Background: The Choosing Wisely campaign discourages CT imaging in low-risk patients with suspected pulmonary embolism (PE) (e.g., low clinical probability and negative d-dimer). Few studies have investigated patient, provider, or operational characteristics associated with overuse of CT imaging. One possible approach to this research question is to examine variability in diagnostic yield (i.e., the percentage of CTAs positive for PE) across characteristics such as patient language. We hypothesized that physicians order more CTAs during times of high occupancy or patient-provider ratios, resulting in a lower diagnostic yield for PE. We conducted a retrospective analysis to examine the association between hospital occupancy and patient:provider ratios with diagnostic yield for PE and pretest probability (PTP ).
Methods: We conducted a retrospective analysis of 1,049 patients aged ≥18 years who received a CT angiography for PE at a large tertiary care hospital between January 1, 2016 and December 31, 2017. Patients identified as having chronic thromboembolic disease, small sub-segmental PE, or indeterminate scans were excluded. The present analysis was additionally restricted to patients on an adult inpatient service at the time of CTA order (N=275; 30.2% of all eligible patients). Trained physicians manually abstracted data on Wells score components; all other clinical data was electronically abstracted from electronic medical records. We retrospectively calculated each patient’s 2-tiered Wells score and categorized patients as PE likely (Wells >4) or unlikely (Wells ≤4) (intra-rater reliability, 0.72; inter-rater reliability, 0.47). We calculated percentiles of adult inpatient occupancy, with the lower and upper percentiles defined using the minimum (109 patients) and maximum (188 patients) observed adult inpatient occupancies. Patient:provider ratio was the number of patients a unique provider billed for on a given day. We compared diagnostic yield and PTP across hospital occupancy and patient:provider ratio at the time of CTA order using Fisher’s exact tests.
Results: Overall diagnostic yield was 10.7%. We did not observe an association between inpatient occupancy and diagnostic yield (diagnostic yield for ≤median occupancy, 6.6%; for 50th-75th percentile occupancy, 13.9%; for 75th-95th percentile occupancy, 10.9%, for ≥95% occupancy, 9.1%; p=0.31. Patient:provider ratio (range: 1:1-18:1) was also not associated with diagnostic yield (diagnostic yield for 1:1-5:1 patient-provider ratios, 10.2%; for 5:2-9:1, 15.7%; for 9:2-14:1, 5.1%; for >14:1, 14.3%; p=0.36). Pretest probability did not vary by inpatient occupancy or patient:provider ratio.
Conclusions: The identification of patient, provider, or operational characteristics associated with increased CTA orders could be used to inform interventions aimed at reducing imaging in low-risk patients. However, in this study, neither patient occupancy nor patient:-provider ratios were associated with diagnostic yield for PE. Future studies should examine associations between markers of provider workload across hospital systems, as the generalizability of findings from a single center are unknown.