EP-4742267-A1 - ALARM FATIGUE MITIGATION

EP4742267A1EP 4742267 A1EP4742267 A1EP 4742267A1EP-4742267-A1

Abstract

An Artificial Intelligence ("Al") solution that can be used to identify alarms that contribute to alarm fatigue is disclosed. Clinicians are directly involved in providing feedback that can be used to train supervised machine learnings models. These insights can then be used by the hospital to make data-driven actions for reducing the number of alarms generated in a hospital. The technological system described herein includes a machine learning ("ML") model deployed in a hospital that is provided with alarms and an associated feature set. The technological system embodies a method by which clinicians can flag potential nuisance alarms, thereby providing direct feedback to the system. The technological system also includes a recommender system that provides a prioritized set of potential nuisance alarms and recommendations.

Inventors

HARROD, John P.

Assignees

Drägerwerk AG & Co. KGaA

Dates

Publication Date: 20260513
Application Date: 20251107

Claims (15)

A computer-implemented method for managing alarm fatigue, comprising: generating a set of candidate nuisance alarms from an aggregation of alarm data to recommend for clinician evaluation; receiving clinician evaluations for the recommended candidate nuisance alarms, the evaluations identifying the recommended candidate nuisance alarms as actual nuisance alarms or non-nuisance alarms; aggregating the identified nuisance and identified non-nuisance alarms into a cumulative set of alarm data; and iterating the above steps.
The computer-implemented method of claim 1, further comprising, in each iteration after receiving the aggregated alarm data: classifying the alarms whose data has been aggregated as potential nuisance alarms or non-nuisance alarms; and prioritizing the classified alarms to generate a prioritized set of potential nuisance alarms and recommendations for corrective actions, wherein the potential nuisance alarms comprise the recommended candidate nuisance alarms.
The computer-implemented method of claim 2, wherein the potential nuisance alarms include a low battery technical alarm, a parameter high limit, a parameter low limit, or a hardware failure, or combinations thereof.
The computer-implemented method of claim 2 or 3, further comprising presenting one or more recommended actions for a potential nuisance alarm, the recommended actions including adjusting the limit value of a parameter.
The computer-implemented method of any one of claims 1 to 4, wherein the candidate alarms are prioritized by one or more factors including one or more of frequency and confidence level that the candidate alarm is a nuisance alarm, alarm severity, alarm location, alarm type, or severity of potential impact on the patient's physical condition.
A computing system, comprising: one or more processors; and a memory on which resides, an aggregation of alarm data; and a set of instructions that, when executed by the one or more processors, performs a method for managing alarm fatigue, the method comprising: generating a set of candidate nuisance alarms from the aggregation of alarm data to recommend for clinician evaluation; receiving clinician evaluations for the recommended candidate nuisance alarms, the evaluations identifying the recommended candidate nuisance alarms as actual nuisance alarms or non-nuisance alarms; aggregating the identified nuisance and identified non-nuisance alarms into a cumulative set of alarm data; and iterating the above steps.
The computing system of claim 6, wherein the programmed method further comprises: in each iteration after receiving the aggregated alarm data: classifying the alarms whose data has been aggregated as potential nuisance alarms or non-nuisance alarms; and prioritizing the classified alarms to generate a prioritized set of potential nuisance alarms and recommendations for corrective actions, wherein the potential nuisance alarms comprise the recommended candidate nuisance alarms.
The computing system of claim 7, wherein the potential nuisance alarms include a low battery technical alarm, a parameter high limit, a parameter low limit, or an oxygen saturation (SpO2) hardware failure, or combinations thereof.
The computer system of claim 7 or 8, further comprising presenting one or more recommended actions for a potential nuisance alarm, the recommended actions including adjusting the limit value of a parameter.
The computing system of any one of claims 6 to 9, wherein the candidate alarms are prioritized.
The computing system of claim 10, wherein the candidate alarms are prioritized by one or more factors including one or more of frequency and confidence level that the candidate alarm is a nuisance alarm, alarm severity, alarm location, alarm type, or severity of potential impact on the patient's physical condition.
A technological system, comprising: a plurality of medical devices programmed to: detect alarm conditions; announce alarms responsive to detecting alarm conditions; and transmit alarm data associated with the alarm; and a computing apparatus, comprising: one or more processors; and a memory on which resides, an aggregation of alarm data; and a set of instructions that, when executed by the one or more processors, performs a method for managing alarm fatigue, the method comprising: generating a set of candidate nuisance alarms from the aggregation of alarm data to recommend for clinician evaluation; receiving clinician evaluations for the recommended candidate nuisance alarms, the evaluations identifying the recommended candidate nuisance alarms as actual nuisance alarms or non-nuisance alarms; aggregating the identified nuisance and identified non-nuisance alarms into a cumulative set of alarm data; and iterating the above steps.
The technological system of claim 12, wherein the programmed method further comprises: in each iteration after receiving the aggregated alarm data: classifying the alarms whose data has been aggregated as potential nuisance alarms or non-nuisance alarms; and prioritizing the classified alarms to generate a prioritized set of potential nuisance alarms and recommendations for corrective actions, wherein the potential nuisance alarms comprise the recommended candidate nuisance alarms.
The technological system of claim 13, further comprising presenting one or more recommended actions for a potential nuisance alarm, the recommended actions including adjusting the limit value of a parameter.
The technological system of any one of claims 12 to 14, wherein the candidate alarms are prioritized by one or more factors including one or more of frequency and confidence level that the candidate alarm is a nuisance alarm, alarm severity, alarm location, alarm type, or severity of potential impact on the patient's physical condition.

Description

TECHNICAL FIELD The present disclosure relates generally to the field of alarm management in a medical care context and, more particularly, to a method and apparatus for managing alarm fatigue. BACKGROUND This section of this document introduces information about and/or from the art that may provide context for or be related to the subject matter described herein and/or claimed below. It provides background information to facilitate a better understanding of the various aspects of the present invention. This is a discussion of "related" art. That such art is related in no way implies that it is also "prior" art. The related art may or may not be prior art. The discussion in this section of this document is to be read in this light, and not as admissions of prior art. Medical care environments can be full of alarms. Many pieces of equipment can issue a number of alarms for a variety of reasons. Many of these alarms may indicate a serious condition that needs to be addressed by a caregiver. Some of these alarms are "nuisance alarms". That is, they are either in error (a "false alarm") or indicate a condition that does not necessarily need to be addressed or, at least, not with much alacrity. The proliferation of alarms has led to a condition known as "alarm fatigue". Those in the art having the benefit of this disclosure will appreciate that whether any particular alarm is a "nuisance alarm" will depend to some degree on well-known factors such as the context in which an alarm occurs and on who is deciding whether the alarm is a nuisance. Some common characteristics of various kinds of nuisance alarms are set forth in Table 1 below. TABLE 1. Characteristics of Common Kinds of Nuisance AlarmsCharacteristicDescriptionExamplesFalse AlarmsAlarms triggered without a valid event, often due to technical issues.Poor electrode contact, signal artifacts.Nonactionable AlarmsTrue alarms that do not require clinical intervention.Alarms triggered by minor fluctuations in heart rate that are not clinically significant.High FrequencyAlarms that occur more frequently.Frequent alarms due to overly sensitive settings.Lack of Clinical RelevanceAlarms that do not provide useful information for patient care.Alarms for minor, non-threatening changes in patient vitals. Alarm fatigue is a complex and pervasive problem that occurs when clinicians are exposed to excessive numbers of alarms, which can result in the desensitization to alarm sounds and an increased rate of missed alarms. This can lead adverse consequences. Research indicates that 72% to 99% of all alarms are false, which has led to the prevalence of alarm fatigue. Alarm fatigue is a sensory overload that occurs when clinicians are exposed to an excessive number of alarms, which can result in desensitization to alarm sounds and an increased rate of missed alarms. The art therefore seeks to mitigate the occurrence and consequences of alarm fatigue. Some approaches attempt to mitigate some kinds of causes of nuisance alarms. For example, some approaches monitor and clean medical equipment as this kind of maintenance can help reduce the frequency of alerts related to technical malfunctions such as low battery or loose a connection. Some approaches seek to decrease clinically inconsequential alerts. One example includes reducing the number of alarms that are announced due to clinically inconsequential events. Still other approaches try to reduce false alarms since reducing the number of false alarms can help reduce alarm fatigue. SUMMARY This disclosure describes an Artificial Intelligence ("Al") solution that can be used to identify alarms that contribute to alarm fatigue. Clinicians are directly involved in providing feedback that can be used to train supervised machine learnings models. These insights can then be used by the hospital to make data-driven actions for reducing the number of alarms generated in a hospital. The technological system described herein includes a machine learning ("ML") model deployed in a hospital that is provided with alarms and an associated feature set. The technological system embodies a method by which clinicians can flag potential nuisance alarms, thereby providing direct feedback to the system. The technological system also includes a recommender system that provides a prioritized set of potential nuisance alarms and recommendations. Accordingly, a ML model is deployed in a hospital or other medical care facility and is provided with alarms and an associated feature set. Clinicians can flag potential nuisance alarms, thereby providing direct feedback to the system. A recommender system that provides a prioritized set of potential nuisance alarms and recommendations may also be provided. In a first aspect, a computer-implemented method for managing alarm fatigue comprises: generating a set of candidate nuisance alarms from an aggregation of alarm data to recommend for clinician evaluation; receiving clinician evaluations for the recommended c