Crying wolf: false alarms in a pediatric intensive care unit

Abstract

Objective: To determine the predictive value of patient monitoring alarms as a warning system in a pediatric intensive care unit (ICU).

Design: Prospective, observational study.

Setting: Pediatric ICU of a university affiliated children's hospital.

Interventions: During a 7-day period, ICU staff were asked to record the type and number of alarm soundings. Alarms were recorded as false, significant (resulted in change in therapy), or induced (by staff manipulations; not significant).

Measurements and main results: Sixty-six percent of nursing shifts (928 patient hours of care) responded. There were 2,176 alarms soundings: 1,481 (68%) false, 119 (5.5%) significant, and 576 (26.5%) induced. Alarm origins were: 44% pulse oximeter, 1% end-tidal PCO2, 31% ventilator, and 24% electrocardiograph (EKG). The positive predictive value of alarms were: 7% pulse oximeter, 16% end-tidal PCO2, 3% ventilator, and 5% EKG. The negative predictive value of all alarms were > 97%. More alarms sounded during the 7:00 am to 3:00 pm shift than during the 3:00 pm to 11:00 pm or 11:00 pm to 7:00 am shifts (167 +/- 19 vs. 64 +/- 39 vs. 75 +/- 43, p < .05, respectively). When corrected for number of patients/shift, the occurrence of soundings differed only between day and night (11.4 +/- 1.5/patient/shift vs. 6.1 +/- 1.0, p < .05).

Conclusions: Over 94% of alarm soundings in a pediatric ICU may not be clinically important. Present monitoring systems are poor predictors of untoward events.