CoChat

The DOI matches its Crossref record; 33 of the 33 references it cites were found in a registry or an index; 18 of the 30 sources its claims cite were checked for support.

Day-4 antipseudomonal and anti-MRSA de-escalation was not associated with higher mortality in critically ill patients with sepsis

Open access
Zhang, Yuhang, Liao, Guoying, Song, Jianli, Liao, Yinsheng, Zhang, Lulu · 2026 · Frontiers in Medicine

Abstract

Background: Broad empiric antibiotics are often appropriate at the start of intensive care unit (ICU) sepsis care; however, continued antipseudomonal or anti-methicillin-resistant Staphylococcus aureus (MRSA) coverage may no longer be justified by the day-4 reassessment. We examined antibiotic de-escalation at this clinical decision point. Materials and methods: We emulated two parallel target trials in Medical Information Mart for Intensive Care IV (MIMIC-IV) v3.1 and conducted a supportive multicenter analysis in the eICU Collaborative Research Database (eICU-CRD) v2.0. The strict MIMIC-IV cohort included patients with ICU sepsis, early receipt of broad-spectrum antibiotics, survival and continued observation through 96 h, available culture information before time zero, no recorded multidrug-resistant organism evidence, and no prespecified exclusions. Time zero was 72 h after antibiotic initiation; exposure was classified during 72–96 h, and follow-up began at 96 h. Trial A compared anti-MRSA de-escalation with continuation, and Trial B compared antipseudomonal de-escalation with continuation. Propensity scores, overlap weighting, and bootstrap confidence intervals were used. Results: The strict MIMIC-IV cohort included 5,762 patients. In Trial B, 1,569 patients were de-escalated, and 3,070 continued antipseudomonal coverage. De-escalation was associated with lower 30-day mortality (24.1% vs. 29.0%; risk difference, −4.91 percentage points; 95% confidence interval, −7.27 to −2.23), lower in-hospital death, lower stage 2–3 acute kidney injury, and more death-adjusted antibiotic-free days. In Trial A, 3,041 patients were de-escalated, and 1,728 continued anti-MRSA coverage; there was no evidence of higher 30-day mortality (26.3% vs. 27.8%; risk difference, −1.44 percentage points; 95% confidence interval, −3.75 to 1.11), and de-escalation was associated with more antibiotic-free days. In eICU-CRD, Trial B showed a directionally consistent association with lower hospital mortality (10.2% vs. 12.7%; cluster-bootstrap risk difference, −2.42 percentage points; 95% confidence interval, −3.87 to −0.87). Conclusion: Day-4 de-escalation was not associated with higher mortality and was accompanied by reduced antibiotic exposure. The lower mortality and kidney-injury estimates in Trial B are exploratory because de-escalation may mark early recovery that is incompletely captured in structured data. The eICU-CRD analysis provides directional support rather than formal external validation. Prospective evaluation is warranted.

Overview

CoChat Summary

TL;DR

In ICU sepsis patients who survived to a day-4 reassessment, antipseudomonal and anti-MRSA de-escalation was not associated with higher mortality and was associated with more antibiotic-free days. Lower mortality and AKI estimates, especially for antipseudomonal de-escalation, were exploratory and may reflect residual confounding by early recovery.

Key findings

  • In MIMIC-IV Trial B (antipseudomonal), 1,569 de-escalated vs. 3,070 continued patients had lower 30-day mortality (24.1% vs. 29.0%; RD −4.91 percentage points, 95% CI −7.27 to −2.23), lower in-hospital death (17.6% vs. 22.8%; RD −5.19, 95% CI −7.65 to −2.88), lower 7-day KDIGO stage 2–3 AKI (17.8% vs. 21.8%; RD −4.06, 95% CI −6.33 to −1.73), and more death-adjusted antibiotic-free days at day 14 (8.57 vs. 6.21 days; MD 2.35, 95% CI 2.06 to 2.65) and day 28 (18.40 vs. 15.27 days; MD 3.12, 95% CI 2.49 to 3.72).
  • In MIMIC-IV Trial A (anti-MRSA), 3,041 de-escalated vs. 1,728 continued patients showed no evidence of higher 30-day mortality (26.3% vs. 27.8%; RD −1.44 percentage points, 95% CI −3.75 to 1.11), lower in-hospital death (20.2% vs. 23.1%; RD −2.91, 95% CI −5.15 to −0.60), lower 7-day KDIGO stage 2–3 AKI (20.4% vs. 22.5%; RD −2.12, 95% CI −4.25 to −0.02), and more death-adjusted antibiotic-free days at day 14 (7.91 vs. 5.85 days; MD 2.06, 95% CI 1.80 to 2.34) and day 28 (17.40 vs. 14.96 days; MD 2.44, 95% CI 1.83 to 3.02).
  • In the supportive eICU-CRD analysis, Trial B hospital mortality was 10.2% with de-escalation vs. 12.7% with continuation (cluster-bootstrap RD −2.42 percentage points, 95% CI −3.87 to −0.87; RR 0.81, 95% CI 0.70 to 0.93); Trial A hospital mortality was 10.7% vs. 12.1% (RD −1.41, 95% CI −3.08 to 0.41).
  • Same-class restart after de-escalation occurred in 287/3,041 Trial A patients (9.4%) and 157/1,569 Trial B patients (10.0%); restarted patients had higher 30-day mortality than non-restarted de-escalated patients (Trial A 34.8% vs. 23.1%; Trial B 29.9% vs. 22.8%).

Methods

  • Target trial emulation using MIMIC-IV v3.1 for the strict primary analysis and eICU-CRD v2.0 for a supportive multicenter analysis. Time zero was 72 h after the first qualifying broad-spectrum antibiotic; exposure was classified during 72–96 h, and follow-up began at 96 h. Trial A compared anti-MRSA de-escalation vs. continuation; Trial B compared antipseudomonal de-escalation vs. continuation.
  • MIMIC-IV eligibility required adult ICU sepsis, early qualifying broad-spectrum antibiotic, survival and observation through 96 h, culture information before time zero, no recorded MDRO evidence, and no prespecified exclusions; the strict cohort included 5,762 patients. eICU-CRD screened adult first ICU encounters with early broad-spectrum antibiotics and a sepsis-compatible union definition, without the strict culture-availability/no-MDRO criteria.
  • Propensity scores with overlap weighting were used; balance was assessed by absolute SMDs, with values below 0.10 considered acceptable. Binary outcomes were weighted risks and risk differences; continuous antibiotic-free days were weighted means and mean differences. CIs used bootstrap procedures (500 MIMIC-IV replicates; 300 eICU-CRD replicates with hospital-cluster bootstrap for the primary uncertainty estimate).

Limitations

  • Stated: treatment was not randomized, so residual confounding by clinical improvement, source-control confidence, and treatment selection is likely and may account for part of the Trial B mortality and AKI associations. Microbiology reporting time could not always be observed directly; storetime was used as a proxy, and residual look-ahead bias cannot be fully excluded. Infection-source classification was not available through a validated framework, so differences across infection syndromes could not be fully assessed.
  • Stated: the strict MIMIC-IV cohort excluded patients without pre-time-zero culture information, chronic dialysis patients, severely immunosuppressed patients, and fluoroquinolone-only initial broad-spectrum exposure, narrowing generalizability. Death was not modeled as a competing event for AKI, and discharge or differential laboratory measurement could affect renal-event ascertainment. The eICU-CRD analysis was supportive rather than strict external validation because microbiology and medication completeness varied across hospitals. Bootstrap uncertainty estimates may contain Monte Carlo variation, and antibiotic restart was descriptive within de-escalated patients rather than a comparative treatment-failure endpoint.
  • Evident from the text: the lower mortality and AKI estimates, particularly in Trial B, are explicitly described as exploratory and not causal; the eICU-CRD absolute mortality risks are not directly comparable with MIMIC-IV because eICU-CRD used hospital mortality in a broader union cohort while MIMIC-IV used 30-day mortality in a strict no-MDRO landmark cohort. Trial A and Trial B cohorts overlapped, so estimates were not treated as statistically independent and no formal between-trial comparison was performed.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.3389/fmed.2026.1924203

Cites

33 of 33 references verified0 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Verification

Claims the paper makes

10 Valid · 0 Not valid · 3 Not checked · 101 Own results and interpretations

30 more claims

Create a free account to see all 114 claims and every source verdict.

See all claims — free account
  • Probably not a substantial claim. 62
  • It may state two separate claims as one. 43

Create a free account to see these claims.

Citations

Cites

Show all 33 references

Metadata

Authors
Zhang, Y., Liao, G., Song, J., Liao, Y., & Zhang, L.
Affiliations
  • Second Affiliated Hospital of Chongqing Medical University (CN)
  • Department of Anesthesiology, Zigong Fourth People's Hospital
Corresponding author
Zhang, Lulu
Venue
Frontiers in Medicine
Published
2026-08-27
Volume
13
Pages
1924203-1924203
ISSN
2296-858X (ISSN-L)
OpenAlex
W7204446328
Cited by
0
References
33
FWCI
0.00
Type
journal article
Language
en
Open access
Locations
  • Frontiers in Medicine (publishedVersion, cc-by)
  • PubMed (publishedVersion)
  • PubMed Central (publishedVersion, cc-by)
Topics
  • Health Sciences › Medicine › Critical Care and Intensive Care Medicine › Sepsis Diagnosis and Treatment
  • Health Sciences › Medicine › Infectious Diseases › Antibiotic Use and Resistance
  • Health Sciences › Medicine › Pharmacology (medical) › Antibiotics Pharmacokinetics and Efficacy
Keywords
antibiotic de-escalation, sepsis, critically ill patients, broad-spectrum antibiotics, 30-day mortality, acute kidney injury, MIMIC-IV, eICU-CRD, antibiotic stewardship

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

The claim verdicts this report counts are experimental: their accuracy is not yet measured.

A language model drew the claims from the paper on 2026-10-02, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).

A second model labelled each claim on 2026-10-02 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).

Cited sources are checked with a language model’s help (gpt-6-sol).

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review

How to cite

Cite CoChat for the checks on this page, and the paper for its own findings.

This page
CoChat. (2026). Citation check: Day-4 antipseudomonal and anti-MRSA de-escalation was not associated with higher mortality in critically ill patients with sepsis. https://app.cochat.ai/papers/10.3389/fmed.2026.1924203
The paper
Zhang, Y., Liao, G., Song, J., Liao, Y., & Zhang, L. (2026). Day-4 antipseudomonal and anti-MRSA de-escalation was not associated with higher mortality in critically ill patients with sepsis. Frontiers in Medicine. https://doi.org/10.3389/fmed.2026.1924203

Data sources

Bibliographic data: Semantic Scholar, OpenAlex, Crossref.

Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).

Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.