Predicting academic success in health professions training and education: Insights from a data mining analysis

Abstract

Dropping-out is a very common phenomenon among students enrolled in Health Professions degree courses. For this reason, it would be extremely useful to have a model with the main variables that can predict, with a certain degree of accuracy, school achievement and success, in order to put forward interventions and counseling. We used administrative available data for the academic years 2007-2008, 2008-2009 and 2009- 2010 for Health Professions degree courses held in Genoa (Liguria Region, Italy), performing a Random Forest (RF) Analysis and a Classification And Regression Tree (CART) Analysis. For the Academic Years 2008-2009 and 2009-2010 results of ad hoc psychological questionnaires investigating, among others, motivation and possible influences from the peer environment were also available. We found that attendance to the so called ‘tirocinio,’ an Italian word describing practical training activities and clinical placements, is the most important variable in terms of variance and that administrative social-demographic variables alone are sufficient to explain school success and achievement (AUC 0.99500 for 2007, AUC 0.98774 for 2008 and AUC 1.00000 for 2009). However taking into account also the results of the questionnaire the goodness-offit of our model slightly decreases (AUC 0.91890 for 2008, AUC 0.99641 for 2009). Therefore, we are planning a series of curriculum counseling activities for high school students, based on the obtained results.