Understanding the “Why”: Using Predictive Modeling to Inform Outcomes
Understanding the “Why”: Using Predictive Modeling to Inform Outcomes

Members: $25
Non-members: $55

After you add this item to your cart, click "view cart" to continue to the payment screen. If you have any questions, please contact acehp@acehp.org.

Presenter: Jamie Reiter, PhD 

Description: In medical education, we frequently ask whether or not an activity was successful, that is, whether it improved outcomes. But rarely do we ask why. If an activity was successful, do we know the formula to ensure this success continues in future activities? Conversely, if an activity was less than successful, do we know which barriers prevented improvement? Factors such as learner demographics, activity format, therapeutic area, question wording, knowledge, and confidence can influence responses to behavior questions or other endpoints. Being able to answer the “why” related to educational outcomes success is an important component of developing activities that ensure best practices are being implemented, resulting in improved patient outcomes.

There are two main approaches to understanding the factors influencing outcomes. The first is to use traditional statistical analysis to compare subgroups of participants (e.g., primary care physicians and neurologists), separating them by variables suspected of influencing outcomes. This approach comes with its share of challenges, primarily: 1) How do we decide which variables to use, and 2) Exploring all possible subgroups from all possible variables can be cumbersome. The second approach is to use predictive modeling, the most common method of which is regression.

The medical education industry is starting to appreciate the value in predictive modeling, but many providers may not feel they have the skills to perform predictive modeling or don’t realize the software is readily available. Statistical analysis is typically best accomplished by a knowledgeable statistician using software specifically designed for statistics (e.g., SAS, SPSS). However, with caution, basic statistical procedures can be conducted by non-experts who have some degree of statistical knowledge, and using software such as Excel or online calculators.

This presentation will provide an overview of predictive modeling and provide an example of how to conduct regression using Excel as well as a few vetted online calculators. It is not the intention of this presentation to provide a full understanding of the mathematics and statistical theory behind predictive modeling, rather a basic overview. Some knowledge of, or at least comfort with, statistics is recommended.

Learning Objectives:

  • Compare/Differentiate standard statistical methods and predictive modeling for gaining insight into educational activity success
  • Obtain skills for conducting a predictive modeling analysis.