Background and Purpose— Detection rates of paroxysmal atrial fibrillation (AF) after acute ischemic stroke increase with duration of ECG monitoring. To date, it is unknown which patient group may benefit most from intensive monitoring strategies. Therefore, we aimed to identify predictors of previously unknown AF during in-hospital ECG monitoring. Methods— All consecutive patients with imaging-confirmed ischemic stroke admitted to our tertiary care hospital from February 2011 to December 2013 were registered prospectively. Patients received continuous bedside ECG monitoring for at least 24 hours. Detection of previously unknown AF during in-hospital ECG monitoring was obtained from medical records. Patients with AF on admission ECG or known history of AF were excluded from analysis. Results— Among 1228 patients (median age, 73 years; median National Institutes of Health Stroke Scale, 4; 43.4% women), previously unknown AF was detected in 114 (9.3%) during a median time of continuous ECG monitoring of 3 days (interquartile range, 2–4 days). Duration of monitoring ( P <0.01), older age ( P <0.01), history of hypertension ( P =0.03), insular cortex involvement ( P <0.01), and higher high-sensitivity cardiac troponin T ( P =0.04) on admission were independently associated with subsequent detection of AF in a multiple regression analysis. Addition of high-sensitivity cardiac troponin T, insular cortex stroke, or both to the CHADS2 score (congestive heart failure, hypertension, age ≥75 years, diabetes mellitus, stroke [2P]) significantly improved c -statistics from 0.63 to 0.68 ( P =0.01), 0.70 ( P <0.01), and 0.72 ( P <0.001), respectively. Conclusions— Insular cortex involvement, higher admission high-sensitivity cardiac troponin T, older age, hypertension, and longer monitoring are associated with new detection of AF during in-hospital ECG monitoring. Patients with higher high-sensitivity cardiac troponin T or insular cortex involvement may be candidates for prolonged ECG monitoring.