Multi-professional image interpretation: Performance in preliminary clinical evaluation of appendicular radiographs

The study evaluated the performance of a cohort of healthcare professionals in image interpretation of appendicular radiographic examinations following a preliminary clinical evaluation (PCE) course.

A systematic review and meta-analysis of image interpretation performance by non-medical staff was undertaken. Thirty three participants (seven radiographers, nineteen nurses, five paramedics, and two physiotherapists) completed an image based Objective Structured Examination (OSE) following a short course on PCE of appendicular radiographs. The case bank comprised 25 appendicular radiographic examinations. Prevalence of abnormal examinations approximated 52% (traumatic conditions). The OSE test scores were analysed with Alternative Free-response Receiver Operating Characteristic (AFROC) calculation of Area Under the Curve (AUC), sensitivity, specificity, and Cohen's Kappa (k) for multi-reader agreement. Professional subgroup results were compared with Analysis of Variance (ANOVA), and compared to the literature review.

The participant subgroup results for radiographers were sensitivity 79.7%, specificity 92.9%, k = 72; AUC 86.5. Nurses sensitivity 76.2%, specificity 80.4%, k = 56; AUC 78.3. Paramedics sensitivity 80.3%, specificity 81.6%, k = 61; AUC 81.5. Physiotherapists sensitivity 90.4%, specificity 91.6%, k = 82; AUC 91.8. ANOVA Samples scores (f = 1.5; p = 0.23), sensitivity (f = 0.5; p = 0.65), specificity (f = 2.8; p = 0.05). The comparison of the radiographers and nurses results to the published literature were encouraging (SROC AUC 93.1 and 91.2, respectively). The paramedics and physiotherapists results could not be compared to published standards or abilities on the task due to a lack of literature available.

The study results of image interpretation and commentary of appendicular radiographs in an academic environment by this multi-professional sample were to a high standard, further work is recommended on a larger sample.