ENTERPRISE AI ANALYSIS
PneumoScore: Risk Prediction Model for 90-Day Mortality After Lung Resection
This study developed 'PneumoScore,' a logistic regression model to predict 90-day mortality after lung cancer resection. Utilizing data from the Brazilian Lung Cancer Registry (2002-2024), the model incorporates age, sex, predicted postoperative FEV1%, ASA > IV, coronary artery disease, cerebrovascular disease, congestive heart failure, pneumectomy, thoracotomy, and extended resection. PneumoScore demonstrated excellent performance with an AUROC of 0.84 and strong calibration, outperforming existing benchmarks for this critical, yet underexplored, outcome. This model enhances preoperative risk assessment and clinical decision-making, supporting improved patient outcomes in thoracic surgery.
Executive Impact at a Glance
Lung resection, while the gold-standard for early-stage lung cancer, still carries significant mortality risks. Existing preoperative risk prediction models often rely on outdated datasets, have insufficient predictive variables, and primarily focus on 30-day or in-hospital mortality, failing to capture a substantial proportion of adverse events occurring between 31 and 90 days post-surgery. This leads to inadequate preoperative evaluation and suboptimal referral of high-risk cases to alternative therapies.
0.84
AUROC
4.40%
90-Day Mortality Rate
0.033
Brier Score
The study proposes 'PneumoScore,' a novel risk prediction model developed using machine learning (logistic regression) on a comprehensive, multicentric database. It specifically targets 90-day all-cause mortality, offering a more complete assessment of short-term postoperative risk. By incorporating a broader set of clinically relevant predictors and continuous performance updates, PneumoScore aims to provide dynamic, accurate, and interpretable risk estimates to guide clinical decision-making and patient counseling.
Deep Analysis & Enterprise Applications
Select a topic to dive deeper, then explore the specific findings from the research, rebuilt as interactive, enterprise-focused modules.
0.84
AUROC (Area Under Receiver Operating Characteristic Curve)
PneumoScore demonstrated an AUROC of 0.84 (0.77–0.91), indicating excellent discriminative performance. This means it accurately distinguishes between patients who will and will not experience 90-day mortality after lung resection.
0.99
AUPR (Area Under Precision-Recall Curve)
The AUPR of 0.99 (0.98–1.00) further confirms the model's high precision and recall, especially valuable in scenarios with imbalanced class distributions like mortality prediction.
| Metric | PneumoScore | Thoracoscore | EuroLung2 |
|---|---|---|---|
| AUROC |
|
|
|
| AUPR |
|
|
|
| Brier Score |
|
|
|
| Brier Skill Score (BSS) |
|
|
|
0.033
Brier Score
A low Brier Score of 0.033 (0.022–0.045) indicates excellent calibration, meaning the predicted probabilities closely match the observed outcomes.
PneumoScore Development Process
Data Extraction (Brazilian Lung Cancer Registry)
→
Data Cleaning & Imputation
→
Feature Selection (Clinical Relevance & Univariable Analysis)
→
Algorithm Evaluation (Logistic Regression, Random Forest, XGBoost, SVM)
→
Model Selection (Logistic Regression for superior balance)
→
Final Model (PneumoScore)
| Predictor | PneumoScore | Thoracoscore | EuroLung2 |
|---|---|---|---|
| Age |
|
|
|
| Male sex |
|
|
|
| ppoFEV1% |
|
|
|
| ASA Classification |
|
|
|
| Coronary artery disease |
|
|
|
| Congestive heart failure |
|
||
| Cerebrovascular disease |
|
|
|
| Pneumectomy |
|
|
|
| Thoracotomy |
|
|
|
| Extended resection |
|
|
Why Logistic Regression Outperformed Advanced ML
Problem: The study initially evaluated various machine learning algorithms (Random Forest, XGBoost, SVM) alongside logistic regression. Surprisingly, logistic regression consistently achieved the best balance between discrimination, calibration, and fold-to-fold stability.
Solution: This outcome is attributed to the inherent class imbalance in clinical scenarios with relatively low event rates (e.g., mortality prediction). Simpler models like logistic regression are less prone to distorting probability estimates and compromising calibration in such contexts, especially when advanced techniques like SMOTE (Synthetic Minority Over-sampling Technique) are avoided to maintain clinical validity.
Impact: Choosing logistic regression for PneumoScore ensures greater interpretability and easier integration into existing perioperative workflows, making it a more practical and reliable tool for clinicians, despite its perceived simplicity.
90
Day Mortality Focus
PneumoScore's focus on 90-day mortality provides a more comprehensive assessment of short-term postoperative risk, capturing a substantial proportion of events (28% of deaths in this study) that occur between 31 and 90 days post-surgery, which is often missed by 30-day or in-hospital models.
Enhancing Preoperative Decision-Making
Problem: Existing risk models often provide insufficient accuracy or are based on outdated data, leading to suboptimal preoperative assessment for lung cancer patients. This can result in high-risk patients undergoing surgery when alternative therapies might be more appropriate.
Solution: PneumoScore offers an accurate, dynamically updated risk prediction for 90-day mortality. By providing clinicians with reliable, interpretable risk estimates, it facilitates informed discussions with patients and supports better stratification for surgical vs. non-surgical treatment paths.
Impact: Improved patient selection for lung resection, potentially reducing adverse outcomes and guiding high-risk patients towards safer, effective alternatives like stereotactic body radiation therapy, thereby enhancing overall patient care and resource allocation.
2001
Patients Included in Study
The model was developed using data from 2001 patients from the Brazilian Lung Cancer Registry, a multicentric database, ensuring a robust and diverse dataset for training and validation.
Estimate Your Potential AI Impact
Input your enterprise details to see the projected annual savings and reclaimed operational hours with AI-driven predictive analytics like PneumoScore.
Projected Annual Savings
$0
Reclaimed Operational Hours
0
Your AI Implementation Roadmap
A phased approach ensures seamless integration and maximum impact. Our experts guide you every step of the way.
Phase 1: Discovery & Strategy
Initial consultations to understand your specific challenges, data infrastructure, and strategic objectives. Develop a tailored AI roadmap.
Phase 2: Data Integration & Model Customization
Securely integrate relevant datasets. Customize PneumoScore or develop new models to align with your unique patient population and clinical workflows.
Phase 3: System Deployment & Training
Deploy the AI solution within your existing EMR/digital health systems. Comprehensive training for clinical and IT staff.
Phase 4: Monitoring, Validation & Optimization
Continuous monitoring of model performance, external validation, and iterative optimization to ensure ongoing accuracy and relevance in an evolving clinical landscape.
Ready to Transform Your Surgical Outcomes?
Schedule a personalized strategy session with our AI specialists to explore how PneumoScore can be tailored for your institution.
Ready to Get Started?
Book Your Free Consultation.
Let's Discuss Your AI Strategy!
Lets Discuss Your Needs
Select a Date
Sun
Mon
Tue
Wed
Thu
Fri
Sat