Data Science Day @ Columbia University

We compared hemoglobin oxygen saturation data (SpO2) from a continuous data (CD) repository called BedMaster with intermittent documentation (ID) by the bedside nurse in the first month of life in premature infants. BedMaster is an archive database that is interfacing with the bedside patient monitors to collect and archive vital signs, waveforms and clinical alarm data every 2 seconds. A novel algorithm is then applied to remove artifact based on heart rate signal. CD showed higher number of desaturation episodes per day compared to ID (35 vs 3; p<0.001). CD was also able to be manipulated to show the percent time infants spent in various categories of oxygenation levels, something not yet available at the bedside or in the literature. Using a CD repository and custom algorithms, we are able to categorize a natural history of oxygen burden in premature infants not previously available to us or in the literature.