Related Experiment Videos
Evaluating obstetric risk scores by receiver operating characteristic curves
A A Herman1, L M Irwig, H T Groeneveld
1Gertrude H. Sergievsky Center, Columbia U., New York, NY 10032.
American Journal of Epidemiology
|April 1, 1988
Summary
Obstetric risk scoring systems identify high-risk pregnancies. Comparing systems using receiver operating characteristic curves (theta) reveals that those including past reproductive history and late antenatal/intrapartum factors offer higher predictive accuracy for poor perinatal outcomes.
Area of Science:
- Obstetrics and Gynecology
- Perinatal Medicine
- Health Informatics
Background:
- Obstetric risk scoring systems aim to predict adverse perinatal outcomes, such as perinatal death.
- Current systems vary in their threshold criteria, complicating direct comparisons of sensitivity and specificity.
- Standardized methods are needed to evaluate and rank the predictive accuracy of different obstetric risk scoring systems.
Purpose of the Study:
- To compare the performance of various obstetric risk scoring systems.
- To establish a method for ranking obstetric risk scoring systems based on predictive accuracy.
- To identify key factors contributing to accurate prediction of poor perinatal outcomes.
Main Methods:
- Utilized receiver operating characteristic (ROC) curves to analyze multiple risk scoring systems.
- Calculated the area under the ROC curve (theta) as a metric for comparing system performance across different thresholds.
- Reviewed published risk scoring systems incorporating prepregnancy, early antenatal, late antenatal, and intrapartum data.
Main Results:
- The area under the ROC curve (theta) values for the reviewed systems ranged from 0.49 to 0.95.
- Risk scores incorporating past reproductive experience and late antenatal/intrapartum factors demonstrated higher predictive accuracy.
- Systems employing statistical weighting of risk factors tended to rely less on late prenatal and intrapartum data.
Conclusions:
- Receiver operating characteristic curve analysis provides a robust method for comparing obstetric risk scoring systems.
- Predictive accuracy is enhanced by including comprehensive data, particularly past reproductive history and intrapartum information.
- Further research should focus on developing and validating risk scoring systems that integrate diverse data sources for improved perinatal outcome prediction.