J L Derrick1, C L Thompson, T G Short. 1. Prince of Wales Hospital and The Chinese University of Hong Kong. jamesderrick@pobox.com
Abstract
OBJECTIVE: We have developed an arterial pressure alarm system based on a modified proportional-derivative (PD) controller algorithm, and prospectively tested its ability to predict significant hypotensive episodes, defined as systolic arterial pressure < 80 mmHg, in comparison to conventional limit alarms. METHODS: The alarm algorithm was tuned to detect hypotension using selected invasive arterial pressure traces taken from ten patients who had large intra-operative arterial pressure changes. The algorithm's performance was then tested prospectively in comparison to conventional limit alarms and median filtered limit alarms, set at 85 mmHg and 90 mmHg, for its ability to predict hypotensive episodes in a further 100 patients who required invasive arterial pressure monitoring. RESULTS: For the PD alarm algorithm, onset times for significant hypotensive episodes were between those of limit alarms set at 85 mmHg and 90 mmHg. Offset times were similar to the 85 mmHg limit alarms. The false positive rate was 34% compared with 45-64% for the other alarms (p < 0.01). Using our definitions, there was one false negative in the PD group, being a 15 second drop in observed arterial pressure, when a non invasive blood pressure cuff was inflated above the arterial line. CONCLUSIONS: An arterial pressure alarm system design based on a closed loop control algorithm offered improved perform ance over conventional limit alarms and in addition provided a graded output of severity of the hypotension.
OBJECTIVE: We have developed an arterial pressure alarm system based on a modified proportional-derivative (PD) controller algorithm, and prospectively tested its ability to predict significant hypotensive episodes, defined as systolic arterial pressure < 80 mmHg, in comparison to conventional limit alarms. METHODS: The alarm algorithm was tuned to detect hypotension using selected invasive arterial pressure traces taken from ten patients who had large intra-operative arterial pressure changes. The algorithm's performance was then tested prospectively in comparison to conventional limit alarms and median filtered limit alarms, set at 85 mmHg and 90 mmHg, for its ability to predict hypotensive episodes in a further 100 patients who required invasive arterial pressure monitoring. RESULTS: For the PD alarm algorithm, onset times for significant hypotensive episodes were between those of limit alarms set at 85 mmHg and 90 mmHg. Offset times were similar to the 85 mmHg limit alarms. The false positive rate was 34% compared with 45-64% for the other alarms (p < 0.01). Using our definitions, there was one false negative in the PD group, being a 15 second drop in observed arterial pressure, when a non invasive blood pressure cuff was inflated above the arterial line. CONCLUSIONS: An arterial pressure alarm system design based on a closed loop control algorithm offered improved perform ance over conventional limit alarms and in addition provided a graded output of severity of the hypotension.