A clinical prediction model can assist doctors in arriving at the most likely diagnosis or estimating the prognosis. The aim of such decision support is to positively impact clinical decision making and patient outcome. However, even when its probability estimates are both accurate and valid in new patients, a prediction model does not necessarily enhance a doctors decision making. In so-called impact studies the prediction model is implemented in daily practice and the effects on clinical outcomes are compared to care-as-usual i.e., care without using the model. The downside of such prospective comparative studies is that they often require a lot of time and resources. Before envisioning such an impact study it is important to ensure a reasonable chance that the use of the prediction model will have a positive effect on clinical practice. This presentation will address the challenges that we faced during the impact studies on a prediction model for postoperative nausea and vomiting. Using this clinical problem as an example, we will arrive at several recommendations on how to prepare a prediction model for implementation, and how to design and analyze a prediction model impact study.
Background Teus Kappen is an anesthesiologist from the Netherlands, who completed his residency at the University Medical Center Utrecht. During his residency, he started his Ph.D. training and research project on the use of prediction models and decision support under supervision of professors Cor Kalkman and Carl Moons (Thesis defense - May 2015). He also earned his Master of Science degree in Clinical Epidemiology at Utrecht University. After residency, he completed a fellowship in pediatric intensive care at the Children's Hospital Utrecht. After the completion of his fellowship, he joined the staff of the Anesthesiology Department of the UMC Utrecht. July 1st 2015 he joined the Vanderbilt Anesthesiology Department as a research scholar. His main interests are how computers may assist physicians in their decision making.