Understanding the effects of technological interventions and using evidence to inform decisions are difficult tasks. Uncertainty remains when clinical studies provide evidence for only surrogate outcomes, have small sample sizes, have limited follow-up durations, have deficiencies in their design, or present insufficient information on relevant patient subgroups. Models and simulations are valuable tools for addressing the uncertainty, trade-offs, and heterogeneous preferences that complicate research questions in health technology assessment.
[Photo: Dr. Issa Dahabreh]
The purpose of this paper, led by Dr. Issa Dahabreh, assistant professor of health services, policy and practice, and member of the Center for Evidence-based Medicine, was to provide recommendations for the conduct and reporting of modeling and simulation studies based on a systematic review of the published literature, a survey of websites of international health technology assessment organizations, and input from experts and other stakeholders.
The recommendations apply to mathematical models that represent relationships among model components and integrate information from multiple sources; they address model identification, estimation, verification, and validation, as well as the conduct of sensitivity, stability, and uncertainty analyses. They are organized into model conceptualization and structure, data, model assessment and consistency, and interpreting and reporting results. For example, when conceptualizing and structuring a model, the scope should be consistent with the research question and modeling perspective. Sources of bias in the model data should be assessed and accounted for when estimating values for model parameters. When assessing the model, it is critical to assess the consistency between model outputs and the data on which the model is based. And finally, it is important to interpret and report results in a way that communicates uncertainty in model outputs.
Investigators tasked with evaluating health technologies face complex questions when assessing the effects of interventions and informing clinical and policy decisions. Addressing such questions requires the integration of evidence from multiple sources using models and simulations. The recommendations presented in this paper can contribute to increased use and better conduct of reporting and modeling and simulation studies in health technology assessment.
This study was published in Annals of Internal Medicine, Volume 165, Issue 8.
To read more: https://www.ncbi.nlm.nih.gov/pubmed/27750326