Dynamic prediction of early antibiotic treatment failure in ischemic stroke-associated pneumonia: development and internal validation of a 48-h landmark model

Background Existing stroke-associated pneumonia scores estimate pneumonia occurrence rather than early response after antibiotics have started. We aimed to develop and internally validate a baseline model and a 48-h landmark model for first antibiotic treatment failure between 48 and 72 h after treatment initiation.Methods This single-center, retrospective cohort study included consecutive adults with acute ischemic stroke-associated pneumonia treated with empirical systemic antibiotics from December 1, 2021 to December 1, 2025. All included patients were event-free at completion of the 48-h landmark assessment. The primary outcome was first treatment failure by the nominal 72-h reassessment, comprising efficacy-driven antibiotic escalation, persistent clinical instability with evidence of uncontrolled pneumonia, or pneumonia-related death. Ridge logistic regression with a fixed inverse penalty strength (C = 1.0) was used, and 500 bootstrap resamples estimated optimism-corrected performance.Results Among 199 patients, 102 (51.3%) experienced treatment failure. At baseline, mechanical ventilation before antibiotic initiation and a higher platelet-to-lymphocyte ratio were associated with failure. In the 48-h model, each 20 mg/L increase in C-reactive protein change was associated with greater failure odds (adjusted odds ratio 1.89, 95% confidence interval 1.52–2.34), as were each five-unit increase in neutrophil-to-lymphocyte ratio change (1.98, 1.25–3.13) and escalation of oxygen support (4.56, 1.09–19.01). The optimism-corrected area under the receiver operating characteristic curve increased from 0.697 at baseline to 0.889 at 48 h. Treatment failure was associated with longer hospitalization, higher hospital cost, and longer antibiotic exposure.Conclusions In this single-center retrospective study, a structured 48-h reassessment combining routinely available inflammatory trajectories and oxygen-support change improved short-horizon prediction of antibiotic treatment failure. The model should prompt clinical review rather than automatic antibiotic escalation; multicenter external validation and prospective impact evaluation are required before clinical implementation.

Authors

Institutions

Publication Details

Journal
Annals of Medicine
Published
2026-11-12
DOI
https://doi.org/10.1080/07853890.2026.2753847
Primary Topic
Pneumonia and Respiratory Infections
Type
article
Field-Weighted Citation Impact
0.00

Funders

Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Dynamic prediction of early antibiotic treatment failure in ischemic stroke-associated pneumonia: development and internal validation of a 48-h landmark model

李思思, Xuebin Qin, Mingjiao Tang, Hao Li et al.
Annals of Medicine
Pneumonia and Respiratory Infections
article

Dynamic prediction of early antibiotic treatment failure in ischemic stroke-associated pneumonia: development and internal validation of a 48-h landmark model

李思思, Xuebin Qin, Mingjiao Tang, Hao Li, Chunhua Lai, Cuiling Tang, Haili Feng, Jufeng Qiu, Quanchan Li, Tengxiang Zhu, Jinsong Li, Yuqing Li
article en

Abstract

Background Existing stroke-associated pneumonia scores estimate pneumonia occurrence rather than early response after antibiotics have started. We aimed to develop and internally validate a baseline model and a 48-h landmark model for first antibiotic treatment failure between 48 and 72 h after treatment initiation.Methods This single-center, retrospective cohort study included consecutive adults with acute ischemic stroke-associated pneumonia treated with empirical systemic antibiotics from December 1, 2021 to December 1, 2025. All included patients were event-free at completion of the 48-h landmark assessment. The primary outcome was first treatment failure by the nominal 72-h reassessment, comprising efficacy-driven antibiotic escalation, persistent clinical instability with evidence of uncontrolled pneumonia, or pneumonia-related death. Ridge logistic regression with a fixed inverse penalty strength (C = 1.0) was used, and 500 bootstrap resamples estimated optimism-corrected performance.Results Among 199 patients, 102 (51.3%) experienced treatment failure. At baseline, mechanical ventilation before antibiotic initiation and a higher platelet-to-lymphocyte ratio were associated with failure. In the 48-h model, each 20 mg/L increase in C-reactive protein change was associated with greater failure odds (adjusted odds ratio 1.89, 95% confidence interval 1.52–2.34), as were each five-unit increase in neutrophil-to-lymphocyte ratio change (1.98, 1.25–3.13) and escalation of oxygen support (4.56, 1.09–19.01). The optimism-corrected area under the receiver operating characteristic curve increased from 0.697 at baseline to 0.889 at 48 h. Treatment failure was associated with longer hospitalization, higher hospital cost, and longer antibiotic exposure.Conclusions In this single-center retrospective study, a structured 48-h reassessment combining routinely available inflammatory trajectories and oxygen-support change improved short-horizon prediction of antibiotic treatment failure. The model should prompt clinical review rather than automatic antibiotic escalation; multicenter external validation and prospective impact evaluation are required before clinical implementation.

Annals of MedicineVol. 58(1)
Guilin Medical University (CN), Affiliated Hospital of Guilin Medical College (CN)
Natural Science Foundation of Guangxi Province
Good health and well-being
Pneumonia and Respiratory Infections
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon