Daily Sepsis Research Analysis
Three studies advance sepsis diagnostics and stewardship: an AI-assisted plasmonic coffee-ring biosensor achieves pg/mL protein detection (including procalcitonin) within 12 minutes; photoplethysmography features from routine pulse oximetry identify early deterioration in ED sepsis patients; and neonatal blood culture time-to-positivity data support stopping antibiotics at 36 hours in well-appearing infants with negative cultures.
Summary
Three studies advance sepsis diagnostics and stewardship: an AI-assisted plasmonic coffee-ring biosensor achieves pg/mL protein detection (including procalcitonin) within 12 minutes; photoplethysmography features from routine pulse oximetry identify early deterioration in ED sepsis patients; and neonatal blood culture time-to-positivity data support stopping antibiotics at 36 hours in well-appearing infants with negative cultures.
Research Themes
- Point-of-care diagnostics and biosensing for sepsis biomarkers
- Physiological waveform analytics to predict early deterioration
- Antibiotic stewardship in neonatal sepsis evaluations
Selected Articles
1. Plasmonic coffee-ring biosensing for AI-assisted point-of-care diagnostics.
This study introduces a dual-droplet, plasmonic coffee-ring biosensor coupled with a deep neural network to quantify proteins from smartphone images within 12 minutes at pg/mL sensitivity. It detected procalcitonin and other clinically relevant proteins across five orders of magnitude and outperformed lateral flow assays by >100-fold in saliva.
Impact: Provides a novel, low-cost, ultra-sensitive platform for point-of-care detection of sepsis biomarkers (e.g., procalcitonin) with AI-enabled readout. It could democratize early sepsis screening in resource-limited settings.
Clinical Implications: If clinically validated in blood or plasma, this platform could enable rapid triage and early antibiotic stewardship decisions using procalcitonin at the point of care, including ambulances and low-resource clinics.
Key Findings
- Dual-droplet coffee-ring biosensor achieved ~3 pg/mL limits of detection within 12 minutes.
- Quantitative readout via a deep neural model using smartphone photos enabled robust classification over five orders of magnitude.
- Procalcitonin (sepsis), SARS-CoV-2 N, CEA, and PSA were all detected; sensitivity surpassed lateral flow immunoassays by >100-fold in saliva.
Methodological Strengths
- Cross-analyte validation across four clinically relevant proteins
- Integration of generative and convolutional deep learning for smartphone-based quantification
Limitations
- No prospective clinical validation in sepsis patients; matrices beyond saliva (e.g., blood) not demonstrated in this report
- Environmental variability (humidity/temperature) may affect droplet evaporation and pattern formation
Future Directions: Prospective ED/ICU trials comparing to standard immunoassays, validation in whole blood/plasma, robustness testing across environments, and regulatory usability studies.
2. Pulse oximetry beyond oxygen saturation: Early waveform characteristics in sepsis patients with adverse outcomes - A proof-of-concept study.
In a secondary analysis of 576 ED patients with early sepsis, specific PPG waveform features (lower SPA/DPA, shorter PI, lower APG b/a) identified those who deteriorated within 48 hours, whereas standard blood pressures did not differ. PPG may provide an early warning signal for clinical deterioration.
Impact: Leverages ubiquitous pulse oximetry to detect early hemodynamic compromise before conventional vitals diverge, supporting scalable risk stratification.
Clinical Implications: Incorporating PPG features into ED triage could identify at-risk sepsis patients earlier, prompting timely monitoring and interventions even when blood pressure appears stable.
Key Findings
- Among 576 early sepsis ED patients, 9.7% met the 48-hour composite endpoint (ICU admission and/or death).
- Deteriorating patients had lower SPA (1218 vs. 1490 AU), lower DPA (462 vs. 621 AU), shorter PI (0.540 vs. 0.609 s), and lower APG b/a (-0.881 vs. -0.802), all p < 0.001.
- Standard BP metrics (SBP/DBP/MAP) did not differ, while HR was higher and strongly correlated with PI (ρ = -0.921).
Methodological Strengths
- Prospective data source (Acutelines biobank) with standardized waveform capture
- Rich feature engineering of PPG/APG enabling physiologic interpretation
Limitations
- Secondary analysis without external validation or full predictive model performance reporting
- Potential confounding from motion artifacts and limited comparison to established risk scores
Future Directions: Develop and externally validate predictive algorithms integrating PPG features, and test clinical impact in prospective triage/intervention trials.
3. Time to positive blood cultures in neonatal sepsis evaluations.
In 151 neonatal blood cultures, the median time to positivity was 17 hours. Nearly all gram-negatives were positive by 24 hours and most gram-positives by 36 hours, supporting antibiotic discontinuation at 36 hours in well-appearing infants with negative cultures.
Impact: Provides organism-specific time-to-positivity data that directly inform neonatal antibiotic stewardship and reduce unnecessary exposure.
Clinical Implications: NICUs can consider stopping empiric antibiotics at 36 hours in well-appearing infants with negative cultures and no source, potentially reducing antimicrobial resistance and adverse effects.
Key Findings
- Overall median time to blood culture positivity was 17 hours (IQR 12–23) in 151 neonatal cultures.
- 97.9% of gram-negative organisms were positive by 24 hours; 93.6% of gram-positive by 36 hours.
- All early-onset sepsis cultures were positive within 24 hours; most late-onset within 36 hours.
Methodological Strengths
- Organism-stratified time-to-positivity analysis with clinically actionable thresholds
- Focus on a defined NICU population with clear endpoints
Limitations
- Single-center retrospective design with modest sample size
- Did not assess clinical outcomes after implementing a 36-hour stop rule
Future Directions: Prospective multicenter validation of a 36-hour antibiotic stop protocol and integration with rapid diagnostics to further shorten unnecessary therapy.