To the Editor—In 2019, the Centers for Disease Control and Prevention reported that >2.8 million antibiotic-resistant infections occur in the United States annually, with nearly 35,000 deaths as a result.¹ One method of minimizing the emergence of antimicrobial resistance is through antimicrobial stewardship. The Infectious Diseases Society of America has published guidelines on antimicrobial stewardship identifying potential interventions to guide appropriate antimicrobial prescribing, which includes selective or cascade reporting of antibiotic susceptibility data.^{Reference Barlam, Cosgrove and Abbo2} In cascade reporting, specific antibiotics in the susceptibility report are deliberately withheld from the view of clinicians when the organism is susceptible to more narrow-spectrum agents.^{Reference Coupat, Pradier and Degand3} Because of the limited data on clinical outcomes, cascade reporting is classified as a weak recommendation.^{Reference Barlam, Cosgrove and Abbo2}

In 2016, antimicrobial data at our 862-bed county hospital in Dallas, Texas, revealed piperacillin-tazobactam (PT) as the most utilized broad-spectrum gram-negative antimicrobial, with an average of 103 days of therapy per 1,000 patient days. During the same time, 11,306 isolates from the Enterobacterales family were identified from various sources, with 85% being Escherichia coli or Klebsiella spp. The objective of this retrospective study was to determine whether a cascade reporting system influenced the de-escalation of empiric PT in patients with E. coli and Klebsiella bacteremia due to a urinary source and subsequent effects on patient outcomes.

On September 7, 2017, the clinical microbiology laboratory implemented a cascading antibiotic algorithm for non–extended-spectrum β-lactamase–producing E. coli and Klebsiella isolates from blood and urine cultures. Susceptibility data for PT was suppressed if the organism was nonsusceptible to ceftriaxone. Isolates were considered susceptible to ceftriaxone at a minimum inhibitory concentration of ≤1 µg/mL, based on Clinical and Laboratory Standards Institute (CLSI) break points. Antibiotic and microbiologic data from September 7, 2016, through September 7, 2018, were retrospectively reviewed. Patients were included if they were at least 18 years old, received empiric PT, and had a monomicrobial non-ESBL E. coli or Klebsiella bacteremia with a concordant monomicrobial urine culture positive for the same organism collected within 24 hours of each other.

The primary end point was de-escalation from PT to a narrower agent, with the hypothesis that a greater percentage of patients would be de-escalated after the intervention. Narrower agents were defined as any narrower-spectrum β-lactam, ciprofloxacin, or sulfamethoxazole/trimethoprim. De-escalation was further stratified as early or late, if de-escalation occurred after release of either the blood or urine culture susceptibility or after release of both blood and urine cultures susceptibilities, respectively.

Of the 197 patients screened with E. coli or Klebsiella in blood and urine cultures receiving PT therapy during the study period, 103 met study criteria. The preimplementation group had 50 patients and the postimplementation group had 53 patients. Baseline characteristics of the study population were well matched regarding age, race, and gender. A similar percentage of patients were de-escalated in both groups: 45 (90%) versus 45 (85%) (P = .56). The most common reason identified for not de-escalating therapy was provider preference based on the patient’s clinical status. Ceftriaxone was the most utilized agent for initial de-escalation, followed by ciprofloxacin and sulfamethoxazole-trimethoprim. Empiric PT therapy was utilized for similar median durations in each group (46.2 vs 44.8 hours; P = .796). Both cohorts displayed rapid de-escalation of antibiotics once susceptibilities were available. Therapy was de-escalated a median of 5.6 hours after the release of susceptibility results in the preintervention group and 4.8 hours in the postintervention group (P = .506). Rates of early de-escalation were similar between both groups: 20 (44.4%) versus 22 (48.9%) (P = .673). Patients in the preintervention group had longer length of stays, although this was not statistically significant (6 vs 5 days; P = .058). The difference in rates of acute kidney injury [3 (6%) vs 9 (20%)] and C. difficile infection (CDI) [1 (2%) vs 0] were not statistically significant.

The impact of cascaded susceptibility reporting has been evaluated and has proven useful on several occasions,^{Reference Liao, Rhodes, Jandarov, DeVore and Sopirala4–Reference Langford, Seah and Chan6} but only 1 previous study has included frequency of de-escalation as the primary outcome. Johnson et al evaluated the impact of a cascaded susceptibility report on de-escalation from a broad-spectrum β-lactam to a narrower agent for gram-negative bacteremia. The cascade resulted in more patients being de-escalated to a narrower agent [15 (48%) vs 30 (71%); P = .43], although this was not statistically significant and did not influence length of stay, CDI, or reinitiation of a broad-spectrum agent within 7 days.^{Reference Johnson, Patel, King and Maslow5}

In our study, we found no significant difference in prescribing patterns after the implementation of cascade susceptibility reporting for Enterobacterales. De-escalation was observed in nearly 90% of patients in both pre- and postintervention groups, and median times to de-escalation were similar in both cohorts.

Because this was a retrospective study focusing on monomicrobial bacteremia secondary to a urinary source, providers may have felt more comfortable in quickly de-escalating, especially after urine culture results were available. Moreover, the overall sample size was small, due to a specific patient population that was included. Finally, providers were allowed to contact the clinical microbiology laboratory for release of suppressed antibiotic results.

Although we observed no difference, we believe that cascade reporting remains a viable intervention that could be added to an program’s arsenal; there may be benefit in patient populations not explored in our study. Additionally, this intervention requires little long-term maintenance, and no negative impact was observed. Further research is needed to better identify patient populations most impacted by a cascade algorithm and its overall effectiveness as a stewardship tool.