Validation of Risk Scores for Predicting Atrial Fibrillation Detected After Stroke Based on an Electronic Medical Record Algorithm: A Registry-Claims-Electronic Medical Record Linked Data Study

Kuo

et al. 2022

IJERPH

Atrial fibrillation (AF)—whether paroxysmal or sustained—increases the risk of stroke. We developed and validated a risk score for identifying patients at risk of paroxysmal atrial fibrillation (pAF) after acute ischemic stroke (AIS). A total of 6033 patients with AIS who received 24 h Holter monitoring were identified in the Chang Gung Research Database. Among the identified patients, 5290 with pAF and without AF were included in the multivariable logistic regression analysis to develop the pAF prediction model. The ABCD-SD score (Age, Systolic Blood pressure, Coronary artery disease, Dyslipidemia, and Standard Deviation of heart rate) comprises age (+2 points for every 10 years), systolic blood pressure (−1 point for every 20 mmHg), coronary artery disease (+2 points), dyslipidemia (−2 points), and standard deviation of heart rate (+2 points for every 3 beats per minute). Overall, 5.2% (274/5290) of patients had pAF. The pAF risk ranged from 0.8% (ABCD-SD score ≤ 7) to 18.3% (ABCD-SD score ≥ 15). The model achieved an area under the receiver operating characteristic curve (AUROCC) of 0.767 in the model development group. The ABCD-SD score could aid clinicians in identifying patients with AIS at risk of pAF for advanced cardiac monitoring.

Section: Discussionmentioning

confidence: 99%

Section: Discussionmentioning

confidence: 99%

Development and Validation of a Novel Score for Predicting Paroxysmal Atrial Fibrillation in Acute Ischemic Stroke

Kuo

et al. 2022

IJERPH

Front. Cardiovasc. Med.

“…For comparison with ML models, we only considered traditional risk scores that are based on variables available from EHRs upon admission. According to a validation study that evaluated eight such risk scores, two risk scores performed better than the others, demonstrating adequate discrimination and calibration (19). These two risk scores were thus used as the baseline models.…”

Section: Baseline Modelsmentioning

confidence: 99%

“…To date, more than twenty risk scores have been proposed to assess the risk of poststroke NDAF (18,19). These risk scores vary in their complexity, target population, outcome definition, predictor variables, and ease of implementation.…”

Section: Introductionmentioning

confidence: 99%

Automated risk assessment of newly detected atrial fibrillation poststroke from electronic health record data using machine learning and natural language processing

Sung

Sung²,

Pan³

et al. 2022

Self Cite

BackgroundTimely detection of atrial fibrillation (AF) after stroke is highly clinically relevant, aiding decisions on the optimal strategies for secondary prevention of stroke. In the context of limited medical resources, it is crucial to set the right priorities of extended heart rhythm monitoring by stratifying patients into different risk groups likely to have newly detected AF (NDAF). This study aimed to develop an electronic health record (EHR)-based machine learning model to assess the risk of NDAF in an early stage after stroke.MethodsLinked data between a hospital stroke registry and a deidentified research-based database including EHRs and administrative claims data was used. Demographic features, physiological measurements, routine laboratory results, and clinical free text were extracted from EHRs. The extreme gradient boosting algorithm was used to build the prediction model. The prediction performance was evaluated by the C-index and was compared to that of the AS5F and CHASE-LESS scores.ResultsThe study population consisted of a training set of 4,064 and a temporal test set of 1,492 patients. During a median follow-up of 10.2 months, the incidence rate of NDAF was 87.0 per 1,000 person-year in the test set. On the test set, the model based on both structured and unstructured data achieved a C-index of 0.840, which was significantly higher than those of the AS5F (0.779, p = 0.023) and CHASE-LESS (0.768, p = 0.005) scores.ConclusionsIt is feasible to build a machine learning model to assess the risk of NDAF based on EHR data available at the time of hospital admission. Inclusion of information derived from clinical free text can significantly improve the model performance and may outperform risk scores developed using traditional statistical methods. Further studies are needed to assess the clinical usefulness of the prediction model.

“…Several validation methods were used, including C-index, decision curve analysis (DCA), net reclassification index (NRI), and integrated discrimination improvement (IDI). However, the highest C-index for CHASE-LESS score was 0.741 (12), and the predictive performance was limited by the lack of imaging variables. To date, specific and practical prediction methods are still lacking.…”

Section: Introductionmentioning

confidence: 98%

A nomogram for predicting atrial fibrillation detected after acute ischemic stroke

Pang¹,

Sun³

et al. 2022

Front. Neurol.

BackgroundAtrial fibrillation detected after stroke (AFDAS) is associated with an increased risk of ischemic stroke (IS) recurrence and death. Early diagnosis can help identify strategies for secondary prevention and improve prognosis. However, there are no validated predictive tools to assess the population at risk for AFDAS. Therefore, this study aimed to develop and validate a predictive model for assessing the incidence of AFDAS after acute ischemic stroke (AIS).MethodsThis study was a multicenter retrospective study. We collected clinical data from 5332 patients with AIS at two hospitals between 2014.01 and 2021.12 and divided the development and validation of clinical prediction models into a training cohort (n = 3173) and a validation cohort (n = 2159). Characteristic variables were selected from the training cohort using the least absolute shrinkage and selection operator (LASSO) algorithm and multivariable logistic regression analysis. A nomogram model was developed, and its performance was evaluated regarding calibration, discrimination, and clinical utility.ResultsWe found the best subset of risk factors based on clinical characteristics and laboratory variables, including age, congestive heart failure (CHF), previous AIS/transient ischemia attack (TIA), national institutes of health stroke scale (NIHSS) score, C-reactive protein (CRP), and B-type natriuretic peptide (BNP). A predictive model was developed. The model showed good calibration and discrimination, with calibration values of Hosmer-Lemeshow χ2 = 4.813, P = 0.732 and Hosmer-Lemeshow χ2 = 4.248, P = 0.834 in the training and validation cohorts, respectively. The area under the ROC curve (AUC) was 0.815, 95% CI (0.777–0.853) and 0.808, 95% CI (0.770–0.847). The inclusion of neuroimaging variables significantly improved the performance of the integrated model in both the training cohort (AUC. 0.846 (0.811–0.882) vs. 0.815 (0.777–0.853), P = 0.001) and the validation cohort (AUC: 0.841 (0.804–0.877) vs. 0.808 (0.770–0.847), P = 0.001). The decision curves showed that the integrated model added more net benefit in predicting the incidence of AFDAS.ConclusionPredictive models based on clinical characteristics, laboratory variables, and neuroimaging variables showed good calibration and high net clinical benefit, informing clinical decision-making in diagnosing and treating patients with AFDAS.