Loading…
This event has ended. Create your own event → Check it out
This event has ended. Create your own
View analytic
Monday, February 27 • 10:00am - 10:45am
DEER VALLEY - [Oral Presentation] 1. Novel Approaches to Predictive Modeling for Understanding Influences of Practice Behavior: An Example Using Alzheimer’s Disease

Sign up or log in to save this to your schedule and see who's attending!

10:00 AM - 10:15 AM

Novel Approaches to Predictive Modeling for Understanding Influences of Practice Behavior: An Example Using Alzheimer’s Disease

J. Reiter, J. Perez, S. Tordoff, W. Faler, CME Outfitters
Abstract Body: Introduction. An essential component of improving patient outcomes through medical education is ensuring healthcare providers (HCPs) perform according to best practices. Traditional statistical comparisons of pre- versus post-activity performance are important for demonstrating performance improvement. However, they do not provide information regarding the factors that influence practice behaviors, which will help guide needs assessments for future activities and ensure the appropriate topics, formats, questions, and audiences are targeted. PredictCME utilizes a form of predictive modeling known as CHAID (chi-square automatic interaction detection). Although frequently used in data mining, CHAID has not been utilized in medical education. PredictCME can be used to determine which variables most impact knowledge, competence, behavior, or other endpoints. It has two main advantages over linear or logistic regression: 1) decision tree-based output which allows for a more informative and user-friendly interpretation, and 2) ability to utilize both continuous and categorical data. This presentation provides results from a CHAID analysis of real-world data from an educational activity on Alzheimer’s disease (AD). Research Question. What factors influence practice behaviors in HCPs seeing patients with Alzheimer’s disease? Methods. Data from 262 HCPs participating in an educational activity on AD were analyzed using PredictCME. A question related to practice behavior was entered into the model as the response variable, with variables such as knowledge, number of patients seen with AD, years in practice, and confidence entered as predictor variables. Results. Results showed that the strongest predictor of practice behavior was confidence. A secondary predictor was the number of patients with AD seen by the learners. Discussion. For future activities, it will be important to consider ways to improve HCP confidence as well as address the needs of HCPs who don’t see a large number of patients with AD. Findings from the PredictCME analysis demonstrate the utility in using predictive modeling to better understand the influences of practice behavior, which in turn will help maximize the impact of future activities, and ultimately patient outcomes. 



Monday February 27, 2017 10:00am - 10:45am
DEER VALLEY

Attendees (3)