RESUMO
BACKGROUND: Changeful seasonal influenza activity in subtropical areas such as Taiwan causes problems in epidemic preparedness. The Taiwan Centers for Disease Control has maintained real-time national influenza surveillance systems since 2004. Except for timely monitoring, epidemic forecasting using the national influenza surveillance data can provide pivotal information for public health response. OBJECTIVE: We aimed to develop predictive models using machine learning to provide real-time influenza-like illness forecasts. METHODS: Using surveillance data of influenza-like illness visits from emergency departments (from the Real-Time Outbreak and Disease Surveillance System), outpatient departments (from the National Health Insurance database), and the records of patients with severe influenza with complications (from the National Notifiable Disease Surveillance System), we developed 4 machine learning models (autoregressive integrated moving average, random forest, support vector regression, and extreme gradient boosting) to produce weekly influenza-like illness predictions for a given week and 3 subsequent weeks. We established a framework of the machine learning models and used an ensemble approach called stacking to integrate these predictions. We trained the models using historical data from 2008-2014. We evaluated their predictive ability during 2015-2017 for each of the 4-week time periods using Pearson correlation, mean absolute percentage error (MAPE), and hit rate of trend prediction. A dashboard website was built to visualize the forecasts, and the results of real-world implementation of this forecasting framework in 2018 were evaluated using the same metrics. RESULTS: All models could accurately predict the timing and magnitudes of the seasonal peaks in the then-current week (nowcast) (ρ=0.802-0.965; MAPE: 5.2%-9.2%; hit rate: 0.577-0.756), 1-week (ρ=0.803-0.918; MAPE: 8.3%-11.8%; hit rate: 0.643-0.747), 2-week (ρ=0.783-0.867; MAPE: 10.1%-15.3%; hit rate: 0.669-0.734), and 3-week forecasts (ρ=0.676-0.801; MAPE: 12.0%-18.9%; hit rate: 0.643-0.786), especially the ensemble model. In real-world implementation in 2018, the forecasting performance was still accurate in nowcasts (ρ=0.875-0.969; MAPE: 5.3%-8.0%; hit rate: 0.582-0.782) and remained satisfactory in 3-week forecasts (ρ=0.721-0.908; MAPE: 7.6%-13.5%; hit rate: 0.596-0.904). CONCLUSIONS: This machine learning and ensemble approach can make accurate, real-time influenza-like illness forecasts for a 4-week period, and thus, facilitate decision making.
Assuntos
Surtos de Doenças/estatística & dados numéricos , Influenza Humana/epidemiologia , Aprendizado de Máquina/normas , Previsões , Humanos , TaiwanRESUMO
This study aimed to evaluate whether quantitative analysis of wrist photoplethysmography (PPG) could detect atrial fibrillation (AF). Continuous electrocardiograms recorded using an electrophysiology recording system and PPG obtained using a wrist-worn smartwatch were simultaneously collected from patients undergoing catheter ablation or electrical cardioversion. PPG features were extracted from 10, 25, 40, and 80 heartbeats of the split segments. Machine learning with a support vector machine and random forest approach were used to detect AF. A total of 116 patients were evaluated. We annotated > 117 h of PPG. A total of 6475 and 3957 segments of 25-beat pulse-to-pulse intervals (PPIs) were annotated as AF and sinus rhythm, respectively. The accuracy of the 25 PPIs yielded a test area under the receiver operating characteristic curve (AUC) of 0.9676, which was significantly better than the AUC for the 10 PPIs (0.9453; P < .001). PPGs obtained from another 38 patients with frequent premature ventricular/atrial complexes (PVCs/PACs) were used to evaluate the impact of other arrhythmias on diagnostic accuracy. The new AF detection algorithm achieved an AUC of 0.9680. The appropriate data length of PPG for optimizing the PPG analytics program was 25 heartbeats. Algorithm modification using a machine learning approach shows robustness to PVCs/PACs.