Performance of Atrial Fibrillation Burden Trends for Stroke Risk Stratification

Figures

• 

  Figure 1.. Study design. An index date was nominally assigned at 21 days after insertable cardiac monitor implantation to initialize a time series for each diagnostic device parameter. Patients with an ischemic stroke post-implantation were included if the first event occurred after the initialization period and with continuous enrollment (period C).

• 

  Figure 2.. Atrial tachycardia (AT)/atrial fibrillation (AF) burden temporal trends. The comparison of a 21-day simple moving average (SMA) with its cumulative moving average (CMA) shows when the daily AT/AF burden amount is above or below average. The onset of relative risk occurs when the SMA crosses above the CMA, remains elevated when the SMA is greater than the CMA, and ends when the SMA crosses below the CMA. The trend is sensitive to a relative increase in daily AT/AF burden (A) and longer periods of increasing burden (B). The patients in both examples experienced an ischemic stroke on the last day of follow-up; follow-up dates were different for each example.

• 

  Figure 3.. Mean variable importance by feature. This figure provides a bar plot of mean variable importance, scaled as a percent of total variable importance, and a color-coded association with stroke risk for each feature in the analysis. For ease of presentation, device parameters show aggregate variable importance across all their respective features. The inset presents the top 30 individual features after stroke/transient ischemic attack (TIA) and history of atrial fibrillation (AF) that were selected as predictors. All trends are defined as the p-day simple moving average offset by its cumulative moving average. AFB indicates atrial fibrillation burden; COPD, chronic obstructive pulmonary disease; DA, daily activity; DHR, daytime heart rate; HRV, heart rate variability; NHR, nighttime heart rate; and OAC, oral anticoagulation.

• 

  Figure 4.. Temporal association of atrial tachycardia (AT)/atrial fibrillation (AF) burden trend and stroke risk. The temporal relationship between AT/AF burden and ischemic stroke risk differed by device indication with AF management prioritizing shorter durations of risk (1–5 days) more frequently than longer durations and cryptogenic stroke selecting a longer duration of risk (21 days) more frequently than shorter durations. While a 21-day simple moving average, offset with its cumulative moving average was the most robust temporal trend across device indications (Figure 3, inset), its frequency of selection as a predictor differed by indication, occurring 94% of the time for cryptogenic stroke and 11% of the time for AF management.

• 

  Figure 5.. Mean area under the receiver operating characteristic curve (AUC) by model structure. This error bar plot shows the mean AUC and 95% credible interval from 1000 bootstrapped, hold-out validation samples for 12 different model structures. Device values indicate the raw device parameter data for atrial tachycardia (AT)/atrial fibrillation (AF) burden, patient activity, daytime heart rate, nighttime heart rate, and heart rate variability; device trends; the temporal trends of the device parameter data; CHA2DS2-Vasc, the eponymous score assessed via claims based on the patient’s clinical history before device implantation; and baseline characteristics, all comorbidities assessed via claims based on the patient’s clinical history before device implantation. Model 6a adds its respective features to the preceding model structure. Model 6b is the same model structure as model 6a and fits with a subset of patients identified as having at least 1 day during follow-up with an AT/AF burden amount of >0.

• 

  Figure 6.. Random sample of atrial fibrillation (AF) burden temporal trends and signaled ischemic stroke risk. Shown are data from a random sample of 16 patients having a CHA2DS2-VASc score >3 and the occurrence of ischemic stroke. The x axis represents follow-up days from the day after device implantation through the day before the stroke event. Risk, denoted in red, is signaled after a relative increase in atrial tachycardia (AT)/AF burden and turned off/on by the daily AT/AF burden 21-day simple moving average crossing below/above its cumulative moving average. A correct classification is shown when any of the last 5 follow-up days is signaled at risk.