Objectives: To assess the relevance of current monitoring alarms as a warning system in the adult ICU.
Design: Prospective, observational study.
Settings: Two university hospital, and three general hospital, ICUs.
Patients: Hundred thirty-one patients, ventilated at admission, from different shifts (morning, evening, night) combined with different stages of stay, early (0-3 days), intermediate (4-6 days) and late (> 6 days).
Interventions: Experienced nurses were asked to record the patient's characteristics and, for each alarm event, the reason, type and consequence.
Measurements and main results: The mean age of the patients included was 59.8 +/- 16.4 and SAPS1 was 15.9 +/- 7.4. We recorded 1971 h of care. The shift distribution was 78 mornings, 85 evenings and 83 nights; the stage distribution was 88 early, 78 intermediate and 80 late. There were 3188 alarms, an average of one alarm every 37 min: 23.7% were due to staff manipulation, 17.5% to technical problems and 58.8% to the patients. Alarms originated from ventilators (37.8%), cardiovascular monitors (32.7%), pulse oximeters (14.9%) and capnography (13.5%). Of the alarms, 25.8% had a consequence such as sensor repositioning, suction, modification of the therapy (drug or ventilation). Only 5.9% of the alarms led to a physician's being called. The positive predictive value of an alarm was 27% and its negative predictive value was 99%. The sensitivity was 97% and the specificity 58%.
Conclusions: The study confirms that the level of monitoring in ICUs generates a great number of false-positive alarms.