ICM monitoring intensity and atrial fibrillation detection: insights from U.S. Medicare data
D Ko, X Cheng, P T Evans, J M Ashburner, S J Shah, D S Frankel, D E Singer, D B Kramer, A J Walkey, E J Benjamin, C D Anderson, D H KimAbstract
Background
Risk of post-stroke AF detection with ICMs in routine care populations has not been studied systematically. Examining routine care populations allows evaluation of AF detection in underrepresented patients and the impact of post-implantation monitoring intensity.
Purpose
To characterize the risk of post-stroke AF detection with ICMs in a diverse, routine care population and to examine how differences in ICM monitoring rates may affect AF detection.
Methods
The study cohort included Medicare beneficiaries aged ≥65 years who underwent ICM implantation during 1/1/2017-12/31/2020; had been hospitalized for ischemic stroke within the year prior to implantation; and had no prior diagnosis of AF or dispensing of an oral anticoagulant. The primary outcome was new AF. Follow-up began the day after ICM implantation and ended at the earliest occurrence of an outcome event, death, 365 days, or disenrollment from Medicare. One-year cumulative incidence of new AF was calculated using the Aalen-Johansen method, accounting for competing risk of death. To evaluate the influence of differential ICM monitoring intensity on AF detection across subgroups, we conducted multivariable Poisson regression analyses with the rate of ICM interrogation visits (total number of interrogation encounters divided by person-time) as an offset. The resulting incident rate ratios (IRRs) compare the incidence of new AF given the rate of ICM monitoring. We then compared these IRRs with the subdistribution HRs from Fine-Gray regression model.
Results
There were 9215 individuals in the study cohort. The mean age was 76 years, 4983 (54%) were women, 125 (1.4%) Asian or Pacific Islander, 647 (7.0%) Black, 355 (3.9%) Hispanic, 251 (2.7%) Other, and 7837 (85%) were White. A third of the patients were prescribed an external rhythm monitor prior to ICM implantation. A third of the patients were considered frail at the time of ICM implantation. The median follow-up time was 365 days (IQR 180-365). The one-year cumulative incidence of new AF in the overall cohort was 27.9% (Figure). When stratified by subgroups, one-year cumulative incidence was 23.6% among individuals aged 66–74 years, 31.9% among those aged ≥75 years, 29.1% in men, and 26.9% in women. By race and ethnicity, the cumulative incidence was 26.0% for Asian or Pacific Islander, 20.6% for Black, 24.0% for Hispanic, 26.7% for Other, and 28.8% for White individuals. The cumulative incidence was 26.5% among those with prior use of an external rhythm monitor and 28.6% among those without. After accounting for the rate of ICM monitoring in the Poisson regression model, the association between some of the patient characteristics and the risks of new AF attenuated substantially (Table).
Conclusion
Although ICMs enable AF detection independent of symptoms or in-person encounters, variation in post-implantation monitoring intensity across subgroups may introduce substantial surveillance bias.