Abstract
AIMS: QTc prolongation is a complex problem linked with multiple risk factors. The RISQ-PATH score was previously developed to identify high-risk patients for QTc prolongation. The aim of this study was to optimize and validate this risk score in a large patient cohort, and to propose an algorithm to generate smart QT signals in the electronic medical record. METHODS: A retrospective study was performed in the Nexus hospital network (n = 17) in Belgium. All electrocardiograms performed in 2015 in both ambulatory and hospitalized patients were collected together with risk factors for QTc prolongation (training database). Multiple logistic regression was performed to obtain the optimal prediction (RISQ-PATH) model. The model was tested in a validation database (electrocardiograms between January and April 2016). RESULTS: In total, 60 208 patients (52.8% males, mean age 63 ± 18 years) were included; 3543 patients (5.9%) had a QTc ≥ 450(♂)/470(♀) ms and 453 (0.8%) a QTc ≥ 500 ms. The optimized RISQ-PATH model has an area under the ROC-curve of 0.772 [95% CI 0.763-0.780] to predict QTc ≥ 450(♂)/470(♀)ms. A predicted probability of ≥0.035 was set as cutoff for a high risk of QTc prolongation. This cutoff resulted in a sensitivity of 87.4% [95% CI 86.2-88.5] and a specificity of 46.2% [95% CI 45.8-46.6]. These results could be confirmed for QTc ≥ 500 ms and in the validation database (n = 28 400). CONCLUSIONS: The RISQ-PATH model, with a cutoff probability of 0.035, predicted a prolonged QTc interval ≥ 450/470 ms or ≥500 ms with a sensitivity of ±87% and a specificity of ±45%. This RISQ-PATH model can be used in clinical decision support systems to create smart QT alerts.