Shaping CME with Predictive Modeling: Part 1

Shaping CME with Predictive Modeling: Part 1

The continuing medical education (CME) field is always changing. That’s a good thing, because it means we are keeping up with the rapid changes in the medical field, accommodating the needs of our learners, and discovering newer and better ways to measure the impact of CME.  But this constant change can be a challenge as well. How do we develop best practices that align with what is essentially a moving target? Part of the answer lies in our ability to share with one another what worked and what didn’t. Another aspect is our willingness to take chances. In outcomes, there are definitely standard methods that have proven value, such as established writing guidelines and statistical comparisons of pre- and post-test data (which I will cover in future blogs). But we can take it a step further. For example, we have an abundance of data at our fingertips that can be used not just to evaluate what we have done, but to guide what we will do in the future. Predictive modeling has the potential to shape CME, so it is certainly a topic worth exploring. For this blog series, I’ll briefly define predictive modeling and list a few examples. Part 2 of the predictive modeling blog will delve further into describing the different models and which of these would most apply to CME.

Background on Predictive Modeling

I have been interested in predictive modeling since my mathematical modeling days in grad school. And if you’ve seen my LinkedIn posts ( as well as some of the CMEO Facebook ( and Twitter (@cmeoutfitters) posts, you’ll know we’ve had a lot of success with it lately. Predictive modeling has been widely used for decades across multiple industries. In general terms, it can be defined as a statistical or mathematical method for predicting the probability of an outcome (aka response variable) based on one or more input variables (aka predictor variables).

What Predictive Modeling Means to CME

Predictive modeling has particular value in the CME world, as it can help us determine variables, such as academic degree, specialty, or experience, that influence outcomes such as knowledge, confidence, and behavior. Of course, the holy grail would be to determine what variables predict patient outcomes! Why do I consider predictive modeling a best practice? Because results from predictive modeling analyses can be applied to development of future activities, since we can then target certain audiences, content areas, formats, etc. to maximize educational impact.

Types of Predictive Models

You may be familiar with one of the most common forms of predictive model, namely, regression. Regression is actually a generic term for a category of predictive models, with linear and logistic regression being the most common. Other forms of predictive modeling include, but are not limited to, neural networks, naïve Bayes, and CHAID (chi-square automatic interaction detection). Which model works for you will depend on multiple factors. My next blog will go into more detail on some of these models. In the meantime, you can start thinking about some research questions, i.e., what types of outcomes you want to predict and what variables may be the most relevant to you as predictors.

Thanks for reading, and stay tuned for Part 2!




About The Author

jr Jamie Reiter, PhD Director, Educational Outcomes