PS 1-58 DTREE: AN OPEN SOURCE TOOL FOR BUILDING DECISION TREES AND COST-EFFECTIVENESS ANALYSES

Sunday, October 23, 2016
Bayshore Ballroom ABC, Lobby Level (Westin Bayshore Vancouver)
Poster Board # PS 1-58

Hawre Jalal, MD, PhD, University of Pittsburgh, Pittsburgh, PA, Anne Cross, Pittsburgh, PA, Fernando Alarid-Escudero, MS, PhD Candidate, Division of Health Policy and Management, University of Minnesota, Minneapolis, MN and M.G. Myriam Hunink, MD, PhD, Center for Health Decision Science, Harvard T.H.Chan School of Public Health, Boston, MA

Purpose: Developing simulation models in high level programming language is gaining increased popularity for the advantages of accuracy, speed, transparency, troubleshooting and performing advanced computations, including calibration and value of information.  However, a major limitation of adopting a programming language is the lack of graphical user interface for visualizing and constructing decision trees and mathematical models.  The purpose of this paper is to introduce DTree, an open source tool that we created for constructing decision trees and converting them into programming code in languages such as R.

Method: The figure illustrates a snapshot of a simple Markov model built with DTree.  DTree is written in Javascript.  Users can build decision trees interactively, by adding, editing or deleting nodes.  In addition, they can change the node type, add properties, payoffs and probabilities to each node.  Because the tool is entirely written in Javascript, it only requires a modern browser, such as Google Chrome or Mozilla Firefox.  Internet connection is not needed.  Users can build new trees, save their progress, load built trees, and edit existing ones. DTree, then uses an intermediary file as a JSON object, which can then be converted to programming code that can be used to run computations. 

Results: Currently the tool is able to create and modify decision trees and Markov models, and translate these models to R code which can then be used independent of the graphical representation of the tree.  The generated R code can then be run to produce the model outcome. 

Conclusion: DTree is an open source effort that is continuously being improving.  It is freely available and open for the community to use and improve. In addition, the tool can be easily extended to other programming languages, such as C++ or MATLAB. We hope that this tool will address the lack of graphical interface for building decision trees in high level programs and help decision analysis to be accessible to a wider range of interested trainees and researchers.

See more of: POSTER SESSION I & WELCOME RECEPTION

See more of: 38th Annual North American Meeting of the Society for Medical Decision Making