Quantitative Modeling Software with Applications to Medical Decision Making

Project Summary/Abstract In recent years, health care systems and physicians have made concerted efforts to practice evidence-based medicine and provide patients with the best available information when making choices about their medical care. However, medical decisions are often complex with many uncertainties and potential outcomes to consider, some beneficial and some adverse. A popular analytic method used to help identify best treatment strategies while accounting for uncertainty is decision analysis, which typically involves computer modeling of a treatment choice outlined in the form of a decision tree, which shows options and health outcomes that may occur as a result of the choice made. Complex decision trees are evaluated via Monte Carlo microsimulation to allow for variability in individual patient characteristics and trace a patient’s path through the tree; when the microsimulation is repeated many times to simulate many individuals, it provides the probability of each potential outcome resulting from the initial decision. From this probability distribution, quantitative measures associated with each decision can be calculated such as life years, quality-adjusted life years (a generic measure of disease burden), and others; furthermore, when costs are also incorporated, cost-effectiveness analysis (CEA) can be performed to compute the incremental cost-effectiveness of each option. In this proposal, we describe plans to add functionality to the mathematical modeling software Berkeley Madonna to allow users to build decision trees and carry out Monte Carlo microsimulations and Markov cohort analysis. Berkeley Madonna’s interface was designed to make mathematical modeling quick and easy for non-technical users by using a simple syntax and graphical images to construct sophisticated differential equations. We will leverage this easy-to-use interface to enable medical researchers to perform microsimulation with software that is more user-friendly, transparent, powerful, and affordable than currently available options. In Aim 1, we propose further development of our decision analysis user interface that allows users to graphically construct decision trees and perform microsimulations. In this aim, in addition to optimizing tools and features for the GUI, we will add CEA output reports and graphics, sensitivity analysis capabilities, and Markov cohort analysis capabilities. We will create tutorials and a user guide as well as ready-made templates that provide users a jumping off point for quickly making their own models. In Aim 2, we propose to optimize code for performance on single CPUs, multiple CPUs, and GPUs. Analysis speed is important because large, complex models can take weeks to months to run with currently available software, none of which harness the power of GPU technology; successful completion of this aim would make Berkeley Madonna the fastest available software by far for performing decision analysis microsimulations. Finally, we will carry out extensive beta testing. Achievement of these goals will provide an easy-to-use, transparent, powerful, and affordable tool to biomedical researchers, educators, and professionals, and positively impact scientific discovery.