Statistical methods for predicting mortality in patients diagnosed with acute pulmonary embolism

Area of Science:

• Cardiology
• Pulmonology
• Medical Informatics

Background:

• Pulmonary embolism (PE) risk stratification is crucial for patient management.
• Accurate mortality prediction models are needed for acute PE.
• Computed tomographic pulmonary angiography (CTPA) is a primary diagnostic tool.

Purpose of the Study:

• Develop and validate novel mortality risk prediction models for acute PE.
• Identify key clinical predictors of 30-day and 90-day mortality.
• Evaluate the performance of adaptive least absolute shrinkage and selection operator (aLASSO) models.

Main Methods:

• Retrospective analysis of 1596 acute PE patients diagnosed via CTPA.
• Utilized landmark analysis for 90-day mortality and aLASSO for variable selection and model building.
• Employed receiver-operating characteristic analysis to assess prediction rule performance.

Main Results:

• For 30-day mortality, malignancy, coronary artery disease, and older age were high-risk factors; anticoagulation was associated with lower risk.
• Recent operations were linked to lower 90-day landmark mortality.
• aLASSO-derived prediction rules demonstrated strong performance compared to standard logistic regression.

Conclusions:

• The aLASSO regression approach combined with landmark analysis offers a novel tool for PE risk stratification in large cohorts.
• Key predictors for increased 30-day mortality post-acute PE diagnosis include malignancy, coronary artery disease, and advanced age.
• These methods simplify risk stratification, with future work to integrate imaging data.