Towards the development of an early warning system for the identification of the student at risk of failing the first year of higher education

Abstract

The purpose of this study was to use first-year test results to develop an early warning system for the identification of freshmen at risk of failing. All students registered between 1989 and 1997 for the six-year programmes chiropractic and homoeopathy were included in this ex post facto study. A descriptive study firstly indicated a serious problem of attrition with on average only 66% of chiropractic and 55% homoeopathy freshmen successfully completing the first year. A relationship was demonstrated between both first and second test results and outcome at the end of the first year of studies. A logistic regression model estimated retrospectively from first test results in physiology, anatomy, biology and chemistry was able to discriminate between successful and non-successful freshmen with an overall predictive accuracy of 80.82%. When this model was validated on a different set of data it was shown to have a very high sensitivity and was thus able to correctly identify >93 % of the potentially at risk freshmen. It also had a low Type II error ( <7%) and thus missed very few of the freshmen at risk of failing. A logistic regression model estimated retrospectively from second test results in physiology, anatomy, biology and chemistry had an overall predictive accuracy of 85.94% . The validated model had a sensitivity of 67% which was too low for the model to be of much use as a management tool for the identification of the freshmen at risk of failing. However, the model was shown to have a high specificity and was able to correctly identify >93% of the potentially successful freshmen. It also had a low Type I error (14.29%). Discriminant analysis models estimated from both first and second test results in physiology, anatomy, biology and chemistry produced strong support for the use of test results for the early identification of those freshmen who would need support in order to be successful. It is suggested that the objective models developed in this research could identify the freshman in need of support at an early enough stage for support measures to still have a positive effect on attrition.