Prognostic Models for Disease Progression and Outcomes in Chronic Obstructive Pulmonary Disease: A Systematic Review...

Area of Science:

• Pulmonary Medicine
• Medical Informatics
• Biostatistics

Background:

• Chronic obstructive pulmonary disease (COPD) poses a growing global health challenge.
• Existing prognostic models for COPD progression and exacerbation risk show inconsistent rigor and performance.
• Machine learning (ML) offers potential but its application in COPD prognostication is limited.

Purpose of the Study:

• To systematically review prognostic models for COPD disease progression and exacerbation risk in adults.
• To evaluate the methodological quality, generalizability, and predictive performance of existing models.
• To compare traditional regression-based methods with ML techniques in COPD prognostication.

Main Methods:

• Systematic literature search of PubMed and Embase for prognostic models in COPD (1-5 year window).
• Inclusion of models for mortality, exacerbations, and hospitalizations; appraisal using PROBAST.
• Descriptive summary of model performance and meta-analysis of discrimination (c-statistic) for externally validated models.

Main Results:

• 193 prognostic models from 87 studies were included; only 7% utilized ML.
• All-cause mortality and severe exacerbations were the most common outcomes studied.
• Meta-analysis for exacerbation models was hindered by heterogeneity; BODE index performed best for mortality (pooled c 0.75).
• Over 40% of studies exhibited a high risk of bias.

Conclusions:

• Meaningful quantitative synthesis was limited to externally validated mortality models due to substantial heterogeneity.
• Regression-based models predominate, with limited adoption of ML in COPD prognostication.
• Persistent methodological limitations necessitate the development of more robust, validated models for complex patient data.