(Taken from application abstract): This proposed study will replicate and extend methodology used in earlier studies and will use extensive clinical data repositories, informatics tools, and expert practitioners for perinatal medical knowledge building. Clinical Data Repository: Duke University's Medical Center (DUMC) TMR (The Medical Record) data repository will be used for this study, and contains 45,922 electronic medical records for both low and high-risk pregnant women (and their infants) who have received prenatal care at DUMC, and its affiliated regional clinics, between 1/1/86 and 12/3l/95. Each patient's electronic data is used for clinical patient care and contains a potential 4000 variables per record. This volume of data requires new approaches for data analysis and medical decision support, since human information processing limitations become quickly overloaded by both an individual patient s data and the aggregate information collected for the perinatal patient population. lnformatics Tools: Informatics techniques for knowledge acquisition and data mining will use machine learning programs, statistical analysis, and domain expert input to articulate relationships between the data and perinatal patent outcomes. The goal is to provide decision support for perinatal care providers to accurately identify patients at risk and assist them with modifiable preterm birth ask factors. An expert system will use data-generated and verified knowledge bases to test its predictive validity when new patient cases are induced to the expert system. Earlier studies found 53-90% predictive accuracies for an expert system prototype, as compared to 17-38% accuracies, reported in the literature, using current manual techniques. Mapping the expert system's knowledge base terms to medical library resources will be explored for additional decision support. Expert Practitioner: The perinatal expert panel will consist of the Principal Investigator, a Board Certified OB-Gyn Physician, and a certified Perinatal RN. Each of the panel members has more than 20 years of perinatal experience. Participating informatics experts are known, both nationally and internationally for their expertise in the field of Medical Informatics.
| Goodwin, Linda K; Prather, Jonathan C (2002) Protecting patient privacy in clinical data mining. J Healthc Inf Manag 16:62-7 |
| Goodwin, Linda K; Iannacchione, Mary Ann (2002) Data mining methods for improving birth outcomes prediction. Outcomes Manag 6:80-5 |
| Goodwin, L K; Iannacchione, M A; Hammond, W E et al. (2001) Data mining methods find demographic predictors of preterm birth. Nurs Res 50:340-5 |
| Goodwin, L; Maher, S; Ohno-Machado, L et al. (2000) Building knowledge in a complex preterm birth problem domain. Proc AMIA Symp :305-9 |
