E-Posters-Predicting Complicated Obstetric Delivery: Defining the Role of the Pelvic Floor

Predicting Complicated Obstetric Delivery: Defining the Role of the Pelvic Floor - Linda S. Burkett, MD

Pdf Summary

This study aimed to analyze a machine learning prediction tool for complicated obstetric delivery, specifically focusing on the role of pelvic floor soft tissue remodeling. Two databases were created, one with over 190,000 deliveries and 31 variables and another with 149 patients and 25 variables. The databases were mined and analyzed using the Waikato Environment for Knowledge Analysis (Weka) suite. Information gain analysis was used for variable selection, and a random forest algorithm was used for prediction modeling. The results showed that the algorithm had a higher accuracy for uncomplicated deliveries compared to complicated ones. Additionally, the inclusion of markers for pelvic floor soft tissue remodeling improved the accuracy of the prediction model. The top features identified by merit included infant's birth weight, maternal age, marital status, and maternal weight. The machine learning model demonstrated superior accuracy with pelvic floor markers compared to traditional sociodemographic and clinical factors. The authors concluded that further research is needed to refine the predictive value of tools for clinical practice, specifically focusing on pelvic floor biomarkers. The analysis is illustrated in a schematic, showcasing the different iterations of the machine learning model.

Keywords

machine learning

prediction tool

obstetric delivery

pelvic floor remodeling

variable selection

random forest algorithm

accuracy

biomarkers

clinical practice

schematic