Impact of Physician Patient Load on Imaging Use in the Emergency Department.

introductionEmergency department (ED) crowding may alter physician decision-making, yet little is known about how real-time physician workload affects diagnostic imaging use. We examined the association between physician patient load and the proportion of patients who received diagnostic imaging, including computed tomography (CT), CT with intravenous (IV) contrast, CT without contrast, plain radiography, magnetic resonance imaging (MRI), and ultrasound (US).

methodsWe conducted a retrospective cohort study of all ED visits at an academic tertiary care center (January 2019-December 2024). Physician workload was defined as the sequential patient assignment order per physician per day (range 1-17). Our primary outcome measures were binary indicators of whether each imaging modality was ordered during the ED visit: CT (any); CT with IV contrast; CT without IV contrast; radiograph, MRI; and US. Multivariable logistic regression with physician-clustered robust standard errors estimated adjusted odds ratios (aOR) for each imaging modality, controlling for age, sex, triage acuity, chief complaint, and temporal factors. We also performed mixed-effects logistic regression with physician random intercepts and a within-between (Mundlak) decomposition. Absolute risk differences (ARD) were computed for the 17th versus the first patient.

resultsAmong 287,925 encounters across approximately 41,132 physician shifts managed by 77 physicians, higher workload was associated with decreased odds of imaging for most modalities. Computed tomography with IV contrast showed the strongest effect (aOR 0.990, 95% CI, 0.986-0.994; within-physician aOR 0.991; ARD -2.2 percentage points [pp], from 21.8% to 19.6%). Results were as follows: any CT, aOR 0.994 (0.991-0.997); ARD -1.9 pp (from 38.7% to 36.8%); radiograph, aOR 0.994 (0.990-0.998, P < .001); ARD -1.6 pp (from 44.2% to 42.6%); US, aOR 0.994 (0.990-0.998, P < .001); ARD -0.7 pp (from 9.9% to 9.2%); and MRI was nonsignificant in the pooled model (aOR 0.993, P = .09) but significant in the mixed-effects model (P < .001). Noncontrast CT was unaffected (aOR 1.000, P = .87). Intraclass correlation coefficients ranged from 0.020-0.047, indicating 2-5% of imaging variance was attributable to physician-level differences.

conclusionHigher physician workload was associated with modestly reduced imaging use, particularly for resource-intensive modalities. Noncontrast CT was unaffected. Within-physician analyses were consistent with the interpretation that these effects reflect real-time behavioral adaptation rather than physician practice style alone. These findings highlight a trade-off between throughput and diagnostic testing that warrants further study of its impact on clinical outcomes.