Impact of atrial fibrillation burden on quality of life and prediction of health care utilization: Primary results from the DEFINE AFib study

BACKGROUND: Atrial fibrillation (AF) can lead to significant cardiovascular events and health care utilization. Continuous monitoring via insertable cardiac monitors (ICMs) and data analytics have the potential to improve care delivery. OBJECTIVE: The DEFINE Atrial Fibrillation (DEFINE AFib) study was designed to develop and evaluate novel algorithms using ICM data to predict AF-associated clinical actions (AFCAs) and guide AF management. METHODS: DEFINE AFib enrolled patients with an ICM (Reveal LINQ/LINQ II; Medtronic) and a history of AF. An Apple iPhone application collected AF-related quality of life (AFEQT) and EQ-5D data. ICM daily AF burden and Apple Watch (AW) irregular rhythm notification (IRN) data were also collected. AFCA was defined as an AF-related procedure or initiation of rate/rhythm control medication. Mutivariable logistic regression was used to identify ICM features in the last 30 days associated with first occurrence of AFCA in the next 30 days in a train and test approach (70%/30%). RESULTS: Among 864 patients (mean 69 ± 10 years; 56% male) meeting inclusion criteria, there were 8963 30-day evaluation windows that included 151 AFCAs. Area under the receiver operating characteristic curve (AUC) was 76% (train) and 70% (test). The model placed participants into high- vs low-risk AFCA groups. At the patient-level, 21% of participants crossing the high-risk threshold for their first time experienced an AFCA at a mean time of 195 ± 164 days compared with 5% in the low-risk group (AUC 65%). Increasing daily mean AF burden was associated with lower quality of life: <6 minutes (reference), 6 minutes to 5.5 hours (AFEQT -7.69; P < .001), and 5.5-12 hours (AFETQ -13.97, P < .001; ≥5.5 hours EQ-5D -0.03, P = .007). In a subanalysis of individuals with smart watch data (n = 53), the ICM model signaled high risk before AFCAs 85% of the time (AUC 57%) compared with 23% for AW IRN (P = .005; AUC 55%). CONCLUSION: DEFINE AFib transformed ICM diagnostic data to predict risk of AFCA with good discrimination, particularly compared with wearable data. These results highlight the potential advantages of ICM-based continuous monitoring for AF management and the utility of ICM prediction models that could help inform pre-emptive therapeutic strategies.