospital, Department of Obstetrics and Gynecology, Ankara /TURKEY
4Ankara Etlik Zubeyde Hanim Women’s Health Training and Research Hospital, Department of Perinatology, Ankara /TURKEY
INTRODUCTION: This study aims to evaluate the predictive capabilities of traditional statistical and machine learning approaches in identifying postpartum hemorrhage (PPH) risk among hypertensive pregnancies, focusing on indicators such as hemoglobin, mean arterial pressure (MAP), and pulse rate. This will enable us to identify accessible and reliable predictors of PPH, thereby contributing to the early clinical decision-making process.
METHODS: Between 2016-2020, this retrospective case-control study was carried out at a tertiary center. Total of 64 cases with hypertensive disorders of pregnancy (HDP) were included, with 32 developing PPH and 32 serving as controls. To evaluate predictive value, hemoglobin, MAP, and pulse rate were analyzed using logistic regression and random forest models. Hyperparameter tuning was performed using GridSearchCV with 5-fold cross-validation. Model performance was evaluated using sensitivity, specificity, F1-score, accuracy, and area under receiver operating characteristic curve (AUC-ROC).
RESULTS: Hemoglobin levels were significantly lower (9.7 g/dL vs.11.9 g/dL, p=0.002), while MAP (103.0 mmHg vs.96.7 mmHg,p=0.002) and pulse rate (90.0 bpm vs.86.0 bpm,p=0.006) were significantly higher in the PPH group. Logistic regression demonstrated higher sensitivity (0.55 vs.0.44) and positive predictive value (0.83 vs.0.80) compared to the random forest model. However, the random forest model had a lower shrinkage value (-0.001 vs. -0.116), suggesting better generalizability. MAP ≥109.4 mmHg, pulse rate ≥97.6 bpm and hemoglobin level <9.23 g/dL were associated with a 94% probability of developing PPH.
DISCUSSION AND CONCLUSION: The findings indicate that monitoring hemoglobin, MAP and pulse rate may aid in recognizing PPH risk among HDP highlighting the importance of early risk assessment for enhancing maternal prognosis.
Keywords: Postpartum hemorrhage, hypertensive disorders of pregnancy, mean arterial pressure, machine learning, logistic regression, random forest.
Manuscript Language: English