Background: Microbiology data are utilized to quantify epidemiology and trends in pathogens, antimicrobial resistance, and bloodstream infections. Understanding variability and trends in rates of hospital-level blood culture utilization may be important for interpreting these findings. Methods: We used clinical microbiology results and discharge data to identify monthly blood culture rates from US hospitals participating in the Premier Healthcare Database during 2012–2017. We included all discharges from months where a hospital reported at least 1 blood culture with microbiology and antimicrobial susceptibility results. Blood cultures drawn on or before day 3 were defined as admission cultures (ACs); blood cultures collected after day 3 were defined as a postadmission cultures (PACs). The AC rate was defined as the proportion of all hospitalizations with an AC. The PAC rate was defined as the number of days with a PAC among all patient days. Generalized estimating equation regression models that accounted for hospital-level clustering with an exchangeable correlation matrix were used to measure associations of monthly rates with hospital bed size, teaching status, urban–rural designation, region, month, and year. The AC rates were modeled using logistic regression, and the PAC rates were modeled using a Poisson distribution. Results: We included 11.7 million hospitalizations from 259 hospitals, accounting for nearly 52 million patient days. The median annual hospital-level AC rate was 27.1%, with interhospital variation ranging from 21.1% (quartile 1) to 35.2% (quartile 3) (Fig. 1). Multivariable models revealed no significant trends over time (P = .74), but statistically significant associations between AC rates with month (P < .001) and region (P = .003), associations with teaching status (P = .063), and urban-rural designation (P = .083) approached statistical significance. There was no association with bed size (P = .38). The median annual hospital-level PAC rate was 11.1 per 1,000 patient days, and interhospital variability ranged from 7.6 (quartile 1) to 15.2 (quartile 3) (Fig. 2). Multivariable models of PAC rates showed no significant trends over time (P = .12). We found associations between PAC rates with month (P = .016), bed size (P = .030), and teaching status (P = .040). PAC rates were not associated with urban–rural designation (P = .52) or region (P = .29). Conclusions: Blood culture utilization rates in this large cohort of hospitals were unchanged between 2012 and 2017, though substantial interhospital variability was detected. Although both AC and PAC rates vary by time of year and potentially by teaching status, AC rates vary by geographic characteristics whereas PAC rates vary by bed size. These factors are important to consider when comparing rates of bloodstream infections by hospital.

Funding: None

Disclosures: None