An Integrated R-Based Framework for Model Development in Health Economics
Daniel Kaschek (1), Henning Schmidt (1), Anna Kondic (2)
(1) IntiQuan, Basel, Switzerland, (2) Economic and Data Sciences, MSD, Rahway, NJ, USA
Introduction: Health economic decisions rely on the evaluation of complex systems. The complexity demonstrates itself in several aspects: (1) each patient might be affected by a series of different events that render possible future consequences more or less probable, (2) the target population for a therapy may be quite different from the overall population and, (3) several alternative therapies might be available that have specific advantages and disadvantages, both in terms of cost and utility. It is the complexity and multitude of possibilities that makes modelling and simulation an indispensable tool for decision making in health economics.
To date, several independent software tools have been used to generate such models. However, models cannot be easily exchanged and exported into a readable, self-explanatory representation. Thus, proving the validity of a model to decision-makers constitutes a major difficulty.
Objectives: To develop a software framework for health economic modelling with a focus on Discretely Integrated Conditions and Events (DICE) models [1] and Markov models [2]. The major features of the core framework are model set-up by a newly developed R-based modelling language, simulation and analysis as well as interoperability with existing modelling tools via a standardized interchange format. The overall goal of the new modelling language is to render model formulation more concise and, therefore, easier to understand for external persons.
Methods: The DICE approach is a recently developed formalism to set up health-economic models for single patients or cohorts of patients. Models consist of “conditions”, describing all relevant aspects of a patient’s state, and “events”, causing sudden changes of conditions and triggering consequent events. In Markov models, events correspond to transitions which occur in every time step. Conditions in DICE models can be mapped to Markov states. However, depending on the multitude of independent patient conditions, the number of representative Markov states can excessively increase due to the combinatorial explosion. The DICE concept covers Markov models as a special case. It is, therefore, especially suited to building the core of a general modelling framework that would allow to set up, simulate and analyze a variety of health-economic models.
Results: Based on a selection of published Markov and DICE models [3, 4], we have developed a concept core modelling framework acting on three different levels. The central level is an R based Health Economic Modelling (RHEM) language by which both, DICE and Markov models can be represented. The RHEM language is designed to be expressive and concise allowing the formulation of models in few lines of humanly readable code with inline model documentation.
Underneath the RHEM language, we have implemented a simulation engine that uses the DICE algorithm to simulate DICE and Markov models. Markov models are internally converted to fit the DICE specification. Various utility functions have been implemented to visualize simulation and analysis results obtained from the simulation of our RHEM models.
To connect to already existing platforms like heRo3 [5] or EviDICE [6], we have developed a standardized JSON-based exchange format called RHEMJ that enabled us to import the published models from those platforms. The RHEMJ specification allows for model-related content like conditions, events, states, transitions to be stored. In addition, data, analysis and simulation results can be accommodated in RHEMJ.
Conclusion: Health-economic modelling has a major impact on the estimated value of a drug. By the conception of an R based Health Economic Modelling framework we set out to simultaneously establish a standard interchange format and a reference implementation to import or set up those models and simulate them. The new RHEM modelling language employed within the framework maximizes readability and intuitive understanding of complex models creating trust in the validity of the model.
References:
[1] Caro JJ, Möller J, Getsios D. Discrete event simulation: the preferred technique for health economic evaluations?. Value in health. 2010 Dec 1;13(8):1056-60.
[2] Beck JR, Pauker SG. The Markov process in medical prognosis. Medical decision making. 1983 Dec;3(4):419-58.
[3] Gaziano TA, Fonarow GC, Claggett B, Chan WW, Deschaseaux-Voinet C, Turner SJ, Rouleau JL, Zile MR, McMurray JJ, Solomon SD. Cost-effectiveness analysis of sacubitril/valsartan vs enalapril in patients with heart failure and reduced ejection fraction. JAMA cardiology. 2016 Sep 1;1(6):666-72.
[4] Möller J, Davis S, Stevenson M, Caro JJ. Validation of a DICE Simulation Against a Discrete Event Simulation Implemented Entirely in Code. PharmacoEconomics. 2017 Oct 1;35(10):1103-9.
[5] heRo3, http://hero3models.info/ [6] EviDICE, “An Efficient Approach to DICE”, http://www.evidera.com/dice/