Development and validation of a risk prediction model in patients with adult congenital heart disease. (1st February 2019)

• Record Type: Journal Article
• Authors: Baggen, Vivan J.M.; Venema, Esmee; Živná, Renata; van den Bosch, Annemien E.; Eindhoven, Jannet A.; Witsenburg, Maarten; Cuypers, Judith A.A.E.; Boersma, Eric; Lingsma, Hester; Popelová, Jana R.; Roos-Hesselink, Jolien W.
• Is Part Of: International journal of cardiology. Volume 276(2019)
• Page Start: 87
• Page End: 92
• Publication Date: 2019-02-01
• Subjects: Congenital heart disease -- Risk -- Prognosis -- Prediction model -- Adverse events; Cardiology -- Periodicals; Electronic journals; 616.12
• DOI: 10.1016/j.ijcard.2018.08.059
• Languages: English
• ISSNs: 0167-5273

Abstract

Abstract: Aims: To develop and validate a clinically useful risk prediction tool for patients with adult congenital heart disease (ACHD). Methods and results: A risk model was developed in a prospective cohort of 602 patients with moderate/complex ACHD who routinely visited the outpatient clinic of a tertiary care centre in the Netherlands (2011−2013). This model was externally validated in a retrospective cohort of 402 ACHD patients (Czech Republic, 2004–2013). The primary endpoint was the 4-year risk of death, heart failure, or arrhythmia, which occurred in 135 of 602 patients (22%). Model development was performed using multivariable logistic regression. Model performance was assessed with C-statistics and calibration plots. Of the 14 variables that were selected by an expert panel, the final prediction model included age (OR 1.02, 95%CI 1.00–1.03, p = 0.031), congenital diagnosis (OR 1.52, 95%CI 1.03–2.23, p = 0.034), NYHA class (OR 1.74, 95%CI 1.07–2.84, p = 0.026), cardiac medication (OR 2.27, 95%CI 1.56–3.31, p < 0.001), re-intervention (OR 1.41, 95%CI 0.99–2.01, p = 0.060), BMI (OR 1.03, 95%CI 0.99–1.07, p = 0.123), and NT-proBNP (OR 1.63, 95%CI 1.45–1.84, p < 0.001). Calibration-in-the-large was suboptimal, reflected by a lower observed event rate in the validation cohort (17%) than predicted (36%), likely explained by heterogeneity and different treatment strategies. The externally validated C-statistic was 0.78 (95%CI 0.72–0.83), indicating good discriminative ability. Conclusion: The proposed ACHD risk score combines six readily available clinical characteristics and NT-proBNP. This tool is easy to use and can aid in distinguishing high- and low-risk patients, which could further streamline counselling, location of care, and treatment in ACHD. Highlights: The ACHD risk calculator is based on readily available clinical characteristics and NT-proBNP. The risk model provides a good discrimination between high- and low-risk patients with ACHD. A web application is presented, which could further streamline counselling, location of care, and treatment in ACHD. Other investigators are invited to collaborate in order to improve the absolute risk estimations of the current model.