Background: Thoughtful laboratory testing is a key component of high-value and cost-effective patient care. Studies estimate that nearly half of laboratory tests may be unnecessary in certain patient populations. The drivers of observed variation in laboratory testing are not clearly established; our study aims to examine whether testing varies by hospitalist years of experience.
Methods: We conducted a retrospective study of a direct-care physician-staffed hospitalist service at a tertiary-referral academic medical center from July 2018 to June 2019. Data on hospitalist characteristics, including years of experience (defined as years since initial American Board of Internal Medicine certification) were obtained for all providers. We identified all orders for complete blood count (CBC), basic metabolic panel (BMP), prothrombin time, serum albumin, serum magnesium, and serum phosphorus from the electronic health record. Using multivariable linear regression, we then tested the association of years of experience with the mean total laboratory tests per patient-day, adjusted for physician gender and patient characteristics (demographics, comorbidity index, length-of-stay).
Results: Over the study period, 81 hospitalists (47 women and 34 men) with years of experience ranging from 0 to 26 years (mean: 5.6 years) on the direct-care service ordered a total of 47268 laboratory tests over 16045 patient-days. For the entire study population, the average number of laboratory tests per patient-day was 2.95. The most common test was a CBC, with or without differential (0.81 per patient-day), followed by a BMP (0.73 per patient-day). For every year of experience, there was a decrease in 0.03 laboratory tests per patient-day (95% confidence interval [CI] -0.05 to -0.002, p = 0.03) (Figure 1). First-year hospitalists (n = 11) ordered significantly more laboratory tests than all others (n = 70) with 3.53 laboratory tests per patient-day vs 2.90 (p = 0.002). Early-career hospitalists (0-3 years of experience, n=32) ordered more laboratory tests than mid-career hospitalists (4-7 years of experience, n=30) (3.19 vs 2.80, p = 0.01) and later-career hospitalists (9 or more years of experience, n=19) (3.19 vs 2.87, p = 0.04) (Figure 2). Physician gender was not significantly associated with laboratory testing.
Conclusions: Our study found an inverse association with hospitalist years of experience and laboratory testing. Early-career hospitalists ordered on average more tests per patient day than did mid- or later-career hospitalists. Hospitalists in their first year of practice ordered over 20% more laboratory tests than all other hospitalists. Efforts to promote high-value care should examine physician experience as a potential contributor to utilization.
