Psychometric Analysis and Validity of the Medical College Admission Test at the Aga Khan University

Objectives. The purpose of the study was to (1) determine the effectiveness of admission criteria used by the Aga Khan University Medical College for predicting students. performance during medical school, and (2) to identify a parsimonious model for admission decisions that is most predictive of students' performance using latent variable path analysis.;Results. English language was shown to be the only predictor in the regression model for clinical reasoning and decision making (F=14.92, p =<0.001). A three factor structure was identified for predicting validity of admission decisions for achievement in medical school employing Maximum Likelihood (ML) estimations (n=112). Fit indices: chi2 (21) = 59.70, p =<.001 CFI=.873; RMSEA = 0.129; SRMR = 0.093.;Conclusion. A three factor model of aptitude for medicine, science knowledge and achievement in medical school was the most theoretically coherent model for predictive validity of admissions at AKUMC. The written tests of cognitive ability used for admission and progress during medical school had adequate validity, but measures of non cognitive ability need further improvement. Admission favors applicants from the British system of education. Prior attainment and scores on science subtest of the admission test predict performance in the first two years of medical school. The performance of written tests of clinical reasoning and decision making are best predicted by the scores on the English language subtest while being a female predicts better performance in assessment of clinical skills.;Method. The data of three cohorts of students (n=276) admitted in the years 2003, 2004 and 2005 was obtained from the AKU Registrar.s Office and the AKUMC Examination Cell. Progress of the three cohorts during medical school was followed longitudinally through years one to five. Psychometric analysis of the admission and medical school measures was done to gather evidence of reliability and validity Descriptive statistics were computed and sub models were run for each of the dependent variable. Factor analysis for identification of latent variables and linear regression analysis using backward regressions models was done to determine the relationship between the independent and dependent variables. Structural equation modeling was used to study latent variable path models assessing the impact of the independent variables on dependent variables.