Methods
Study Population: Using patient-level data from the Optum® de-identified EHR dataset from 01-01-2007-to 06-30-2019 linked with Medtronic cardiac implantable electronic device (CIED) data, patients who underwent an AF ablation procedure following CIED implantation were identified. Only the first AF ablation procedure observed in the dataset was considered. Following a 90-day blanking period from the time of the first catheter AF ablation procedure observed in the dataset, patient level daily AT/AF burden data (collected by the CIED) was analyzed for AF recurrence using varying definitions of recurrence, which were then compared against simulated external monitoring strategies used clinically to detect recurrence.
Inclusion and Exclusion criteria: Patients were required to have an AF ablation procedure (identified by the CPT codes 93656 & 93657), and only the patient’s first AF ablation procedure observed in Optum® dataset was considered. Patients were also required to have a Medtronic CIED capable of continuous AF monitoring implanted prior to the AF ablation procedure, have at least 6 months between first record in the Optum EHR and the Medtronic CIED implant date, have daily AT/AF burden available for every day during a period of 30 days prior to AF ablation procedure, with at least one day having greater than 23 hours of available burden data, and have at least 1 daily AT/AF burden measurement available after the end of the 90-day blanking period.
Medical history prior to AF ablation: History of hypertension, prior stroke/TIA, heart failure, diabetes, and vascular diseases were extracted from the Optum® EHR data based on the presence of diagnosis codes (ICD-9/ICD-10) in the diagnosis table on or prior to the AF ablation procedure date. The list of codes considered for each of these conditions is available in Supplement Table 1.
AF type prior to AF ablation: Patients were classified as having persistent AF if they had 7 consecutive daily AT/AF burden > 23 hours in the 30 days prior to the AF ablation procedure; all other patients were considered to have paroxysmal AF.
Atrial Fibrillation Recurrence: AF recurrence definitions included daily AT/AF burden duration of at least 6 min, 1 hour, 6 hours, 12 hours, 23 hours, 7 consecutive days with >23 hours, and a rolling 30-day average burden > 3.6 hours (0.5% burden). Kaplan-Meier estimator was used to estimate the survival probability for each recurrence definition post-ablation following a 90-day blanking period.
Atrial Fibrillation Burden Reduction Following Ablation: Two metrics were considered to quantify the change in AF burden following ablation: percent of time in AT/AF and median daily AT/AF duration.
For each patient, the percent of time in AT/AF was computed as the division of the total AT/AF burden hours by the total observed time, while the median daily AT/AF was the median duration of the observed daily AT/AF burden measurements. Both metrics were computed stratifying by the time period with respect to ablation procedure: pre-ablation (all observed daily AT/AF burden measurements 1 year prior to the AF ablation procedure), and post ablation (all daily AT/AF burden measurements observed between the days 90 and 455 (i.e. one-year post-blanking period) post AF ablation).
Monitoring Strategy Simulation: Using daily AT/AF burden measurements as the source of truth,
sensitivity and negative predictive value (NPV) were computed to assess performance of various external AF monitoring strategies used in clinical practice with respect to the detection of events of daily AT/AF burden > 6 min, daily AT/AF burden > 6 hours and daily AT/AF burden > 23 hours. The simulated monitoring strategies were: 24-hour, 48-hour, 7-day, 21-day, 30-day, quarterly 24-hour, quarterly 48-hour, quarterly 7-day and monthly 24-hour continuous external monitoring. All these monitoring strategies were considered to occur within 1 year after the 90-day blanking period and 100% patient compliance with external monitoring was assumed. Only data from patients who had uninterrupted daily AT/AF burden measurements during this time period were included in the simulation. Lastly, a bootstrap method (with 10,000 samples) was used to build 95% confidence intervals (CI) for the performance metrics. The initial day of the AF monitoring strategy was randomly assigned within the first two weeks post blanking period for each patient.