Bayesian Networks: A Powerful and Flexible Tool for Analyzing Health Economics and Outcomes Research Data

Lawrenceville, NJ, USA—April 25, 2019—Value in Health, the official journal of ISPOR—the professional society for health economics and outcomes research, announced today the publication of a new report introducing Bayesian networks, a knowledge representation and machine-learning tool for risk estimation in medical science. The article discusses several challenges associated with traditional risk prediction methods and then describes Bayesian networks and their construction, application, and advantages in risk prediction based on examples in cancer and heart disease. The paper, “Bayesian Networks for Risk Prediction Using Real-World Data: A Tool for Precision Medicine,” was published in the April 2019 issue of Value in Health.

Bayesian networks provide a robust and flexible analytic approach to the challenge of complex health datasets, which pose specific analytic challenges because of missing data, large size, complexity (of relationships not only between variables but also in the datasets themselves), changing populations, and nonlinear relationships between exposures and outcomes. In contrast to regression-based models—the tools historically most commonly used in clinical risk prediction analysis in medicine—Bayesian networks are compact and intuitive graphical representations that can be used to conduct causal reasoning and risk prediction analysis.

Relying on the Bayesian approach to statistical inference, Bayesian networks offer several advantages over regression-based methods:

  • Explicit representation of model structure (dependencies and independences between variables)
  • So-called “diagnostic reasoning” (from effect to cause) is naturally implemented
  • Models can be learned from data, expert knowledge (no data), or a combination of the two approaches
  • “What-if” scenarios can be explored to conduct individual-level risk prediction
  • Extension to decision models by incorporating decision and utility nodes
  • Makes no a priori assumptions of linearity or independence between variables
  • Multiple outcomes and exposures can efficiently be handled in a single model

“The increasing availability of large real-world datasets has brought about a growing interest in machine-learning algorithms for extracting knowledge from observations and for constructing personalized risk prediction models,” said author Paul Arora, PhD, Division of Epidemiology, Dalla Lana School of Public Health, University of Toronto, and Lighthouse Outcomes, Toronto, Ontario, Canada. “Bayesian networks can be embedded in real-world data sources so they can continually learn and be updated with new information thus generating individual-level risk prediction that is up-to-date and locally relevant. They offer a novel approach to risk prediction and decision analysis while maintaining a high degree of flexibility to accommodate developments in knowledge, new therapies, database size and complexity. We are pleased to see an expanding body of literature demonstrating the value of these approaches to problems in health and medicine.”

###

ABOUT ISPOR

ISPOR, the professional society for health economics and outcomes research (HEOR), is an international, multistakeholder, nonprofit dedicated to advancing HEOR excellence to improve decision making for health globally. The Society is the leading source for scientific conferences, peer-reviewed and MEDLINE®-indexed publications, good practices guidance, education, collaboration, and tools/resources in the field.

Web: www.ispor.org | LinkedIn: www.linkedin.com/company/ispororg | Twitter: www.twitter.com/ispororg (@ISPORorg) | YouTube: www.youtube.com/ispororg | Facebook: www.facebook.com/ispororg | Instagram: www.instagram.com/ispororg

ABOUT VALUE IN HEALTH

Value in Health (ISSN 1098-3015) is an international, indexed journal that publishes original research and health policy articles that advance the field of health economics and outcomes research to help healthcare leaders make evidence-based decisions. The journal’s 2017 impact factor score is 5.494. Value in Health is ranked 3rd among 94 journals in healthcare sciences and services, 3rd among 79 journals in health policy and services, and 6th among 353 journals in economics. Value in Health is a monthly publication that circulates to more than 10,000 readers around the world.

Web: www.ispor.org/valueinhealth | Twitter: www.twitter.com/isporjournals (@ISPORjournals)

Posted in Announcements, ISPOR Journals, ISPOR News & Press, Value in Health Tagged with: