Key result
A machine learning model using 13 routinely available ICD-coded clinical variables predicted incident atrial fibrillation with an AUC of 0.829 (95% CI 0.826-0.833) in internal validation.
Why the study?
Over a third of AF cases remain undiagnosed, and tailoring AF screening is particularly pressing in primary care and stroke survivors.
Does a machine learning model using ICD-coded clinical variables accurately identify patients at risk of newly diagnosed atrial fibrillation in primary care and post-stroke?
Observational (n=1,476,391)
Does a machine learning model using ICD-coded clinical variables accurately identify patients at risk of newly diagnosed atrial fibrillation in primary care and post-stroke?
Effect estimate: AUC 0.829 (95% CI 0.826-0.833)
Machine learning models using routinely available ICD codes can effectively identify patients at high risk for incident and post-stroke atrial fibrillation to target screening efforts.
Enables scalable AF risk stratification from routine codes; leaves open external validation and outcome impact.
AIMS: Atrial fibrillation (AF) carries a substantial risk of ischemic stroke and other complications, and estimates suggest that over a third of cases remain undiagnosed. AF detection is particularly pressing in stroke survivors. To tailor AF screening efforts, we explored German health claims data for routinely available predictors of incident AF in primary care and post-stroke using machine learning methods. METHODS AND RESULTS: We combined AF predictors in patients over 45 years of age using claims data in the InGef database (n = 1 476 391) for (i) incident AF and (ii) AF post-stroke, using machine learning techniques. Between 2013-2016, new-onset AF was diagnosed in 98 958 patients (6.7%). Published risk factors for AF including male sex, hypertension, heart failure, valvular heart disease, and chronic kidney disease were confirmed. Component-wise gradient boosting identified additional predictors for AF from ICD-codes available in ambulatory care. The area under the curve (AUC) of the final, condensed model consisting of 13 predictors, was 0.829 (95% confidence interval (CI) 0.826-0.833) in the internal validation, and 0.755 (95% CI 0.603-0.890) in a prospective validation cohort (n = 661). The AUC for post-stroke AF was of 0.67 (95% CI 0.651-0.689) in the internal validation data set, and 0.766 (95% CI 0.731-0.800) in the prospective clinical cohort. CONCLUSION: ICD-coded clinical variables selected by machine learning can improve the identification of patients at risk of newly diagnosed AF. Using this readily available, automatically coded information can target AF screening efforts to identify high-risk populations in primary care and stroke survivors.
No takes yet. Share an insight, caveat, or question.
Schnabel et al. (2022) conducted an observational in Atrial fibrillation (n=1,476,391). Machine learning-based risk-factor signature (13 predictors) was evaluated on Incident atrial fibrillation prediction (AUC) (AUC 0.829, 95% CI 0.826-0.833). A machine learning model using 13 routinely available ICD-coded clinical variables predicted incident atrial fibrillation with an AUC of 0.829 (95% CI 0.826-0.833) in internal validation.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: