Done or Almost Done? Improving OSCE Checklists to Better Capture Performance in Progress Tests.

Construct: The impact of using nonbinary checklists for scoring residents from different levels of training participating in objective structured clinical examination (OSCE) progress tests was explored.

BACKGROUND: OSCE progress tests typically employ similar rating instruments as traditional OSCEs. However, progress tests differ from other assessment modalities because learners from different stages of training participate in the same examination, which can pose challenges when deciding how to assign scores. In an attempt to better capture performance, nonbinary checklists were introduced in two OSCE progress tests. The purposes of this study were (a) to identify differences in the use of checklist options (e.g., done satisfactorily, attempted, or not done) by task type, (b) to analyze the impact of different scoring methods using nonbinary checklists for two OSCE progress tests (nonprocedural and procedural) for Internal Medicine residents, and (c) to determine which scoring method is better suited for a given task.

APPROACH: A retrospective analysis examined differences in scores (n = 119) for two OSCE progress tests (procedural and nonprocedural). Scoring methods (hawk, dove, and hybrid) varied in stringency in how they awarded marks for nonbinary checklist items that were rated as done satisfactorily, attempted, or not done. Difficulty, reliability (internal consistency), item-total correlations and pass rates were compared for each OSCE using the three scoring methods.

RESULTS: Mean OSCE scores were highest using the dove method and lowest using the hawk method. The hawk method resulted in higher item-total correlations for most stations, but there were differences by task type. Overall score reliability calculated using the three methods did not differ significantly. Pass-fail status differed as a function of scoring methods and exam type, with the hawk and hybrid methods resulting in higher failure rates for the nonprocedural OSCE and the dove method resulting in a higher failure rate for the procedural OSCE.

CONCLUSION: The use of different scoring methods for nonbinary OSCE checklists resulted in differences in mean scores and pass-fail status. The results varied with procedural and nonprocedural OSCEs.