Alarm fatigue has been recognised as a significant health technology safety risk. 'Probability matching', in which clinicians respond to the alarm at a rate identical to the perceived reliability of the alarm, has been postulated as a model to explain alarm fatigue. In this article, we quantitatively explore the implications of probability matching for systolic blood pressure alarms. We find that probability matching could have a profound effect on clinician response to the alarm, with a response rate of only 8.6% when the alarm threshold is 90 mm Hg and the optimal threshold for a systolic blood pressure alarm would only be 77 mm Hg. We use the mathematical framework to assess a mitigation strategy when clinicians have a limit to the capacity to respond. We find that a tiered alarm in which clinicians receive information on the severity of vital sign perturbation significantly improves the opportunity to rescue patients. Practitioner Summary: Using a theoretical model, we predict that probability matching, a postulated model of clinician behaviour, can result in a profound decrease in clinician response to alarms for decreased blood pressure. A mitigating strategy is to create alarms that convey information on the degree of vital sign perturbation.
Keywords: Bayesian model; alarm fatigue; clinical alarms; probability matching; systolic blood pressure.