Later ED Assignments Linked to Fewer Contrast CT Orders

Key Takeaways

• Among patients seen by attending physicians in an Arizona academic ED, later positions in an attending's daily assignment sequence were associated with less contrast-enhanced CT ordering.
• Any CT, radiography, and ultrasound ordering also declined modestly at later assignment positions.

Investigators examined attendings' daily assignment sequence in a single urban academic tertiary-care emergency department (ED) in Arizona in a study published in the Western Journal of Emergency Medicine. They retrospectively analyzed electronic records from January 1, 2019, through December 31, 2024, covering 287,925 encounters across approximately 41,132 physician shifts and 77 attending physicians. Patients were automatically assigned to physicians in rotation, and imaging orders were attributed to the attending linked to each encounter.

Investigators treated each patient's sequential physician-day assignment position, from first through 17th, as a continuous exposure. Encounters beyond that range or without a linked attending were excluded. Visit-level, non-mutually-exclusive outcomes recorded orders for any CT, CT with IV contrast, CT without contrast, radiography, MRI, and formal or point-of-care ultrasound. Adjusted pooled logistic models used physician-clustered robust errors. Mixed-effects models included physician random intercepts, and within-between estimates separated assignment patterns within and across physicians. Models adjusted for age, sex, triage acuity, chief complaint, arrival hour, night shift, and weekend. The absolute contrast compared marginal predicted ordering probabilities at the ends of the assignment sequence.

Contrast-enhanced CT ordering fell with each later assignment in the pooled analysis: adjusted odds ratio (aOR) 0.990 (95% confidence interval, 0.986–0.994). The within-physician analysis showed a similar association. These estimates describe adjusted odds, not an absolute probability change or a causal response by individual physicians. Physician-level variation was present, but tests comparing within- and between-physician coefficients did not establish distinct effects.

Marginal predicted contrast-enhanced CT ordering was 21.8% at the first assignment versus 19.6% at the 17th, an absolute risk difference of −2.2 percentage points. These modeled rates concern orders, not observed patient outcomes. Adjusted ordering of any CT, radiographs, and ultrasound also declined modestly. Noncontrast CT showed no significant pooled association; MRI was nonsignificant in the pooled analysis but significant in the mixed-effects analysis.

At one institution, assignment position could also track proximity to shift-end sign-out; it captured neither active census nor acuity-weighted load. Within-physician patterns therefore could not distinguish workload adaptation from handoff behavior. Incomplete clinical details, variation in imaging resources, possible pandemic-era confounding, and no adjustment for multiple imaging-outcome comparisons further qualify the associations. Orders cannot establish whether scans were indicated, whether foregoing them was appropriate, or whether missed diagnoses, return visits, mortality, or other patient outcomes changed.

Clinician Questions

How do ED imaging patterns by physician assignment order compare with findings based on department-wide crowding?

Martini and colleagues noted that earlier research using department-level crowding or overall physician-volume measures generally described more imaging as volume rose. By contrast, their per-encounter comparison by each physician's daily assignment position found modestly less ordering for several modalities.

Why did the analysis exclude physician assignments beyond the 17th patient?

In this Arizona academic ED, automatic rotational allocation capped assignments to an attending at 17 per shift before routing a patient to another physician or prompting a handoff. This was an institutional operational rule, not a clinical workload threshold. Encounters beyond the cap and those without a linked attending identifier were excluded.