Construct validity and generalizability of pediatrics clerkship evaluation at a problem-based medical school, Bahrain.

The aim of this study is to identify the sources of construct validity evidence and determine generalizability of the scores given to medical students on pediatrics clerkship evaluation at the College of Medicine and Medical Sciences, Arabian Gulf University (CMMS-AGU). The CMMS-AGU is a problem-based medical school in Bahrain. The sample was composed of 140 students. Validity evidence and generalizability theory were sought. Results indicated that pediatrics clerkship evaluation at CMMS-AGU has content validity. Some limitations may have occurred in response process, but the students performance was observed directly by the tutors. There was evidence of range restriction and "halo'' effect. Interrater reliability was 0.74, and the generalizability coefficient for three raters was 0.61. Students' contribution to the variance component of the global clerkship evaluation scores was 24%. Decision study indicated that seven tutors are required to achieve a generalizability coefficient of about 0.80. In conclusion, this study indicated that the pediatrics clerkship evaluation at the CMMS-AGU has, overall, some acceptable content and internal structure evidence of validity and that the raters are the major source of error. To improve the construct validity, there is a need to increase the number of raters and to train raters about the proper process of clerkship evaluation.