Background: Real time trended physiological data in the intensive care unit varies within limits accepted as the normal range. Limit alarms address variation outside the normal range, but a trend may be abnormal well before these limits are transgressed. We have explored derivatives of real time data to identify artifact as this must be eliminated before intelligence can be built into the monitors.

Hypothesis: Derivatives of real time data can be used 1. To filter artifact from physiological trend data display or 2. To exaggerate artifact thus allowing easy identification and the possibility of exclusion.

Methods: Data is collected on our physiological trend monitoring system at 1 Herz. Derivatives of real time data were used to eradicate or exaggerate artifact. Data channels examined were (a) heart rate where electronic artifact usually spikes down (b) tcPO2, where probe artifact usually goes up and (c) tcPCO2, where probe artifact usually goes down. The derivatives used were (1) median (MD) filter over 5, 10,and 20mins (2) mean(MN) filter over 5, 10, 20mins (3) standard deviation (SD) filter over 5, 10, 20mins and (4) maximum (MX) filter over 5 mins.

Results: Artifact smoothing, slight=-, moderate=--, marked=---. Artifact exaggeration, slight=+, moderate=++, marked=+++. Little change=0.Table

Table 1

Conclusion: Derivatives of real time data can be used to smooth away artifact or exaggerate it so it may be excluded from decision support algorithms.