Does student confidence on multiple-choice question assessments provide useful information?

Feedback from multiple-choice question (MCQ) assessments is typically limited to a percentage correct score, from which estimates of student competence are inferred. The students' confidence in their answers and the potential impact of incorrect answers on clinical care are seldom recorded. Our purpose was to evaluate student confidence in incorrect responses and to establish how confidence was influenced by the potential clinical impact of answers, question type and gender.

This was an exploratory, cross-sectional study conducted using a convenience sample of 104 Year 3 dental students completing 20 MCQs on implant dentistry. Students were asked to select the most correct response and to indicate their confidence in it for each question. Identifying both correctness and confidence allowed the designation of uninformed (incorrect and not confident) or misinformed (incorrect but confident) responses. In addition to recording correct/incorrect responses and student confidence, faculty staff designated incorrect responses as benign, inappropriate or potentially harmful if applied to clinical care. Question type was identified as factual or complex. Logistic regression was used to evaluate relationships between student confidence, and question type and gender.

Students were misinformed more often than uninformed (22% versus 8%), and misinformed responses were more common with complex than factual questions (p < 0.05). Students were significantly more likely to be confident of correct than incorrect benign, incorrect inappropriate or incorrect harmful answers (p < 0.001), but, contrary to expectations, confidence did not decrease as answers became more harmful.

Recording student confidence was helpful in identifying uninformed versus misinformed responses, which may allow for targeted remediation strategies. Making errors of calibration (confidence and accuracy) more visible may be relevant in feedback for professional development.