The advent and development of clinical Artificial Intelligence (AI) has been groundbreaking, showing its relevance beyond the workstation, impacting patient care paths, particularly in cardiovascular and neuro spaces. AI’s benefits to physician workflows are numerous. However, one AI challenge is alert fatigue, which can potentially hamper efficiency and efficacy in care scenarios if not managed properly.
Alert fatigue refers to the overabundance of notifications specialists receive about a specific pathology, particularly when the case does not warrant their intervention. This situation can lead to a less serious interpretation of AI notifications, particularly when specialists are inundated with alerts.
AI’s role in pulmonary embolism (PE) cases is critical, given PE’s high mortality rate when left untreated. A facility’s pulmonary embolism response team (PERT) needs to know about potential PE cases as quickly as possible, making AI invaluable. However, not all PEs require PERT activation, with their severity varying, hence the risk of alert fatigue. For example, a massive PE, which obstructs more than 50% of the pulmonary arterial tree, requires immediate PERT intervention. A sub-massive PE, which has notable myocardial necrosis, is often treated with anticoagulants. Therefore, PERT consultations are not usually required unless right ventricular dysfunction is present. Finally, non-massive PEs, stable with normal RV function, do not necessitate PERT consultation, but rather anticoagulant therapy.
To reduce alert fatigue, AI must incorporate additional risk stratification measures, use this data to assess PE’s severity, and notify the PERT only if they meet specific criteria. AI must also integrate with electronic health records (EHRs), enabling real-time lab values to use in patient triage, and when appropriate, refer relevant patients to the PERT.
Without such AI intervention, clinicians risk alert fatigue, leading to notifications being taken less seriously over time. This could affect a patient’s treatment path. For AI to truly breakthrough, it has to be useful, usable and used. Parameters should be in place to limit alert fatigue and ensure PE notifications are significant and require immediate attention. The ultimate value of AI in the healthcare field lies in its ability to reduce unnecessary alerts, help prioritize cases, and enhance the delivery of timely and appropriate care.
