Goto Michihiko, Cho Hyunkeun, Merchant James A, Perencevich Eli N, Goetz Matthew B, Marra Alexandre R, Alexander Bruce, Hanks Tyler C, Beck Brice F, Richards Christopher, Hernandez David M, Livorsi Daniel J
Center for Access and Delivery Research and Evaluation, Iowa City Veterans Affairs (VA) Health Care System, Iowa City.
Department of Internal Medicine, University of Iowa Carver College of Medicine, Iowa City.
JAMA Netw Open. 2025 Jun 2;8(6):e2514989. doi: 10.1001/jamanetworkopen.2025.14989.
The Centers for Disease Control and Prevention offers a standardized antimicrobial administration ratio (SAAR) as an evaluation metric for inpatient antibiotic use through rankings and peer comparisons (ie, benchmarking). However, the SAAR model only accounts for facility- and unit-level factors without considering the hierarchical nature of the health care data, and it does not directly reflect patient-level factors or stewardship efforts to avoid overly broad-spectrum therapy.
To examine the use of antimicrobial use risk adjustment methods and choice of basic metrics (eg, days of therapy [DOT] and days of antimicrobial spectrum coverage [DASC], which do not and do consider antimicrobial spectrum, respectively) in hospital benchmarking.
DESIGN, SETTING, AND PARTICIPANTS: This retrospective cohort study was conducted using data from 117 acute care hospitals within the Veterans Health Administration (VHA) system. All patients admitted between January 1, 2021, and December 31, 2023, were included.
Monthly antibiotic use was measured with 2 basic metrics and risk adjustment models created using baseline data for 2021 to 2022. Hospitals were benchmarked for 2023 use via 3 methods: (1) unadjusted comparison, (2) risk adjustment with hospital- and unit-level factors with single-level negative binomial regression models (method 1, similar in approach to the SAAR), and (3) risk adjustment with hospital-, unit-, and patient-level factors with hierarchical zero-inflated negative binomial regression models (method 2).
This study included 736 810 patients (median age, 70 [IQR, 61-76] years; 94.7% male). There was wide variability in unadjusted antibiotic use among hospitals (median, 477 [IQR, 420-523] DOT per 1000 days present [DP]; and median, 3115 [IQR, 2739-3602] DASC per 1000 DP). Risk adjustments with methods 1 and 2 resulted in moderate ranking changes, but there were only weak correlations between benchmarking results by the 2 methods (τB = 0.43 for DOT and 0.44 for DASC). The choice of basic metrics with or without consideration of antimicrobial spectrums (DOT vs DASC) had a modest correlation after risk adjustment (τB = 0.84).
In this cohort study of the nationwide VHA system, there were substantial differences in risk-adjusted benchmarking results between models with only hospital- and unit-level factors and models with hospital-, unit-, and patient-level factors. Future studies should evaluate whether these models with higher content validity also have better construct validity and can inform hospitals and stewardship programs about their objective performance compared with other programs.
疾病控制与预防中心提供了一种标准化抗菌药物使用比率（SAAR），作为通过排名和同行比较（即基准对比）来评估住院患者抗生素使用情况的指标。然而，SAAR模型仅考虑了机构和科室层面的因素，未考虑医疗保健数据的层级性质，也没有直接反映患者层面的因素或避免过度广谱治疗的管理措施。
研究抗菌药物使用风险调整方法的应用以及医院基准对比中基本指标（如治疗天数[DOT]和抗菌谱覆盖天数[DASC]，分别不考虑和考虑抗菌谱）的选择。
设计、设置和参与者：这项回顾性队列研究使用了退伍军人健康管理局（VHA）系统内117家急性护理医院的数据。纳入了2021年1月1日至2023年12月31日期间入院的所有患者。
每月抗生素使用情况通过2个基本指标以及使用2021年至2022年基线数据创建的风险调整模型进行测量。2023年医院通过3种方法进行基准对比：（1）未调整比较；（2）使用单水平负二项回归模型对医院和科室层面因素进行风险调整（方法1，在方法上与SAAR类似）；（3）使用分层零膨胀负二项回归模型对医院、科室和患者层面因素进行风险调整（方法2）。
本研究纳入了736810名患者（中位年龄70岁[四分位间距，61 - 76岁]；94.7%为男性）。各医院未调整的抗生素使用情况差异很大（每1000天住院日[DP]的中位DOT为 477[四分位间距，420 - 523]；每1000 DP的中位DASC为3115[四分位间距，2739 - 3602]）。方法1和方法2的风险调整导致排名有适度变化，但两种方法的基准对比结果之间只有微弱的相关性（DOT的τB = 0.43，DASC的τB = 0.44）。风险调整后，是否考虑抗菌谱的基本指标选择（DOT与DASC）有适度的相关性（τB = 0.84）。
在这项针对全国VHA系统的队列研究中，仅包含医院和科室层面因素的模型与包含医院、科室和患者层面因素的模型在风险调整后的基准对比结果存在显著差异。未来的研究应评估这些具有更高内容效度的模型是否也具有更好的结构效度，以及与其他项目相比，能否为医院和管理项目提供有关其客观绩效的信息。
