Meta-analysis of predictors of dental school performance

Accurate prediction of which candidates show the most promise of success in dental school is imperative for the candidates, the profession, and the public. Several studies suggested that predental GPAs and the Dental Admissions Test (DAT) produce a range of correlations with dental school performance measures. While there have been similarities, such as the finding by Bergman et. al., 2006 and DeBall, et. al., 2002 that the DAT’s Reading Comprehension (RC) section was a significant predictor for portions of the National Board Dental Examinations Part 1 (NBDEI), correlations were disparate. A narrative review conducted by Ranney et. al., (2005) did not apply quantitative methods and changes in the DAT and NBDE over time suggest prediction has changed as well. Ranney et. al., (2005) found extensive variation in correlations. Dental school admissions officers perceive distinctions among the GPAs of their “feeder schools,” and know that grade inflation is a greater issue at some schools than others. The DAT measures students on a common rubric. Yet, correlations of the DAT with dental school performance span from low to high. The literature is unclear as to how well and how consistently the DAT and grades predict future performance for various groups (Kramer, 1999). Improved understanding of prediction may enhance its implementation. That is what this paper attempts to do. A literature search found nine articles with data that could be used toward this purpose. This dissertation then applied two different methods of meta-analysis, one more streamlined, espoused by Rosenthal (1991), simply combined results, calculated confidence intervals and tested for heterogeneity of results. A second analysis followed the direction of Hunter & Schmidt (2004). After combining results, through a series of corrections, it produced correlations without the effects of range restriction and unreliability of measures. Across both meta-analyses methods, standardized tests were more closely associated with performance measured soon thereafter. Associations between grades and criterion increased over time. Extrapolation of results to other populations is not endorsed due to shortcomings associated with available data and methodology.Ph. D.Includes bibliographical referencesby Jeanette E. DeCastr