Related Experiment Videos
Data mining methods find demographic predictors of preterm birth.
L K Goodwin1, M A Iannacchione, W E Hammond
1Health Systems and Primary Care, and School of Nursing and Community and Family Health Medicine, Duke University, Durham, NC, USA. goodw010@mc.duke.edu
Nursing Research
|December 1, 2001
Summary
Predicting preterm birth is crucial. Simple demographic factors accurately identify at-risk pregnancies, showing data mining methods confirm traditional findings for preterm birth prediction.
Area of Science:
- Obstetrics and Gynecology
- Public Health
- Biostatistics
Background:
- Preterm birth rates in the U.S. present a persistent and complex healthcare challenge.
- The incidence of preterm births has shown an increasing trend in recent years.
Purpose of the Study:
- To compare the efficacy of traditional statistical methods against emerging data mining techniques.
- To identify accurate predictors for preterm birth using advanced analytical approaches.
Main Methods:
- Utilized an ethnically diverse dataset of 19,970 pregnant women with 1,622 variables.
- Applied both traditional statistical analyses and data mining methods to evaluate preterm birth predictors.
Main Results:
- Seven demographic variables achieved an area under the curve of 0.72 in predictive accuracy using Receiver Operating Characteristic curves.
- Incorporating hundreds of additional variables resulted in only a marginal improvement (0.03) in predictive accuracy.
Conclusions:
- Data mining methods consistently produced similar results, indicating findings are data-driven.
- A concise set of demographic variables demonstrates reasonable accuracy for predicting preterm birth outcomes in diverse populations.