# P-905. Applying Encounter-level Risk-Adjustment for Standardized Antimicrobial Administration Ratios (SAAR): Is it meaningful?

## TL;DR

This study examines how adjusting for patient-specific factors changes how hospitals assess antimicrobial use, finding that most antimicrobial stewardship teams prefer the new method.

## Contribution

The study introduces encounter-level risk-adjusted standardized antimicrobial administration ratios (R-SAARs) and evaluates their impact on antimicrobial stewardship assessments.

## Key findings

- Encounter-level risk adjustment changed antimicrobial stewardship program (ASP) assessments and perceptions of antimicrobial use.
- Most ASPs preferred the newly derived R-SAARs over traditional SAARs.
- R-SAARs provided tighter distributions for commonly used antimicrobial groups compared to traditional SAARs.

## Abstract

Whether encounter-level risk-adjustment impacts antimicrobial stewardship program (ASP) assessments and perceptions of standardized antimicrobial administration ratio (SAAR) comparisons is unknown.

This retrospective analysis included patient encounter-level data among 50 US hospitals participating in the CDC Epicenters collaborative, including 2022 antimicrobial use (AU), demographics, and diagnosis and procedure categories. We developed two-staged, xgboost machine learning models on 2020-2021 training sets for adult and pediatrics (aged 1-17) to produce expected days of therapy (DOT) for each NHSN antimicrobial group and each 2022 encounter. Observed and expected DOT were aggregated to create ratios (O:E), summarized as “robust” SAARs (R-SAARs) for hospital and unit. Methods and interpretation guidance were provided to ASPs in documents and recorded videos. Two hospital-specific reports were delivered sequentially over a 3-month period in 2024. Part 1 included unadjusted hospital-level AU comparisons and 2017 unit-level SAARs using existing methods. Part 2 included newly derived R-SAARs. Hospital ASP leads were surveyed to assess their team’s response after each part.

Data included 2.5 million DOT and 5.6 million days present among 50 hospitals in 2022. Unit-level SAARs were calculated for 272 adult and 32 pediatric units. R-SAARs were derived for 474 adult and 45 pediatric units as a larger variety of unit types could be included. Data feedback included hospital- and unit-level O:E values and percentiles (Figures 1 and 2). R-SAARs produced extreme values for antimicrobial groups and units with infrequent AU, but tighter distributions for commonly used groups as compared to SAARs (Figure 3). Forty-five of 50 ASPs voluntarily participated in data feedback; half actively used SAARs at baseline. ASPs perceived Part 2 as being more helpful in developing goals but did not score Part 2 higher in other questions of value (Table 1). Most ASPs (88%) indicated preference for R-SAARs.

When encounter-level risk adjustment was applied, O:E and percentile estimates changed ASP assessments. ASPs preferred encounter-level risk adjustment when asked about desire for future AU comparisons.

Rebekah W. Moehring, MD, MPH, FIDSA, FSHEA, UpToDate, Inc.: Author Royalties Elizabeth Dodds Ashley, PharmD, MHS, HealthtrackRx: Advisor/Consultant|UpToDate, Inc.: Author Royalties Eili Klein, PhD, Beckman-Coulter: Grant/Research Support|Diasorin: Grant/Research Support|MeMed: Grant/Research Support Carlos Santos, MD, MPHS, UpToDate, Inc.: Author Royalties Michael J. Smith, M.D., M.S.C.E, Pfizer: Grant/Research Support