A Bayesian decision support model for assessment of endodontic treatment outcome.

OBJECTIVE This article presents a decision support model that describes the mutual relationships among multiple variables for assessment of the outcome of endodontic treatment. STUDY DESIGN The model was built on the data-driven Bayesian network (BN) methodology. Randomized controlled trials of nonsurgical endodontic treatment from January 1966 through August 2007 were chosen to be our data source. The total sample size in the included studies was 8783 cases. The structure and conditional probability distributions of the BN were learned from the data using Necessary Path Condition algorithm and Expectation-Maximization learning algorithm respectively. RESULTS Receiver operating characteristic curve analysis showed that the model was highly accurate in predicting the endodontic treatment outcome; the area under the curve (AUC) was 0.902. The predictions generated by the model are in line with majority consensus predictions of the endodontists. In the cases where the endodontists' predictions were uncertainty, the proposed model predicted them more accurately. CONCLUSION A decision support model can be constructed from clinical trails to successfully predict endodontic treatment outcome.