Discussion

This antimicrobial stewardship initiative introduces a novel approach for feedback and peer comparisons. To our knowledge, this is the first publication on semi-automated prescriber report cards that integrate multiple, objective data streams, including spectrum scores and patient acuity. Existing studies on inpatient antibiotic peer comparisons have not yet incorporated this level of automation, focused on ID prescribers, or addressed concerns that prescribing differences may be driven by patient acuity. ^{Reference Fowler, Webber and Cooper3–Reference Elligsen, Walker and Pinto5} Interestingly, two of our highest antibiotic utilizers, in both DASC/encounter and DOT/encounter, had a lower average case mix index than the median. Prior literature suggests a positive correlation between the case mix index and antibiotic utilization, and this may highlight opportunities for targeted feedback to these outlier prescribers. ^{Reference Kuster, Ruef and Bollinger6}

One key strength is the integration of automation into the workflow. Semi-automation reduces the time required for data processing and generating report cards. It also enhances consistency and reproducibility. Furthermore, it reduces manual errors, ensuring that any observed changes in prescribing habits result from true changes in prescriber patterns rather than errors in data collection or manipulation.

Another strength of this method is its objective assessment nature. Spectrum scores quantify antibiotic coverage, and the case mix index is routinely used at institutions nationwide and standardized in measurement. ^{Reference Kakiuchi, Livorsi and Perencevich7–Reference Mendez, Harrington, Christenson and Spellberg9} Objective feedback is provided with specific data on key antimicrobials that influence the prescriber’s score. This approach is also beneficial from an antibiotic stewardship perspective because it considers all prescribed antibiotics.

Our approach has several limitations. First, we analyzed prescription data as DOT/encounter versus DASC/encounter. This was a practical choice and requires further validation. Normalization through encounters can enable fair comparisons in peer feedback comparisons, given ID prescriber patient volumes at our institutions vary widely. Traditional metrics such as 1,000 patient days may be less meaningful at the individual prescriber level. Utilizing a surrogate of patient volume has also been utilized in other stewardship literature evaluating individualized prescriber peer comparisons. ^{Reference Song, Wilson and Bej10} More importantly, our intention in sharing this initiative is not to advocate for a specific measure of antibiotic consumption but rather to present an adaptable and scalable approach.

Second, this method relies on administrative data, so we may miss encounters if antibiotics were continued in a setting that was not captured or if an ID prescriber saw a patient without prescribing any antimicrobials. Additionally, if patient care transfers occur without any changes in anti-infective orders, duration and spectrum may be attributed to the incorrect ID prescriber. It remains challenging to adequately attribute the spectrum and duration of therapy to the recommending consultant. Unfortunately, our database system is limited in the granularity of data to distinguish this. We maintained our methodology as we were aiming for an approximation of antibiotic utilization to facilitate a peer comparison visualization and do anticipate that actionable outlier trends would still emerge. Furthermore, the inclusion of 1 years’ worth of prescription data into the analysis helps partially overcome these limitations.

Ultimately, we aim for outlier scores to shift toward an inlier center, reflecting improved prescribing practices over time. We plan to continue to analyze the impact of these reports in conjunction with other antimicrobial stewardship initiatives on antibiotic utilization.

We present a semi-automated antimicrobial stewardship initiative designed to generate individualized prescriber feedback report cards, which is traditionally a resource-intensive task. The model’s scalable nature facilitates adaptability across various prescriber groups and facility types. This model will assist ASP teams to optimize stewardship workflows and support antimicrobial stewardship goals.