Abstract Manager

Oral

OB2-3-2

no

General Call (Workshop, Oral Presentation, Poster Presentation)

Oral

Education Research

Student

yes

Image Interpretation: Evidence Informed Learning Opportunities

Learning analytics is the measurement, collection, analysis and reporting of data for the purposes of understanding and optimizing learning. We derived learning analytics on a challenging radiograph to determine variables that predict for an incorrect diagnostic interpretation. Furthermore, we determined image review processes inherent to novice versus experienced participants that were associated with a higher diagnostic performance.

Physician participants attempted to detect pneumonia on 200 pediatric chest radiograph (pCXR) on a digital platform. We examined associations with diagnostic success with respect to physician demographics and pCXR variables.

We enrolled 83 participants (20 medical students, 40 postgraduate trainees and 23 faculty), obtaining 12,178 case interpretations. Variables that predicted for pCXR interpretation difficulty were pneumonia present vs. absent (β = 8.7; 95% CI 7.4, 10.0); low vs. high visibility of pneumonia (β = -2.2; -2.7, -1.7); non-specific lung pathology present vs. absent (β = 0.9; 0.4, 1.5); and, localized vs. multifocal pneumonia (β = -0.5; -0.8, -0.1). Novices reviewed both available radiograph views less often than faculty and were more accurate when they reviewed both views (p<0.0001). Novices also spent less time reviewing images, despite lower accuracy scores (p<0.0001). Physician certainty was associated with an increased probability of case correctness, and this effect was more prominent in faculty (p<0.0001).

Learning analytic information can be used to allow for a customized weighting of which cases to practice and predict participant review processes that may lead to diagnostic error.

Clinical Skill

Radiographs

Assessment

General

Teaching and learning

Postgraduate

Continuing Medical Education

Faculty Development

Yes

Yes

Yes

Yes

Yes