Making Health Economic Models More Efficient in Dealing with Contemporary Complexities
Author(s)
Poirrier JE1, Maervoet J2, Bergemann R3
1Parexel, HEOR Modeling, Wavre, WBR, Belgium, 2Parexel, HEOR Modeling, Schoten, VAN, Belgium, 3Parexel International, Loerrach, Germany
Presentation Documents
OBJECTIVES:
Health economic models (HEM) are widely used to assess the cost-effectiveness of healthcare interventions and inform decision-making by policymakers and payers. However, their complexity is increasing, as they are being used to analyze more diverse and detailed data sets, more complex interventions and patient pathways. This has led to the development of increasingly more complex analysis frameworks (like mixed-cure or individual-based models) and to an increasing uncertainty tentatively addressed by extensive sensitivity analyses. This poster brings conceptual solutions to these computational issues and illustrates them with examples in Excel.METHODS:
A first solution is to understand and review the model concepts. Often modeling frameworks have requirements and inefficiencies (e.g. keeping a Markov trace) that can lead to computational issues (e.g. out-of-memory execution, inefficient loops). Understanding the modeling framework allows to focus computation on effective parts. A second solution is to understand the implemented algorithm, the language, its intricacies, environment and limitations. For instance, Visual Basic for Applications (VBA) code inherits from Excel user-interface features that slow down data processing and can be removed. VBA code is weakly typed and its type-determination is sometimes problematic, with issues arising “far” from where the initial code is. Once the framework is appropriate and the code optimized, a third solution is to increase the computing power available for the model. Even without code parallelization, the variety of powerful processors available in commercial cloud offerings makes it easy and cheap for HEMs to run much faster. Our benchmarks show 2-10 times decrease in computing times depending on the task.RESULTS: and CONCLUSIONS
: Modelers have an array of solutions at their disposal to reduce the computing time required by HEMs. This freed time can therefore be used to explore more modeling possibilities and/or to reflect more on results. Standardization of HEM programming practices will greatly help modelers’ efficiencies.Conference/Value in Health Info
2023-05, ISPOR 2023, Boston, MA, USA
Value in Health, Volume 26, Issue 6, S2 (June 2023)
Code
MSR85
Topic
Economic Evaluation, Methodological & Statistical Research, Study Approaches
Topic Subcategory
Cost-comparison, Effectiveness, Utility, Benefit Analysis, Decision Modeling & Simulation
Disease
No Additional Disease & Conditions/Specialized Treatment Areas