Background: Recent studies have suggested a high prevalence of subclinical atrial fibrillation (AF) in various patient populations, and interest in AF screening has increased. However, knowledge about episode duration is scarce, and risk factors for short or long subclinical AF episodes have yet to be recognized. The aim of the study was to assess AF by long-term continuous screening and to investigate predictors of episodes lasting ≥6 minutes, ≥5.5 hours, or ≥24 hours, respectively. Methods: A total of 597 patients aged ≥70 years and diagnosed with ≥1 of hypertension, diabetes, previous stroke, or heart failure were recruited from the general population to receive implantable loop recorder with remote monitoring. Exclusion criteria included history of AF or cardiac implantable electronic device. AF episodes were adjudicated by senior cardiologists. Results: During 40 (37; 42) months of continuous monitoring, AF was detected in 209 (35%) of the patients. The cumulative incidences at 3 years were 33.8% (30.2%-37.8%), 16.1% (13.4%-19.4%), and 5.7% (4.1%-7.9%) for AF episodes lasting ≥6 minutes, ≥5.5 hours, and ≥24 hours, respectively. Slower resting sinus rate and higher body mass index, N-terminal prohormone of brain natriuretic peptide, and troponin T at baseline were independently associated with AF detection. Addition of these markers to a model of sex, age, and comorbidities improved prediction of AF episodes ≥24 hours (time-dependent area under the receiver operating characteristic curve 79% vs 65%, P =.037). Conclusions: A considerable burden of previously unknown AF was detected when long-term monitoring was applied in at-risk patients. Biomarkers were associated with AF incidence and improved prediction of long AF episodes.