838 - Drivers of empiric carbapenem use: How important is history of extended-spectrum beta-lactamase (ESBL) infection?

Background: CARs are first line agents for serious infections caused by ESBL producers. Likelihood of developing subsequent ESBL infection is unknown. In patients (pts) with a history (hx) of ESBL positive (ESBLP) culture, empiric therapy with a CAR has become common in hospitals. The purpose of this study was to evaluate the microbiology of subsequent infections (SI) among pts with hx of ESBLP culture and determine risk factors associated with ESBLP SI that may justify an empiric CAR.

Methods: This retrospective observational study was conducted at a multicenter health system. The electronic medical record (EMR) was used to generate a report of all E. coli (EC) or K. pneumoniae (KP) ESBLP cultures during 2017, an analogous report was generated for ESBL-negative (ESBLN) EC or KP. These were termed index cultures (IC). Pts were randomly selected from each report until 200 total pts were enrolled. Inpatients, outpatients, and all culture specimens were included. Pts with an ESBLP culture prior to 2017 were excluded. The EMR was reviewed up to 1 year after the IC. Pt and culture characteristics were recorded. The primary outcome was proportion of pts who developed an ESBLP SI. Risk factors associated with ESBLP SI were determined. Relapsed infection (same site, same bacteria) that occurred within 2 weeks of the IC was excluded.

Results: 200 pts were included, 100 with ESBLP IC and 100 with ESBLN IC. The mean age was 58 years, 84% were female, and 69% were outpatients. 86% of IC were EC and 86% were urine specimens. Within 1 year of IC, 100 pts (50%) developed a SI. 22 of these were ESBLP, 43 were ESBLN, and 35 had no or negative culture. The mean time since IC for ESBLP SI and ESBLN SI was 85 (26-226) days and 140 (15-363) days, respectively (p=0.014). When comparing time to SI, 21 (96%) ESBLP and 26 (61%) ESBLN occurred < 6 months after IC (p=0.003). Among SI with culture data (n= 65), the number of ESBLP SI was higher if the IC was ESBLP (22 vs 0, p< 0.001). Incidence of ESBLP or ESBLN SI in all pts with an ESBLP IC was similar (22 vs 18, p=0.428). Factors associated with ESBLP SI were hx of ESBLP IC, male gender, and time between IC and SI.Table 1. Index Culture Characteristics of Culture Positive Subsequent Infections
Figure 1. Cumulative rate of ESBL-positive SI in 180 days (6 months) following IC
Table 2. Univariate Analysis of Patient Characteristics Comparing ESBL-positive and ESBL-negative Culture Positive Subsequent Infections


Conclusion: Hx of positive culture for ESBL-producing EC or KP is associated with SI caused by ESBLP EC or KP. Pts presenting < 6 months after ESBLP IC are at increased risk for ESBLP SI, justifying empiric CAR therapy.