Performance of Risk Models for Antimicrobial Resistance in Adult Patients With Sepsis
Authors' conclusion
In this cohort study of 39 893 patients with sepsis, variable model performance was associated with prevalence rates of antimicrobial resistance rather than patient case mix.
Study context
- Population
- Adults with sepsis (blood cultures, 4 days of antibiotics, organ dysfunction) at 10 acute-care hospitals in Missouri and Illinois, 2016–2021
- Participants
- 39,893 adults · 85,238 sepsis episodes
- Exposure
- Community onset vs Hospital onset
- Outcome
- Performance (AUROC, AUPRC) of deep learning models predicting ceftriaxone-resistant but cefepime-susceptible (RS) and ceftriaxone- and cefepime-resistant (RR) GNB across hospitals and patient subgroups
Methodological detailsmulticentre, deep learning prediction model, Analyses completed in April 2024
- Multicentre
- Deep learning prediction model
- Analyses completed in April 2024
Main finding
As reported in the abstract
Model performance across hospitals and subgroups
Correlation of area under precision recall curve (AUPRC) with prevalence of resistant gram-negative bacilli (GNB) (R)
0.79
P = .001
Observational data: associations, not effects.
Further findings
Prevalence of RS and RR GNB · events/value = RS episodes
RR Community onset 1.8% (796 episodes) · Hospital onset 3.9% (1626 episodes)
Previous antimicrobial resistance, community-onset sepsis (subgroup)
- RR episodes in patients with previous resistance
- 375 (47.1%)
- RS episodes in patients with previous resistance
- 420 (25.2%)
- SS episodes in patients with previous resistance
- 3483 of 40 744 (8.5%)
- P
- P < 0.001
- Previous infections and antibiotic exposure associated with risk of resistant GNB
Original publication
Performance of Risk Models for Antimicrobial Resistance in Adult Patients With Sepsis. JAMA Netw Open. Published 4 November 2024. doi:10.
Composed by AI from the published abstract and checked against it. How Briefs are made →