Derivation and Validation of a General Predictive Model for Long Term Risks for Mortality and Invasive Cardiovascular Interventions in Congenital Heart Disease
Year: 2021
Type: article
Abstract: Introduction. Accurate assessment of prognosis is a key driver of clinical decision making in congenital heart disease (CHD), but is complicated because CHD represents such a diverse collection of conditions. The aim of this investigation is to derive, validate, and calibrate multivariable predictive models for time to surgical or catheter-mediated intervention (INT) in CHD and for time to death in CHD. Methods. 4108 unique subjects were prospectively and consecutively enrolled, and randomized to derivation and validation cohorts. Total follow up was 26,578 patient-years, with 102 deaths and 868 INTs. Accelerated failure time multivariable predictive models for the outcomes, based on primary and secondary diagnoses, pathophysiologic severity, age, gender, genetic comorbidities, and prior interventional history, were derived using piecewise exponential methodology. The model predictions were validated, calibrated, and evaluated for sensitivity to changes in the independent variables. Results. Model validity was excellent for prediction of both mortality and INT at 4 months, 1 year, 5 years, 10 years, and 22 years (areas under receiver operating characteristic curves ranged from 0.809 to 0.919), and predictions calibrated well with observed outcomes. Although age, gender, secondary diagnoses, and genetic comorbidities were significant independent contributors to the survival and/or freedom from intervention models, predicted outcomes were most sensitive to variations in a composite predictor incorporating primary diagnosis, pathophysiologic severity, and history of prior intervention. An active cohort effect is identified in which predicted mortality and intervention both increased throughout the 22 years of study. Conclusions. Time to INT and time to death in CHD can be predicted with accuracy based on clinical variables. The objective predictions available through these models could educate both patient and provider, and inform clinical decision making in CHD.
Source: DigitalCommons@UNMC
Author: Danford, David A.
Institution: University of Nebraska Medical Center
Language: English
Subfield: Cardiology and Cardiovascular Medicine
Field: Medicine
Domain: Health Sciences
Open Access status: green
OpenAlex ID: W7064733134
