Background: Hospital medicine is a maturing field that increasingly focuses on providing high value patient care. A key question in the value equation is how different patient census sizes impact care. Our recent research showed that higher patient censuses were associated with longer hospital length of stay and costs. This study follows up on this initial analysis hypothesizing that increases in physician relative value units (RVU) and patient census will be associated with increased utilization of medical imaging and specialty consults, potentially representing medical overuse.
Methods: Retrospective cohort study examining patients admitted to a hospitalist service between February 1, 2008 and January 31, 2011 in a single two-hospital academic community health system. We focused on patients discharged with a primary diagnosis of Pneumonia, Chronic Obstructive Pulmonary Disease (COPD), Syncope, Acute Kidney Injury (AKI), and Congestive Heart Failure (CHF) as defined by selected ICD-9-CM codes. We created separate logistic regression models to examine the outcomes of the number of imaging tests and sub-specialty consults ordered for each disease condition. The primary exposure of interest was physician workload, which we defined separately as 1) the physician RVU and 2) physician census on the day of admission. Each logistic regression model was adjusted for age, sex, race, insurance, teaching, hospital, prior hospitalization within 30 days, and average hospital occupancy.
Results: During the study period 1478 Pneumonia, 571 COPD, 1151 Syncope, 601 AKI, and 778 CHF admissions met inclusion criteria. Models for patients with pneumonia showed that an increase in 10 units of average RVU was associated with 28% greater odds of ordering a Chest CT, while an increase in 10 patients on the census was associated with a 13 and 15% increase in odds of an infectious disease and total consults respectively. In COPD increases in RVU decreased odds of Chest CT by 26% while changes in census had no impact on imaging or consultations. In syncope, AKI, and CHF increases in RVU had no impact on imaging and consult ordering. However in syncope, increases in census led to decreased odds of MRI and Carotid US by 15% and 20% increased odds of total consults ordered. In AKI, increases in census decrease odds of Abdominal US by 20% and increase odds of Nephrology consults by 27%. Lastly for CHF, increases in census lead to decreased echocardiograms by 20% and increases in Cardiology consults by 34%.
Conclusions: Counter to our proposed hypothesis increases in physician workload were not associated with increased imaging utilization except for RVU being associated with increased Chest CT in pneumonia. The results did show broad association with patient census and increased specialist consults across the 5 diagnoses of interest. While we hypothesized that higher workloads would lead to an over-reliance on imaging and consults compared to clinical evaluation, the results suggest that higher workload does result in increased consults but that when managing patients without a consultant hospitalists may over utilize imaging.