Korean standardized antimicrobial administration ratio in long-term care hospitals: a benchmark tool

Antimicrobial stewardship in long-term care hospitals (LTCHs) is increasingly important; however, standardized benchmarking tools for antimicrobial use in these settings are lacking. The Korean Standardized Antimicrobial Administration Ratio for long-term care hospitals (K-SAAR-LTCHs) was developed to benchmark antimicrobial use in these facilities.

We used 2021 National Health Insurance claims data from 1432 LTCHs (387,211 inpatients, including untreated patients) in Korea. Data were split into training (80%) and test (20%) sets at the hospital level for model validation, with 5-fold hospital-level cross-validation within the training set. The main outcome measure was antimicrobial consumption expressed as days of therapy (DOT) per 1000 patient-days. For the fixed-rate cohort, a two-part Hurdle Gamma model addressed zero-inflation; for fee-for-service (FFS) cohorts (pneumonia, septicemia), a Gamma generalized linear model with log link and length of stay as offset was used. Variable selection used stepwise regression with the Akaike Information Criterion (AIC). Predictive performance was evaluated using mean absolute percentage error (MAPE) and calibration metrics (overall observed-to-expected ratio, intercept, slope). The standardized antimicrobial administration ratio (SAAR) was calculated as observed-to-predicted antimicrobial use for each LTCH. For hospital-level benchmarking, 95% funnel plot control limits were calculated using the Byar approximation.

Antimicrobial use under FFS was substantially higher than under the fixed-rate system (1579.3 vs. 108.3 DOT per 1000 patient-days), reflecting that FFS episodes primarily comprised diagnosis-defined infectious conditions. In the FFS pneumonia and septicemia cohorts, delirium, fever, oxygen therapy, and stage 3 pressure ulcer were consistently associated with higher antimicrobial use, although many rate ratios were modest in magnitude. Hospital-level calibration was strongest for the total-antimicrobial FFS models, whereas broad-spectrum models showed greater between-hospital dispersion. Median hospital-level SAAR values were close to 1.0 in the FFS pneumonia and septicemia cohorts. Funnel plot analysis identified hospitals above the upper control limits in 0.8-1.2% for total antimicrobials, while broad-spectrum antimicrobials showed wider variation, with 4.6-5.7% of hospitals falling below the lower control limits. The fixed-rate hurdle model was developed and evaluated separately, but hospital-level benchmarking was less stable because antimicrobial use was sparse and highly right-skewed.

The K-SAAR-LTCH framework provides a risk-adjusted approach for monitoring antimicrobial utilization in Korean LTCHs. Hospital-level benchmarking was primarily demonstrated for diagnosis-specific FFS pneumonia and septicemia cohorts, while fixed-rate reimbursement episodes require further refinement before routine benchmarking use. The framework may inform antimicrobial utilization surveillance and targeted stewardship review, but it should not be interpreted as a direct measure of prescribing appropriateness.