It Only Takes Two Minutes: Predicting Successful Diagnoses in Medical Training Simulations Based on Task Behavior

• In Event: Advanced Technologies for Learning SIG Individual Paper Session

Abstract

Simulated learning environments provide a way to train complex skills such as making critical medical diagnoses in a safe and controlled way. These environments often require scaffolds, which should optimally be tailored to the learners’ individual needs. This study investigates whether it is possible to predict performance in a simulated medical learning environment based on broad diagnostic activities. A random-forest algorithm was very effective in classifying correct and incorrect diagnoses based on the behavior displayed. Moreover, the final algorithm was able to classify behaviors that lead to incorrect diagnoses based on only the first two minutes of behavior. Our study, thus, provides a first starting point to individualizing scaffolds based on behavior in simulated learning environments.

Authors

• Matthias Stadler, Ludwig-Maximilians-Universität München

• Constanze Richters, Ludwig-Maximilians-Universität München

• Anika Radkowitsch, Ludwig-Maximilians-Universität München

• Ralf Schmidmaier, Clinics LMU Munich

• Martin Richard Fischer, Clinics LMU Munich

• Frank Fischer, Ludwig-Maximilians-Universität München