Computerized order-entry prompts using patient-specific MDRO risk safely reduced empiric extended-spectrum antibiotics for SSTI by ~28% without increasing ICU transfers or length of stay.

Background

Despite guideline support for standard-spectrum therapy in most nonpurulent, nonsurgical skin and soft tissue infections (SSTIs), extended-spectrum antibiotics are frequently prescribed to cover MRSA, Pseudomonas, and other multidrug-resistant organisms (MDROs). Overuse can cause harms, including C. difficile, nephrotoxicity, hepatotoxicity, allergies, and cost. This trial evaluated whether real-time, patient- and pathogen-specific MDRO risk estimates delivered via computerized provider order entry (CPOE) could improve empiric antibiotic selection.

Patients

• Adults ≥18 years hospitalized with SSTI in non-ICU settings across 92 HCA Healthcare hospitals in 15 U.S. states.
• Identified by discharge diagnosis codes for SSTI marked present-on-admission; excluded if transferred to ICU within 2 calendar days (primary analysis).
• Total: 118,562 admissions; mean age 58.0 years; 56.7% male.
• Baseline period: 57,837 patients (2019). Intervention period: 60,725 patients (2023).

Intervention

• Antibiotic stewardship bundle with CPOE prompts during the first 3 hospital days (empiric period):
  • Prompts triggered when an extended-spectrum agent was ordered and the patient’s absolute risk of Pseudomonas/MDR gram-negative SSTI was <10% (antibiotic-specific risk logic); one-click substitution to standard-spectrum (e.g., cefazolin).
  • Risk model: recursive partitioning using >60 EHR variables plus hospital-specific MDRO prevalence.
  • Education, monthly coaching, site visits/webinars, and clinician prescribing feedback reports.

Control

• Routine stewardship per national guidance (no CPOE prompts): local guidelines, required documentation of indications, prospective feedback to de-escalate, educational materials, quarterly coaching calls.

Outcome

• Primary: Empiric extended-spectrum antibiotic days of therapy (DOT) per 1000 empiric days (targeting Pseudomonas and/or MDR gram-negative bacteria) over hospital days 1–3.
• Secondary: Antipseudomonal DOT per 1000 empiric days; post hoc vancomycin DOT.
• Safety: Time to ICU transfer; hospital length of stay (LOS).

Study Design

• Cluster randomized clinical trial (hospital-level), pair-matched by baseline prescribing and case mix.
• 12-month baseline (2019), 5-month phase-in (excluded), 12-month intervention (2023).
• Difference-in-differences analysis using generalized linear mixed-effects models accounting for clustering.
• Trial registration: ClinicalTrials.gov NCT05423756.

Level of Evidence

Level I (cluster randomized clinical trial).

Follow up period

• Effectiveness outcomes assessed over the empiric period (hospital days 1–3).
• Safety outcomes assessed through hospital discharge (LOS capped at 14 days in analyses).

Results

Primary outcome

• Extended-spectrum DOT/1000 empiric days:
  • Routine stewardship: 511.5 (baseline) → 488.7 (intervention).
  • CPOE bundle: 496.2 (baseline) → 359.1 (intervention).
• Effect size: Rate ratio 0.72 (95% CI, 0.67–0.79), indicating a ~28% relative reduction with the CPOE bundle.
• Dichotomous utilization (receipt of any empiric extended-spectrum antibiotic):
  • Routine: 57.0% (baseline) → 56.0% (intervention).
  • CPOE: 55.4% (baseline) → 43.0% (intervention).
  • Approximate NNT: 9 to prevent one patient from receiving any empiric extended-spectrum antibiotic (based on an absolute difference-in-differences of 11.4%).

Secondary outcomes

• Antipseudomonal DOT/1000 empiric days:
  • Routine stewardship: 484.7 (baseline) → 495.1 (intervention).
  • CPOE bundle: 407.3 (baseline) → 338.7 (intervention).
  • Effect size: Rate ratio 0.68 (95% CI, 0.64–0.72).
• Post hoc vancomycin DOT/1000 empiric days:
  • Routine stewardship: 582.0 (baseline) → 566.0 (intervention).
  • CPOE bundle: 596.3 (baseline) → 524.7 (intervention).
  • Effect size: Rate ratio 0.90 (95% CI, 0.86–0.95).

Safety outcomes

• Hospital length of stay (mean [SD]): Routine 6.5 (3.8) days; CPOE 6.4 (3.8) days; hazard ratio 0.99 (90% CI, 0.95–1.04) — no evidence of inferiority.
• Days to ICU transfer (mean [SD]): Routine 6.3 (3.2) days; CPOE 6.3 (3.1) days; hazard ratio 1.14 (90% CI, 1.00–1.31) — no evidence of inferiority.

Additional findings

• Risk algorithm classified >95% of patients as low risk for Pseudomonas/MDR gram-negative SSTI; among those, <3% had subsequent Pseudomonas/MDRO-positive cultures.
• For MRSA, >76% classified low risk; among those, <6% had MRSA-positive cultures.
• Mechanisms of reduction included lower initial extended-spectrum selection (31.4% vs 40.9% in intervention vs routine hospitals during 2023) and 9.5% switching to standard-spectrum after seeing the prompt.

Limitations

• Positive skin cultures included regardless of specimen quality (colonization vs infection not always distinguishable).
• The <10% MDRO-risk threshold may be conservative; higher thresholds might retain safety and improve effectiveness.
• Conducted primarily in community hospitals; generalizability to other settings may vary.
• Concurrent prompts for abdominal infections could have influenced adoption (familiarity vs alert fatigue).
• Unable to account for physician-level effects.
• Effect of prompts cannot be fully separated from accompanying education/feedback, though rapid prescribing changes suggest a prominent prompt effect.

Funding

Funded by the National Institute of Allergy and Infectious Diseases (NIH) within the NIH Pragmatic Trials Collaboratory (U01 AI153005), with additional Collaboratory support (U24 AT009676) and in-kind support from HCA Healthcare. The content is the authors’ responsibility and does not necessarily represent official views of NIH or HCA Healthcare.

Citation

Gohil SK, Septimus E, Kleinman K, et al. Improving Empiric Antibiotic Selection for Patients Hospitalized With Skin and Soft Tissue Infection: The INSPIRE3 Skin and Soft Tissue Randomized Clinical Trial. JAMA Internal Medicine. 2025;185(6):680-691. doi:10.1001/jamainternmed.2025.0887. ClinicalTrials.gov: NCT05423756.