Dataset Information

Machine learning does not improve upon traditional regression in predicting outcomes in atrial fibrillation: an analysis of the ORBIT-AF and GARFIELD-AF registries.

ABSTRACT:

Aims

Prediction models for outcomes in atrial fibrillation (AF) are used to guide treatment. While regression models have been the analytic standard for prediction modelling, machine learning (ML) has been promoted as a potentially superior methodology. We compared the performance of ML and regression models in predicting outcomes in AF patients.

Methods and results

The Outcomes Registry for Better Informed Treatment of Atrial Fibrillation (ORBIT-AF) and Global Anticoagulant Registry in the FIELD (GARFIELD-AF) are population-based registries that include 74 792 AF patients. Models were generated from potential predictors using stepwise logistic regression (STEP), random forests (RF), gradient boosting (GB), and two neural networks (NNs). Discriminatory power was highest for death [STEP area under the curve (AUC) = 0.80 in ORBIT-AF, 0.75 in GARFIELD-AF] and lowest for stroke in all models (STEP AUC = 0.67 in ORBIT-AF, 0.66 in GARFIELD-AF). The discriminatory power of the ML models was similar or lower than the STEP models for most outcomes. The GB model had a higher AUC than STEP for death in GARFIELD-AF (0.76 vs. 0.75), but only nominally, and both performed similarly in ORBIT-AF. The multilayer NN had the lowest discriminatory power for all outcomes. The calibration of the STEP modelswere more aligned with the observed events for all outcomes. In the cross-registry models, the discriminatory power of the ML models was similar or lower than the STEP for most cases.

Conclusion

When developed from two large, community-based AF registries, ML techniques did not improve prediction modelling of death, major bleeding, or stroke.

SUBMITTER: Loring Z

PROVIDER: S-EPMC7657384 | biostudies-literature |

REPOSITORIES: biostudies-literature

ACCESS DATA

Similar Datasets

Project description:AimsThis study aimed to identify relationships in recently diagnosed atrial fibrillation (AF) patients with respect to anticoagulation status, use of guideline-directed medical therapy (GDMT) for comorbid cardiovascular conditions (co-GDMT), and clinical outcomes. The Global Anticoagulant Registry in the FIELD (GARFIELD)-AF is a prospective, international registry of patients with recently diagnosed non-valvular AF at risk of stroke (NCT01090362).Methods and resultsGuideline-directed medical therapy was defined according to the European Society of Cardiology guidelines. This study explored co-GDMT use in patients enrolled in GARFIELD-AF (March 2013-August 2016) with CHA2DS2-VASc ≥ 2 (excluding sex) and ≥1 of five comorbidities-coronary artery disease, diabetes mellitus, heart failure, hypertension, and peripheral vascular disease (n = 23 165). Association between co-GDMT and outcome events was evaluated with Cox proportional hazards models, with stratification by all possible combinations of the five comorbidities. Most patients (73.8%) received oral anticoagulants (OACs) as recommended; 15.0% received no recommended co-GDMT, 40.4% received some, and 44.5% received all co-GDMT. At 2 years, comprehensive co-GDMT was associated with a lower risk of all-cause mortality [hazard ratio (HR) 0.89 (0.81-0.99)] and non-cardiovascular mortality [HR 0.85 (0.73-0.99)] compared with inadequate/no GDMT, but cardiovascular mortality was not significantly reduced. Treatment with OACs was beneficial for all-cause mortality and non-cardiovascular mortality, irrespective of co-GDMT use; only in patients receiving all co-GDMT was OAC associated with a lower risk of non-haemorrhagic stroke/systemic embolism.ConclusionIn this large prospective, international registry on AF, comprehensive co-GDMT was associated with a lower risk of mortality in patients with AF and CHA2DS2-VASc ≥ 2 (excluding sex); OAC therapy was associated with reduced all-cause mortality and non-cardiovascular mortality, irrespective of co-GDMT use.Clinical trial registrationClinical Trial Registration-URL: http://www.clinicaltrials.gov. Unique identifier: NCT01090362.

Project description:OBJECTIVES:To externally validate the accuracy of the Global Anticoagulant Registry in the FIELD-Atrial Fibrillation (GARFIELD-AF) model against existing risk scores for stroke and major bleeding risk in patients with non-valvular AF in a population-based cohort. DESIGN:Retrospective cohort study. SETTING:Danish nationwide registries. PARTICIPANTS:90 693 patients with newly diagnosed non-valvular AF were included between 2010 and 2016, with follow-up censored at 1 year. PRIMARY AND SECONDARY OUTCOME MEASURES:External validation was performed using discrimination and calibration plots. C-statistics were compared with CHA2DS2VASc score for ischaemic stroke/systemic embolism (SE) and HAS-BLED score for major bleeding/haemorrhagic stroke outcomes. RESULTS:Of the 90 693 included, 51 180 patients received oral anticoagulants (OAC). Overall median age (Q1, Q3) were 75 (66-83) years and 48 486 (53.5%) were male. At 1-year follow-up, a total of 2094 (2.3%) strokes/SE, 2642 (2.9%) major bleedings and 10 915 (12.0%) deaths occurred. The GARFIELD-AF model was well calibrated with the predicted risk for stroke/SE and major bleeding. The discriminatory value of GARFIELD-AF risk model was superior to CHA2DS2VASc for predicting stroke in the overall cohort (C-index: 0.71, 95% CI: 0.70 to 0.72 vs C-index: 0.67, 95% CI: 0.66 to 0.68, p<0.001) as well as in low-risk patients (C-index: 0.64, 95% CI: 0.59 to 0.69 vs C-index: 0.57, 95% CI: 0.53 to 0.61, p=0.007). The GARFIELD-AF model was comparable to HAS-BLED in predicting the risk of major bleeding in patients on OAC therapy (C-index: 0.64, 95% CI: 0.63 to 0.66 vs C-index: 0.64, 95% CI: 0.63 to 0.65, p=0.60). CONCLUSION:In a nationwide Danish cohort with non-valvular AF, the GARFIELD-AF model adequately predicted the risk of ischaemic stroke/SE and major bleeding. Our external validation confirms that the GARFIELD-AF model was superior to CHA2DS2VASc in predicting stroke/SE and comparable with HAS-BLED for predicting major bleeding.