Assessing the accuracy of the lung allocation score

Area of Science:

• Transplantation Science
• Medical Informatics
• Biostatistics

Background:

• The United States (US) Lung Allocation Score (LAS) utilizes two survival models to predict waitlist and post-transplant outcomes.
• These models were developed using data from 2005-2008, raising questions about their current accuracy.

Purpose of the Study:

• To evaluate the accuracy of the LAS waitlist and post-transplant survival models.
• To assess if the current LAS accurately ranks lung transplant candidates and recipients.

Main Methods:

• An observational cohort study of US lung transplantation candidates and recipients (age >12) from 2015-2019.
• Evaluation of LAS models using concordance probability and comparison of predicted vs. observed survival times.
• Comparison of nonparametric observed LAS with predicted LAS across recipient percentiles.

Main Results:

• The waitlist model correctly ranked candidates 72% of the time and underestimated survival by 136 days in the highest risk decile.
• The post-transplant model correctly ranked recipients 57% of the time and underestimated survival by 70 days in the highest risk decile.
• The LAS explained only 56% of outcome variation and showed decreased accuracy for higher predicted values.

Conclusions:

• The LAS waitlist and post-transplant models are inaccurate, impairing the accurate risk stratification of candidates.
• Modernization and updating of the LAS survival models are recommended to improve the lung allocation system.