Decision-Theoretic Analysis of Prenatal Testing Strategies

We have constructed a Bayesian influence diagram and implemented an analytic framework to recommend strategies for selecting prenatal diagnostic tests and deciding whether to terminate a pregnancy based on their results. Our model is an improvement over previous applications of decision analysis to prenatal testing because it includes multiple diseases (Down’s syndrome, neural-tube defects, cystic fibrosis, sickle-cell anemia, and fragile X mental retardation) and multiple diagnostic tests (chorionic villus sampling, triple marker screening, ultrasonography, and amniocentesis). Our software framework handles influence diagrams with multiple decisions and the conditional plans (decision strategies) that such models imply. Our software allows us not only to generate optimal decision strategies for specific cases, but also to examine decision strategies that we suggest for expected utility and sensitivity to other variables in the decision model.