Development and validation of claims-based algorithms to identify interstitial lung disease among Japanese patients with cancer in routine clinical practice using real-world data sources - Report - MDSpire
Conexiant’s news site is now MDSpire News. Learn more
Advertisement
Creation and assessment of claims-based algorithms for detecting interstitial lung disease in Japanese cancer patients utilizing real-world data in standard clinical settings
Clinical Report: Claims-Based Algorithms for Detecting ILD in Japanese Cancer Patients
Overview
This study developed and validated claims-based algorithms to identify interstitial lung disease (ILD) in Japanese cancer patients using machine learning. The LASSO reduced model demonstrated high specificity but limited sensitivity.
Background
Interstitial lung disease (ILD) poses significant safety concerns, particularly in Japan where its incidence is notably high. Accurate identification of ILD is crucial for managing cancer treatment risks.
Data Highlights
Model
Sensitivity
Specificity
Positive Predictive Value (PPV)
LASSO Reduced Model
33.5%
99.3%
76.7%
External Database Validation
19.8%
99.4%
65.5%
Key Findings
The study analyzed data from 13,601 eligible patients, identifying 415 as ILD cases.
The LASSO reduced model was the highest performing, with a sensitivity of 33.5% and specificity of 99.3%.
Key variables for ILD identification included confirmed ILD diagnosis codes and biomarkers such as Krebs von den Lungen-6.
External validation showed comparable performance with a sensitivity of 19.8% and specificity of 99.4%.
The algorithm's positive predictive value supports its use in confirming ILD cases in retrospective studies.
Clinical Implications
The algorithm has limited sensitivity, high specificity, and positive predictive value.
Conclusion
The development of claims-based algorithms for ILD detection is reported.
