Human- Versus System-Level Factors and Their Effect on Electronic Work List Variation: Challenging Radiology’s Fundamental Attribution Error

PURPOSE:The aim of this study was to analyze sources of variation influencing the unread volume on an electronic abdominopelvic CT work list and to compare those results with blinded radiologist perception. METHODS:The requirement for institutional review board approval was waived for this HIPAA-compliant quality improvement effort. Data pertaining to an electronic abdominopelvic CT work list were analyzed retrospectively from July 1, 2013, to June 30, 2014, and modeled with respect to the unread case total at 6 pm (Monday through Friday, excluding holidays). Eighteen system-level factors outside individual control (eg, number of workers, workload) and 7 human-level factors within individual control (eg, individual productivity) were studied. Attending radiologist perception was assessed with a blinded anonymous survey (n = 12 of 15 surveys completed). RESULTS:The mean daily unread total was 24 (range, 3-72). The upper control limit (48 CT studies [3 SDs above the mean]) was exceeded 10 times. Multivariate analysis revealed that the rate of unread CT studies was affected principally by system-level factors, including the number of experienced trainees on service (postgraduate year 5 residents [odds ratio, 0.83; 95% confidence interval, 0.74-0.92; P = .0008] and fellows [odds ratio, 0.84; 95% confidence interval, 0.74-0.95; P = .005]) and the daily workload (P = .02 to P < .0001). Individual faculty productivity had a weak effect (Spearman ρ = 0.13, P = .03; adequacy: 3% of variance explained). The majority (67%) of radiologists (8 of 12) completing the survey believed that variation in faculty effort was the most important influence on the daily unread total. CONCLUSIONS:System-level factors best predict the variation in unread CT examinations, but blinded faculty radiologists believe that it relates most strongly to variable individual effort.