Computerized selection of pulse oximeter arterial oxygen saturation data

In recent years, technology has produced several new bedside patient monitors that continuously display important physiological parameters. Simply processing more monitoring data, however, does not automatically ensure better patient care. A selection process must be applied to the data to identify that information that will be most useful for human decision-making and computerized decision support. This research dealt specifically with the pulse oximeter in the critical care setting. The research sought to demonstrate that computers can be used to automatically record and report appropriate saturation values and identify real-time clinically important changes in the saturation. To investigate how clinicians use pulse oximetry data and what constitutes a clinically important change in saturation, data collection, informal interviews and questionnaires/structured interviews were conducted with clinicians. These activities yielded general perceptions of clinicians' practices and of the notions of "clinically important" events, but it was apparent that a clear consensus did not exist among them. To address the possibility of computerized charging of saturation values, computer-calculated median saturation values were compared to values nurses had charted. The computer-selected median oxygen saturation values were reasonable substitutes for the values manually charted by the nurses. The computer-selected values were sometimes more timely and representative. In an investigation examining real-time detection of important events, two detection algorithms were implemented in a medical intensive care unit. The attending nurses recorded their impressions of the events that were detected. Results suggested that it will be difficult to identify clinically important changes in oxygen saturation on the basis of saturation data alone. Other factors such as mechanical ventilation settings and patient vital signs will need to be incorporated. Computerized selection of pulse oximetry data holds great promise. However, this research suggested that pulse oximetry usage practices are far form uniform and interpretation of the oximeter's saturation signal is equivocal. These problems, in addition to the problem of eliminating artifacts in pulse oximeter's signal, need to be resolved before a reliable recording and detection algorithm can be developed and perhaps even before the fairly expensive and widely used pulse oximetry technology can be justified. In the meantime, recommendations are made for a computerized oxygen saturation recording scheme that may result in the recording of pulse oximetry data that is more timely and representative and that also may assist cost-benefit analysis of pulse oximetry and in further refinement of clear guidelines for uniform pulse oximetry use.