Methods: We performed a retrospective cohort study using a national dataset of adult hospitalizations from The Advisory Board Company from 2009 to 2014. We excluded hospitalizations that were pregnancy-related, had a urologic procedure, or had LOS greater than 30 days. The exposure was a urine culture on day 1 of hospitalization. Hospitalizations were matched with coarsened exact matching by facility, patient age, gender, Medicare Severity-Diagnosis Related Group (MS-DRG), DRG severity level, DRG mortality level, Elixhauser comorbidity score, and ICD-9 codes for infection on admission. A multi-level linear Poisson model and a multi-level linear regression model were used to determine the impact of urine culture on inpatient antibiotic use and LOS.
Results: Matching produced a cohort of 88,481 (n=41,070 with a urine culture on day 1, n=47,411 without a urine culture). A urine culture on day 1 of admission was associated with an increase in days of inpatient antibiotic use (incidence rate ratio 1.26; p-value <0.001). The impact of urine culture testing on antibiotic days was greatest for diagnoses where antibiotic use was uncommon (Table 1). The difference in antibiotic use between admissions with and without a urine culture on day 1 resulted in an additional 36,607 days of inpatient antibiotic use. Urine culture on day 1 of the admission resulted in a 2.1% (SE 0.7%) increase in LOS. The predicted difference in bed days of care between hospital admissions with and without a urine culture on the first day of admission resulted in 6,071 excess bed days of care overall. The impact of urine culture testing varied by MS-DRG. For example, urine culture testing on day 1 of the hospitalization resulted in a 4.5% increase in LOS (875 excess days) for patients admitted for major joint replacement of the lower extremity and 1,006 excess days of antibiotic use.
Conclusions: Collecting a urine specimen for culture is a simple procedure that has major downstream impacts. Overall, patients with a urine culture sent on the first day of the hospital admission receive more days of antibiotics and have a longer hospital stay than patients who do not receive a urine culture. These findings varied by diagnosis group. Targeted interventions for specific diagnoses may achieve the best balance between reducing the potential harm associated with low-yield urine cultures and supporting clinician autonomy to order a culture in the appropriate clinical setting.