System and Method for Predicting Pulmonary and Other Complications Following Rib Fractures - (HJF 690-24)

Introduction

Scientists at the Henry M. Jackson Foundation (HJF), the Uniformed Services University of the Health Sciences (USUHS), and Duke University (Duke) have developed innovative systems, methods, and computational models designed to predict clinical complications such as acute lung injury (ALI), acute kidney injury (AKI), pneumonia, and respiratory distress following rib fractures. Utilizing serum biomarker data combined with machine learning algorithms, the invention provides a Clinical Decision Support Tool (CDST) that enables early risk stratification of patients with thoracic trauma, facilitating personalized treatment strategies and improving patient outcomes.

Applications and Advantages:

Early Risk Assessment: Enables clinicians to use serum biomarkers to predict(within 6 hours of blood collection), the likelihood of adverse complications from a rib fracture.
Improved Patient Outcomes: By identifying high-risk patients early, the decision tool system helps reduce morbidity, mortality, and unnecessary interventions.
Minimally-Invasive Testing: Uses serum samples for biomarker analysis, avoiding more invasive diagnostic procedures.
Robust Predictive Models: Employs advanced machine learning techniques such as random forest and Lasso regression models with demonstrated high accuracy (AUC up to 0.9).
Dynamic Immune Response Analysis: Accounts for temporal changes in proinflammatory and anti-inflammatory biomarkers post-injury to improve prediction accuracy.

Description of the Invention:

The invention arose from recognizing the complex interplay between rib fractures, immune responses, and subsequent clinical complications—a triad termed the “RFx-immune responses-complication” (RIC) system. Researchers analyzed data from 142 patients with rib fractures, including 431 serum samples assayed for 34 cytokines and chemokines.

Key findings include:

Changes in regulation for two major clusters of serum biomarker profiles
Correlation of biomarker clusters with groups of ribs and incidence of complications such as ALI, AKI, and pneumonia.
Development of predictive models using twelve selected biomarkers (eotaxin-3, IL-6, TNF-α, FGF basic, IFN-γ, G-CSF, IL-8, IP-10, MIG, MCP-4, IL-2Rα, and IL-16) representing both pro- and anti-inflammatory mediators.
Use of machine learning algorithms (random forest and Lasso regression) trained on initial assessment data (within 72 hours post-injury) and validated on follow-up data.
Demonstration of strong predictive performance with area under the curve (AUC) values ranging from approximately 0.78 to 0.89 for various complications.
Inclusion of additional clinical predictors such as rib fracture counts by location, age, sex, and Injury Severity Score (ISS) further improved model discrimination.
Implementation of post-training calibration techniques (Platt-scaling) to ensure reliable probability estimates.

The system can be implemented via software running on computing platforms, utilizing non-transitory computer-readable media storing executable instructions for training and applying the predictive models.