Objective: To develop a risk scorecard for prediction of failure in preterm labor inhibition by selecting the most significant risk factors.

Materials and Methods: Retrospective cross-sectional analytic study collected 200 medical records of pregnant women between 24⁰⸍⁷ and 33⁶⸍⁷ weeks of gestation with the diagnosis of preterm labor who received tocolytic drugs at Chonburi Hospital. The 200 data were split into train and test datasets. Seventy percent (140) were allocated for training the model. The remaining 30% (60) were for testing the model’s accuracy. The present study analyzed risk factors related to failure of preterm labor inhibition and selected those with most predictive power. Finally, a risk scorecard was developed using logistic regression.

Results: The significant risk factors were age, previous preterm labor, cervical dilatation, cervical effacement, duration, and interval of uterine contraction. Models were created and tested for accuracy of prediction of failure in preterm labor inhibition. The final model achieved 90% of AUC, 89% of sensitivity, and 78% of specificity in the test dataset. These four risk factors, which are history of previous preterm labor, cervical dilatation, duration, and interval of uterine contraction, were included in the final scorecard, and ranged from 0 to100. The score of 0 means high risk and 100 means low risk of failure in preterm labor inhibition.

Conclusion: The authors developed a scorecard that can accurately classify preterm labor patients into two groups, failed and successful. The model was tested and proved that it has good AUC, sensitivity, and specificity statistics.

Keyword: Inhibition of preterm labor, Risk score, Tocolytic drugs

DOI: doi.org/10.35755/jmedassocthai.2020.08.9420

Received 16 Oct 2018 | Revised 7 Jan 2019 | Accepted 8 Jan 2019