Abstract
Despite powerful efforts to maximize nursing school enrollment, schools and colleges of nursing are faced with high rates of attrition and low rates of completion. Early identification of at-risk students and the factors associated with graduation outcomes are the main foci for the studies that have addressed attrition and completion rates in nursing programs. Machine learning has been shown to perform better in prediction tasks than traditional statistical methods.
The purpose of this study was to identify adequate models that predict, early in a students career, if an undergraduate nursing student will graduate within six college years. In addition, factors related to successful graduation were to be identified using several of the algorithms.
Predictions were made at five time points: the beginning of the first, second, third, fourth years, and the end of the sixth year. Fourteen scenarios were built for each machine learning algorithm through the combinations of different variable sections and time points.
College of Nursing in a private university in an urban Midwest city, USA.
Seven hundred and seventy-three full time, first time, and degree-seeking students who enrolled from 2004 through 2012 in a traditional 4-year baccalaureate nursing program.
Eight popular machine learning algorithms were chosen for model construction and comparison. In addition, a stacked ensemble method was introduced in the study to boost the accuracy and reduce the variance of prediction.
Using one year of college academic performance, the graduation outcome can be correctly predicted for over 80% of the students. The prediction accuracy can reach 90% after the second college year and 99% after the third year. Among all the variables, cumulative grade points average (GPA) and nursing course GPA are the most influential factors for predicting graduation.
This study provides a potential mode of data-based tracking system for nursing students during their entire baccalaureate program. This tracking system can serve a large number of students automatically to provide customized evaluation on the dropout risk students and enhance the ability of a school or college to more strategically design school-based prevention and interventional services.