1. Validation of a Simple to Use PGD Prediction Algorithm.
- Author
-
Diamond, J.M., Localio, R., Michelson, A., Cantu, E., Clausen, E., Kalman, L., Oyster, M., Criner, R., Anderson, M., Gallop, R., Hachem, R., and Christie, J.D.
- Subjects
- *
LUNG transplantation , *BODY mass index , *LUNG volume measurements , *PREDICTION models , *DECISION making - Abstract
Primary graft dysfunction (PGD) is the leading cause of early morbidity and mortality after lung transplantation. We previously developed an easy to use PGD predictive model utilizing the Lung Transplant Outcomes Group (LTOG) cohort. We therefore sought to demonstrate the net benefit of the model at an external site as a measure of model generalizability. We tested external validation in a well-phenotyped lung transplant cohort from Washington University in St. Louis (WashU) for a PGD predictive model based on lung allocation score, body mass index, pulmonary arterial pressure, recipient total lung capacity, donor and recipient age and gender, donor mode of death, and donor distance to transplant center. We primarily employed decision curve analysis (DCA) to quantify the net benefit of the predictive model and also assessed model discrimination using the c-statistic and calibration using the Hosmer-Lemeshow (H-L) test. The incidence of PGD at WashU was lower than in the LTOG cohort (7% vs 25%); the validation model accounts for the differences in PGD incidence. The model demonstrated good discrimination (c-statistic=0.67) and calibration (H-L p=0.2, where non-significant p-value is good calibration). DCA describes whether the model performs better than alternatives of classifying everyone (or no one) as high PGD risk as well as the net benefit of the model across a range of decision thresholds. The DCA (Figure 1) shows that the predictive model demonstrates significant net benefit in the PGD incidence range 2-15%, incorporating the PGD incidence seen in the WashU cohort. Our simple to implement PGD predictive model demonstrates excellent net benefit, even at centers with differing PGD incidence than the discovery cohort centers. This model can be used to identify recipients at high and low risk for PGD across a range of transplant centers, allowing treatment teams to be prepared for post-transplant complications and identify patients for potential enrollment in targeted intervention studies based on patient risk. [ABSTRACT FROM AUTHOR]
- Published
- 2022
- Full Text
- View/download PDF