Daily Sepsis Research Analysis
Three papers advanced sepsis research today: a massive multicenter cohort linked peripartum maternal inflammatory markers to neonatal early-onset sepsis risk, a robust machine learning model predicted sepsis-induced coagulopathy with external validation and explainability, and a national surveillance study detailed device-associated bacteremia patterns in nursing homes to guide prevention.
Summary
Three papers advanced sepsis research today: a massive multicenter cohort linked peripartum maternal inflammatory markers to neonatal early-onset sepsis risk, a robust machine learning model predicted sepsis-induced coagulopathy with external validation and explainability, and a national surveillance study detailed device-associated bacteremia patterns in nursing homes to guide prevention.
Research Themes
- Risk stratification and predictive analytics in sepsis
- Maternal–neonatal inflammatory pathways and early-onset sepsis
- Long-term care epidemiology and device-associated bacteremia prevention
Selected Articles
1. The predictive value of maternal inflammation markers for neonatal early-onset sepsis.
In a 380,455-pair multicenter cohort, markedly elevated maternal CBC-derived indices (WCC, ANC, NLR, PLR) within 24 hours before delivery were strongly associated with neonatal early-onset sepsis, with very high likelihood ratios, especially at ≥30 weeks' gestation. Earlier elevations (24–12 hours pre-delivery) conferred even higher risk, supporting temporal causality.
Impact: This study quantifies maternal inflammatory thresholds with high likelihood ratios for neonatal EOS at population scale, offering actionable metrics for risk stratification.
Clinical Implications: Integrating maternal CBC indices and CRP measured hours before delivery into neonatal EOS risk calculators could refine early evaluation, guide targeted laboratory workup, and support antibiotic stewardship in at-risk infants.
Key Findings
- Among 380,455 dyads, 460 neonates developed early-onset sepsis.
- Maternal WCC, ANC, NLR, and PLR elevations within 24 hours pre-delivery were strongly associated with neonatal EOS, especially at ≥30 weeks' gestation.
- Within 12 hours pre-delivery, LRs were very high (e.g., ANC >30×10^9/L: LR up to 92.9 in term neonates).
- Earlier marker elevations (24–12 hours pre-delivery) were associated with even higher risk, indicating a temporal relationship.
Methodological Strengths
- Very large multicenter cohort with precise temporal windowing around delivery
- Clear reporting of likelihood ratios for multiple CBC-derived markers across gestational strata
Limitations
- Retrospective design may entail residual confounding and selection bias
- Generalizability outside Hong Kong and to settings with different obstetric practices is uncertain
Future Directions: Prospective validation and integration into clinical decision support systems, combining maternal markers with obstetric and neonatal factors to optimize EOS risk algorithms.
PURPOSE: Maternal infection is a known risk factor for neonatal early-onset sepsis (EOS). However, the relationship between maternal inflammatory makers near delivery and neonatal EOS remains unclear. This study aimed to evaluate these associations and explore whether maternal blood parameters could contribute to EOS risk assessment strategies. METHODS: In this retrospective multicenter study in Hong Kong, we included mother‑neonate pairs where the mother underwent a complete blood count (CBC) or C-reactive protein (CRP) test between 48 h before and 72 h after delivery from January 1, 2006 to December 31, 2017. We assessed associations between maternal white blood cell count (WCC), absolute neutrophil count (ANC), neutrophil-to-lymphocyte ratio (NLR), platelet-to-lymphocyte ratio (PLR), and CRP with subsequent neonatal EOS. RESULTS: Among 380,455 mother‑neonate dyads, 460 neonates developed EOS. Markedly elevated maternal WCC, ANC, NLR, and PLR within 24 h before delivery were significantly associated with neonatal EOS, particularly in neonates born at ≥30 weeks' gestation. Within 12 h before delivery, the estimated likelihood ratios (LRs) for WCC > 30 × 10^9/L, ANC > 30 × 10^9/L, NLR > 50, and PLR > 800 were 23.8, 73.1, 45.7, and 45.6, respectively, in neonates born at 30-36 weeks, and 36.4, 92.9, 20.9, and 17.5, respectively, in term neonates. LRs were even higher when markers were elevated earlier (within 24 to 12 h) before delivery, suggesting a temporal relationship between maternal inflammation and neonatal EOS risk. CONCLUSIONS: Although maternal sepsis biomarkers are insufficient to diagnose neonatal EOS independently, their elevation is associated with increased risk and may support clinical risk stratification, particularly when occurring well before delivery.
2. A machine learning model for robust prediction of sepsis-induced coagulopathy in critically ill patients with sepsis.
Using 15,479 MIMIC-IV records plus external validation, a gradient boosting model with 17 features predicted sepsis-induced coagulopathy with AUCs of 0.773 (train), 0.730 (internal), and 0.966 (external). SHAP analysis highlighted total bilirubin, RDW, SBP, heparin, and BUN as key contributors, outperforming SOFA-based prediction.
Impact: Provides an explainable, externally validated ML tool for early SIC risk prediction, potentially enabling proactive monitoring and anticoagulation strategies.
Clinical Implications: Embedding the GBM model into ICU EHRs could flag high SIC risk to prompt intensified coagulation monitoring, judicious anticoagulant use, and targeted lab testing, potentially reducing hemorrhagic and thrombotic complications.
Key Findings
- Gradient boosting machine with 17 features achieved AUC 0.773 (train), 0.730 (internal), and 0.966 (external) for SIC prediction.
- Model outperformed SOFA-based prediction approaches.
- SHAP identified total bilirubin, RDW, SBP, heparin exposure, and BUN as leading predictors.
- SIC incidence in derivation cohort was 38.9% (6,036/15,479).
Methodological Strengths
- Large derivation cohort with internal and external validation
- Model explainability via SHAP facilitates clinical interpretation
Limitations
- Retrospective design and potential overfitting (noted by high external AUC) without prospective impact evaluation
- Reliance on database coding and ISTH-derived SIC criteria may affect generalizability across ICUs
Future Directions: Prospective, multi-ICU implementation studies to assess calibration, clinical workflow integration, fairness across subgroups, and outcome impact of model-guided care.
INTRODUCTION: Sepsis-induced coagulopathy (SIC) is a common disease in patients with sepsis. It denotes higher mortality rates and a poorer prognosis in these patients. This study aimed to develop a practical machine learning (ML) model for the prediction of the risk of SIC in critically ill patients with sepsis. METHODS: In this retrospective cohort study, patients were extracted from the Medical Information Mart for Intensive Care IV (MIMIC-IV) database and the Inner Mongolia Autonomous Region People's Hospital database. Sepsis and SIC were defined based on the Sepsis-3 criteria and the criteria developed based on the International Society of Thrombosis and Haemostasis (ISTH), respectively. We compared nine ML models using the Sequential Organ Failure Assessment (SOFA) score in terms of SIC prediction ability. Optimal model selection was based on the superior performance metrics exhibited by the model on the training dataset, the internal validation dataset, and the external validation dataset. RESULTS: Of the 15,479 patients in MIMIC-IV included in the final cohort, a total of 6,036 (38.9%) patients developed SIC during sepsis. We selected 17 features to construct ML prediction models. The gradient boosting machine (GBM) model was deemed optimal as it achieved high predictive accuracy and reliability across the training, internal, and external validation datasets. The areas under the curve of the GBM model were 0.773 (95%CI = 0.765-0.782) in the training dataset, 0.730 (95%CI = 0.715-0.745) in the internal validation dataset, and 0.966 (95%CI = 0.938-0.994) in the external validation dataset. The Shapley Additive Explanations (SHAP) values illustrated the prediction results, indicating that total bilirubin, red cell distribution width (RDW), systolic blood pressure (SBP), heparin, and blood urea nitrogen (BUN) were risk factors for progression to SIC in patients with sepsis. CONCLUSIONS: We developed an optimal and operable ML model that was able to predict the risk of SIC in septic patients better than the SOFA scoring models.
3. Incidence trends and epidemiology of invasive device-associated bacteremia in French nursing home residents, 2020-2024: Insights from the SPIADI Prospective Multicenter Study.
A prospective national surveillance across 1,233 facilities identified 2,117 nursing home–acquired bacteremias with stable incidence (0.009/1,000 resident-days). Urinary tract sources predominated, invasive devices were implicated in 20% (mostly urinary catheters), Enterobacterales and S. aureus were common, and MDR organisms occurred in 15.2%.
Impact: Provides large-scale, contemporary epidemiology of device-associated bacteremia in long-term care, pinpointing urinary catheters as key prevention targets and quantifying MDR burden.
Clinical Implications: Strengthens rationale for catheter stewardship bundles in nursing homes (e.g., indication review, aseptic maintenance, early removal), enhanced surveillance of Enterobacterales and S. aureus, and stewardship plans addressing MDR risk.
Key Findings
- Recorded 2,117 nursing home–acquired bacteremias across 1,233 facilities (2020–2024), with incidence 0.009 per 1,000 resident-days and stable trend.
- Urinary tract (52.1%) and respiratory tract (11.9%) were main sources; invasive devices implicated in 20% of cases, predominantly urinary catheters.
- Enterobacterales (64.0%) and Staphylococcus aureus (14.6%) were most frequent; MDR organisms in 15.2% of cases.
- Intravascular device–related bacteremia was rare (38 cases), highlighting different prevention priorities.
Methodological Strengths
- Prospective, national multicenter surveillance covering 1,233 institutions over four years
- Source attribution and pathogen distribution including MDR characterization
Limitations
- Surveillance design may under-detect cases depending on reporting practices and diagnostics
- Limited clinical granularity (e.g., sepsis severity, comorbidity adjustment) restricts risk modeling
Future Directions: Evaluate the impact of catheter stewardship bundles on bacteremia incidence and MDR patterns; link surveillance to patient-level outcomes and sepsis management pathways.
Healthcare-associated bacteremia is associated with increased morbidity and mortality. In nursing homes, these infections remain under-documented. We investigated invasive device-associated bacteremia in residents. We analyzed bacteremias acquired in nursing homes using data from a national surveillance program conducted between 2020 and 2024, involving 1,233 French healthcare institutions. A total of 2,117 bacteremias acquired in the nursing home were recorded. The main sources of infection were the urinary tract (52.1%) and the respiratory tract (11.9%). An invasive device was involved in 20.0% of cases, primarily urinary catheters (386 cases), while bacteremia related to intravascular devices was rare (38 cases). Enterobacterales (64.0%) and Staphylococcus aureus (14.6%) were the most frequently identified pathogens, with multidrug-resistant bacteria detected in 15.2% of nursing home-acquired bacteremias. The incidence rate was 0.009 per 1,000 resident-days, remaining stable over the study period. The study highlights the burden of bacteremias in nursing homes and underscores the importance of targeted infection prevention measures, particularly in relation to urinary catheter management, and long-term intravascular central lines.