Background
Alarm oversaturation and subsequent alarm fatigue experienced by clinicians are patient safety hazards. Clinicians desensitized to alarms are more likely to miss or ignore patients with actionable alarms. This effect is proportional to the perceived reliability of the alarm system. This study sought to quantify alarm burden and clinical utility.
Methods
A VitalConnect patch and monitor system was attached to the study patients to collect vitals data and produce and record fall, respiratory rate, and heart rate. Clinicians prospectively annotated any actions taken in response to the triggered alarms.
Results
In a study of 593 patients, a total of 6324 alarms were triggered: 2184 HR, 4047 RR, and 93 fall alarms. Only 22 HR, 8 RR, and 3 fall alarms showed clinical significance. Patients had an average of 10.7 alarms per stay only 0.06 of which were actionable on average. Most commonly, patients had no alarms (19%) and only 10% of patients had 28 or more throughout their stay.
Conclusions
The alarm systems studied produced large alarm burdens, with high false-positive rates, and few actionable alerts. This shows the need for new systems built via AI and machine learning, that provide fewer, more clinically useful alerts, better informing hospital staff of patient needs.