About this Application: This app predicts 1-year survival for liver transplant
recipients with a Model for End-stage Liver Disease (MELD) score greater than or equal to 40.
To get started, click on the risk models tab and enter your donor and recipient characteristics. All inputs must be filled.
Project Abstract:Background: Although transplant is beneficial for
patients with the highest acuity (MELD greater or equal to 40), mortality in this group
is high. Predicting which patients are likely to survive for
>1 year would be medically and economically helpful.
Methods: The Scientific Registry of Transplant Recipients database was reviewed
to identify adult liver transplant recipients from 2001 through 2016
with MELD score at or above 40 at transplant. The relationships between 44
recipient and donor factors and 1-year patient survival were examined
using random survival forests methods. Variable importance measures
were used to identify the factors with the strongest influence on
survival, and partial dependence plots were used to determine the
dependence of survival on the target variable while adjusting for
all other variables.
Results: We identified 5309 liver transplants that met
our criteria. The overall 1-year survival of high-acuity patients
improved from 69% in 2001 to 87% in 2016. The strongest predictors
of death within 1 year of transplant were patient on mechanical
ventilator before transplantation, prior liver transplant,
older recipient age, older donor age, donation after cardiac
death, and longer cold ischemia.
Conclusions: Liver transplant outcomes continue to improve even
for patients with high medical acuity. Applying ensemble learning
methods to recipient and donor factors available prior to
transplant can predict survival probabilities for future transplant
cases. This information can be used to facilitate donor/recipient
matching and to improve informed consent.
Notes: This data used to create this app is from SRTR from 2001 through 2016.
The original dataset of 5309 patients was 65.7% male and 83.3% white with a
median age of 55 years old.
Key References: