Prior to implementation, the antimicrobial stewardship pharmacist educated CSPs on the use of these tools, including simulated orders for mock sepsis alerts to ensure competency. A copy of the algorithm and nomogram were emailed to all CSPs and posted in a prominent location in the pharmacy.
As part of continuous performance improvement efforts of the ISC, a retrospective cohort study was conducted through chart review on patients at the Lexington VAHCS with an order for a sepsis alert in the ED from December 3, 2018 to May 31, 2020 to assess the accuracy of the CSPs’ antibiotic selection and dosing. Patients were excluded if they had a vancomycin allergy or if the ED practitioner ordered antibiotics prior to the CSPs placing orders. Patients could be included more than once in the study if they had sepsis alerts placed on different dates.
The primary outcomes were CSPs’ accuracy in antimicrobial selection with the antibiotic selection algorithm and vancomycin dosing nomogram. The antibiotic selection was deemed accurate if the appropriate antibiotic regimen was selected based on allergy status and previous cultures as directed in the algorithm. The vancomycin dose was considered accurate if the dose chosen was appropriate based on the patient’s weight at the time of ED presentation. Secondary outcomes included time to administration of antibiotics from ED presentation as well as time to antibiotics administration from sepsis alert initiation. Time of administration was considered the time the antibiotics were scanned in the bar code medication administration (BCMA) system.
Descriptive statistics were used with data presented as percentages for nominal data and median as IQR for continuous data. In accordance with our facility’s project assessment process, this project was determined not to constitute human subjects research; therefore, this QI project did not require review by the institutional review board.
Results
Between December 3, 2018 and May 31, 2020, 160 sepsis alerts were ordered by ED practitioners. Of the 160 patients, 157 were included in the final data analysis. Two patients were excluded due to vancomycin allergy, and 1 patient because the physician ordered antibiotics prior to pharmacist order entry. The population was largely composed of male patients (98%) with a median age of 72 years (Table 2).
Of 157 sepsis alerts, the antibiotic selection algorithm was used appropriately in 154 (98%) instances (Table 3). Chart reviews were performed in instances of antimicrobial selection different from the algorithm. Of the 3 patients who received antibiotics not consistent with the algorithm, 1 patient without a history of ESBL-producing organisms in their culture history received meropenem instead of piperacillin/tazobactam. Another patient without a penicillin allergy received cefepime (plus metronidazole ordered separately from the ED practitioner) instead of piperacillin/tazobactam, and the third patient received piperacillin/tazobactam instead of meropenem despite a culture history of ESBL-producing organisms. Vancomycin dose was appropriate according to the weight-based nomogram in 147 cases (94%). The median time to administration of first dose antibiotics was 39 minutes after the sepsis alert order was placed and 96 minutes after initial ED presentation.
Discussion
This study found extremely high rates of accuracy among CSPs for both the antibiotic selection algorithm (98%) and the vancomycin dosing nomogram (94%). Moreover, analysis of the 3 patients who received antibiotics that were inconsistent with the algorithm revealed that 2 of these patients arguably still received adequate empiric coverage, increasing the percentage of patients receiving appropriate empiric antibiotics to 99.4%. Similarly, chart review of 10 patients who received vancomycin doses that deviated from the nomogram revealed that in at least 3 cases, patients were likely given correct vancomycin doses based on the patient’s last known weight. However, when actual current weights were recorded soon after admission, the updated weights rendered the initial vancomycin loading dose incorrect when this analysis was performed. Thus, the adherence to the vancomycin dosing nomogram is higher than it appears.