Rationale: High frequency activity (HFA) shows promise as an epileptogenic biomarker. Current processing methods assess limited durations of recordings, typically ranging from ten to 90 minutes during non-REM sleep on the first night of monitoring. Considering longer durations may improve the predictive value of HFA biomarkers for many patients. However, considering longer durations could lead to unstable baseline levels of HFA which would alter high frequency oscillation (HFO) analyses and comparisons. As a first step in assessing long duration HFA, we evaluated the stability of baseline-HFA over 24 hours (hrs). Methods: Baseline-HFA from the first 24 hrs of data for six pediatric patients undergoing intracranial EEG (iEEG) monitoring for epilepsy surgery was evaluated. Baseline-HFA was computed as the standard HFO threshold, i.e., five times the standard deviation above the mean of the band power (80-400 Hz) for ten minutes. HFA across the day was compared to the HFA from quiet data during two hours of contiguous sleep. Results: A total of 514 channels (referential montage) were evaluated (60-110 per patient) from patients ages 1.92-18.50 yr). Data from 5 p.m. day-1 to 5 pm day-2 were considered with 2-hr sleep-baseline times happening between 11:30 p.m. – 4:15 a.m. Baseline-HFA was significantly different from sleep-HFA for 33.3% of channels (Ranksum with Bonferroni adjustment). The median difference between non-sleep and sleep HFA was 14.1±18.1 µV2 (median±IQR) across all patients. Two patients had one or no channels with significant differences. A secondary analysis looking only at channels with pathological EEG activity similarly found 15 of the 49 (30.6%) channels had significantly different baseline-HFA from sleep-HFA. Conclusions: Relying on a static baseline-HFA for detecting HFOs can lead to misleading results. In our cohort, using a static baseline-HFA during sleep may produce more false positive HFOs detections across the 24-hrs. The mechanisms of HFA fluctuations require further investigation. When analyzing long-durations, HFO algorithms should not rely on quiet data to determine parameters, but should instead consider the context of the surrounding data. Funding: Please list any funding that was received in support of this abstract.: Supported by the Phoenix Children’s Hospital Innovators in Neuroscience for Kids Foundation, NSF I/UCRC Cooperative Grant for Building Reliable Innovation and Advances in Neurotechnology, and Phoenix Children’s Hospital Foundation Leadership Circle Award. Its contents are the responsibility of the authors and do not necessarily represent the view of the funding agencies.