J O Nnodim
ABSTRACT
An analysis of 596 multiple-choice questions (MCQs) on human anatomy given at three First Professional Examinations for medical students is reported. The MCQ paper at each examination was 200 items long and consisted of three item-types: A, K and T/F. Each A-type item comprised a stem and five options, only one of the latter being the correct or best answer. Items of the K-type consisted of a stem and four responses, any number of which may be correct. The T/F items were of the three-response kind, the available options being 'true', 'false' and 'don't know'. Test reliability was computed by internal analysis, using the Kuder-Richardson 20 formula. Measures of concurrent validity were obtained by correlating the scores in the MCQ papers with the overall outcome of the First Professional Examination. Indices of item facility, discrimination and abstention were calculated. The effects of item-type and the availability of the 'don't know' option on examinee performance were also determined. Reliability (alpha) and concurrent validity (Pearson r) coefficients in the ranges of 0.71-0.85 and 0.80-0.93 (P less than 0.05) respectively were recorded. Regression analysis revealed the MCQ papers to be less sensitive predictors of the aggregate performance than the essay papers. The proportion of highly discriminatory and excessively difficult items was highest for the K-type. When the same K-type questions were re-exhibited in the indeterminate format, the examinees performed significantly better. Higher scores were also recorded when candidates were required to respond to all the questions than when they were offered the 'don't know' option and the percentage gain was higher for the low-scoring examinees. The appropriateness of multiple-choice testing as a tool for assessing student achievement in human anatomy is discussed.