Daily Sepsis Research Analysis
Risk-stratified nutrition in sepsis, antimicrobial resistance in urinary bacteraemia among cancer patients, and pre-dialysis risk prediction for life-threatening dialysis events emerged as today’s most impactful themes. A post hoc multicentre analysis supports ≥60% early energy delivery in high-risk sepsis (by mNUTRIC) to reduce 28-day mortality. Large cohort data underscore MDR burden and recurrence in urinary bacteraemia in oncology, while a pragmatic model anticipates severe intradialytic eve
Summary
Risk-stratified nutrition in sepsis, antimicrobial resistance in urinary bacteraemia among cancer patients, and pre-dialysis risk prediction for life-threatening dialysis events emerged as today’s most impactful themes. A post hoc multicentre analysis supports ≥60% early energy delivery in high-risk sepsis (by mNUTRIC) to reduce 28-day mortality. Large cohort data underscore MDR burden and recurrence in urinary bacteraemia in oncology, while a pragmatic model anticipates severe intradialytic events using pre-HD variables.
Research Themes
- Risk-stratified nutrition for sepsis
- Antimicrobial resistance and recurrence in urinary bacteraemia among cancer patients
- Pre-dialysis prediction of life-threatening hemodialysis complications
Selected Articles
1. Early energy delivery and 28-day mortality in critically ill patients with sepsis: Post hoc analysis of a multicenter cluster-randomised controlled trial.
In a post hoc analysis of 1162 ICU sepsis patients, early energy delivery ≥60% of a 25 kcal/kg (ideal body weight) target reduced 28-day mortality in high mNUTRIC patients, with no benefit in low-risk patients. Findings support risk-based nutrition targets rather than uniform goals.
Impact: Provides large, multicentre evidence that individualized energy targets based on nutritional risk can influence mortality in sepsis. Bridges a gap between guideline ambiguity and pragmatic thresholds.
Clinical Implications: Adopt mNUTRIC-guided early energy delivery in sepsis: aim for ≥60% of energy target during the first week in high-risk patients, while avoiding unnecessary escalation in low-risk patients.
Key Findings
- Among 1162 sepsis ICU patients, 28-day mortality was 15.7%.
- Optimal early energy thresholds differed by risk: 100% of target for low-risk and 60% for high-risk (mNUTRIC ≥5).
- In high-risk patients, achieving ≥60% of target reduced 28-day mortality (HR 0.588, 95% CI 0.388–0.891); no benefit in low-risk patients.
- Spline analysis suggested decreasing mortality with increasing energy delivery in high-risk patients.
Methodological Strengths
- Large multicentre cohort embedded within a cluster-RCT framework with prespecified stratification by mNUTRIC
- Robust time-to-event analyses including Cox models, subgroup tests, and restricted cubic splines
Limitations
- Post hoc observational nature with potential residual confounding and no randomization to energy targets
- Energy target fixed at 25 kcal/kg (ideal body weight) without indirect calorimetry; generalizability across ICUs may vary
Future Directions: Prospective randomized trials stratified by mNUTRIC comparing personalized energy (by indirect calorimetry) and protein dosing thresholds; evaluate functional outcomes and infection rates.
2. Multidrug resistance and recurrence in urinary bacteraemia among cancer patients.
In 561 episodes of urinary bacteraemia among cancer patients, Gram-negative bacilli predominated (87.3%), MDR-GNB accounted for 19.6%, and 23.4% received inappropriate empiric therapy. Recurrence occurred in 14% with a simple predictive score, and 30-day mortality was 15.3% (bUTI-related 10.7%), with septic shock, absence of fever, and carbapenemase-producing Enterobacterales linked to higher related mortality.
Impact: Quantifies MDR burden, treatment delays, and recurrence in a vulnerable oncology population, and provides a pragmatic recurrence score to inform prevention and empiric therapy.
Clinical Implications: Optimize empiric coverage for high-risk patients (instrumentation, prior admissions), incorporate the recurrence score for follow-up planning, and implement MDR stewardship and source control to reduce mortality and recurrence.
Key Findings
- 561 bUTI episodes in 478 oncology patients; 62.2% had tumor-related urinary tract involvement and 59.4% had instrumentation.
- Gram-negative bacilli caused 87.3% of cases; MDR-GNB in 19.6%; inappropriate empiric therapy in 23.4%.
- Recurrence occurred in 14.0% with a simple predictive score identifying high-risk patients.
- 30-day mortality was 15.3% (bUTI-related 10.7%); absence of fever, septic shock, and carbapenemase-producing Enterobacterales were linked to higher related mortality.
Methodological Strengths
- Large single-centre cohort over 11 years with detailed clinical and microbiological data
- Multivariable regression to identify independent risk factors and development of a simple recurrence score
Limitations
- Single-centre retrospective design limits generalizability and may include residual confounding
- Recurrence score requires external validation; appropriateness of empiric therapy judged retrospectively
Future Directions: External validation and calibration of the recurrence score across centers; interventional studies testing stewardship and device-related strategies to reduce MDR and recurrence.
3. Machine Learning-Based Prediction of Life-Threatening Complications During Hemodialysis in Hospitalized Patients With Poor General Conditions.
Using 739 inpatients receiving hemodialysis, a pre-dialysis variable model predicted sudden life-threatening events during or within 24 hours after HD with an AUC of 0.889. Predictive factors included emergency admission, recent surgery, shorter HD vintage, higher heart rate, hypoalbuminemia, and elevated CRP.
Impact: Provides a practical pre-dialysis risk tool for acute care settings, enabling proactive monitoring and resource allocation in high-risk inpatients, including those treated for sepsis.
Clinical Implications: Use pre-HD variables to triage monitoring intensity, adjust ultrafiltration and vasopressor readiness, and coordinate perioperative/sepsis care around HD sessions.
Key Findings
- Among 739 HD inpatients, 17 (2.3%) experienced sudden events (fatal arrhythmia, refractory hypotension, or respiratory arrest) during or within 24 hours after HD.
- A logistic regression model using 23 pre-HD covariates achieved an AUC of 0.889.
- Key predictors included emergency hospitalization (present in 71% of events), recent surgery (76%), shorter HD history, higher pre-HD heart rate, lower albumin, and higher CRP.
Methodological Strengths
- Comprehensive pre-HD variable set (51 candidates) with clear, clinically relevant composite outcomes
- Model performance reported with discrimination (AUC 0.889) and interpretable covariates
Limitations
- Single-centre retrospective design with small number of events (n=17) raising overfitting risks; no external validation
- Described as machine learning but implemented as stepwise logistic regression; calibration and clinical utility not prospectively tested
Future Directions: External validation across centres, prospective impact analysis, and integration with real-time monitoring to trigger preventive interventions during HD.