A data mining based clinical decision support system for survival in lung cancer

Area of Science:

• Oncology
• Clinical Decision Support Systems
• Biostatistics

Background:

• Lung cancer patient outcomes are crucial for treatment planning.
• Existing clinical guidelines offer a framework but may lack personalized predictive power.
• Routine clinical data holds potential for enhanced prognostic accuracy.

Purpose of the Study:

• To develop and validate a clinical decision support system (CDSS) for predicting overall survival in lung cancer patients.
• To compare the predictive performance of the CDSS against established clinical guidelines.
• To assess the CDSS's utility in complementing guideline-based care.

Main Methods:

• Prospective, multicenter data from 543 lung cancer patients (2013-2017) with 1167 variables were utilized.
• Development of the CDSS employed Data Mining techniques, including XGBoost and Generalized Linear Models algorithms.
• Comparative analysis of prediction accuracy between the CDSS and clinical guidelines was performed using receiver-operating characteristic curves (AUCs).

Main Results:

• The CDSS demonstrated superior predictive performance for overall survival compared to clinical guidelines, with AUCs generally above 0.70 versus below 0.70 for guidelines.
• Statistically significant improvements in AUCs were observed for the CDSS across most comparisons (p < 0.05).
• The highest AUCs (> 0.90) were achieved for predicting survival in small cell lung cancer patients using the CDSS.

Conclusions:

• The developed CDSS shows significant potential to enhance the prediction of survival in lung cancer patients.
• The CDSS can support physicians in providing evidence-based management recommendations.
• This tool facilitates individualized prognosis discussions, improving patient care.