Amith Shetty1,2,3,4, Margaret Murphy4,5, Catriona Middleton-Rennie6, Angelo Lancuba4, Malcolm Green6, Harvey Lander6, Mary Fullick6, Ling Li7, Jonathan Iredell2,3, Naren Gunja3,4. 1. Patient Experience and System Performance Division, NSW Ministry of Health, Sydney, New South Wales, Australia. 2. Centre for Infectious Diseases and Microbiology, Westmead Institute for Medical Research, Sydney, New South Wales, Australia. 3. Sydney Medical School, The University of Sydney, Sydney, New South Wales, Australia. 4. Western Sydney Local Health District, Sydney, New South Wales, Australia. 5. The University of Sydney Susan Wakil School of Nursing and Midwifery, Sydney, New South Wales, Australia. 6. Clinical Excellence Commission, NSW Health, Sydney, New South Wales, Australia. 7. Australian Institute of Health Innovation, Macquarie University, Sydney, New South Wales, Australia.
Abstract
OBJECTIVE: Electronic medical records-based alerts have shown mixed results in identifying ED sepsis. Augmenting clinical patient-flagging with automated alert systems may improve sepsis screening. We evaluate the performance of a hybrid alert to identify patients in ED with sepsis or in-hospital secondary outcomes from infection. METHODS: We extracted a dataset of all patients with sepsis during the study period at five participating Western Sydney EDs. We evaluated the hybrid alert's performance for identifying patients with a discharge diagnosis related to infection and modified sequential sepsis-related organ functional assessment (mSOFA) score ≥2 in ED and also compared the alert to rapid bedside screening tools to identify patients with infection for secondary outcomes of all-cause in-hospital death and/or intensive care unit admission. RESULTS: A total of 118 178 adult patients presented to participating EDs during study period with 1546 patients meeting ED sepsis criteria. The hybrid alert had a sensitivity - 71.2% (95% confidence interval 68.8-73.4), specificity - 96.4% (95% confidence interval 96.3-96.5) for identifying ED sepsis. Clinician flagging identified additional alert-negative 232 ED sepsis and 63 patients with secondary outcomes and 112 alert-positive patients with infection and ED mSOFA score <2 went on to die in hospital. CONCLUSION: The hybrid alert performed modestly in identifying ED sepsis and secondary outcomes from infection. Not all infected patients with a secondary outcome were identified by the alert or mSOFA score ≥2 threshold. Augmenting clinical practice with auto-alerts rather than pure automation should be considered as a potential for sepsis alerting until more reliable algorithms are available for safe use in clinical practice.
OBJECTIVE: Electronic medical records-based alerts have shown mixed results in identifying ED sepsis. Augmenting clinical patient-flagging with automated alert systems may improve sepsis screening. We evaluate the performance of a hybrid alert to identify patients in ED with sepsis or in-hospital secondary outcomes from infection. METHODS: We extracted a dataset of all patients with sepsis during the study period at five participating Western Sydney EDs. We evaluated the hybrid alert's performance for identifying patients with a discharge diagnosis related to infection and modified sequential sepsis-related organ functional assessment (mSOFA) score ≥2 in ED and also compared the alert to rapid bedside screening tools to identify patients with infection for secondary outcomes of all-cause in-hospital death and/or intensive care unit admission. RESULTS: A total of 118 178 adult patients presented to participating EDs during study period with 1546 patients meeting ED sepsis criteria. The hybrid alert had a sensitivity - 71.2% (95% confidence interval 68.8-73.4), specificity - 96.4% (95% confidence interval 96.3-96.5) for identifying ED sepsis. Clinician flagging identified additional alert-negative 232 ED sepsis and 63 patients with secondary outcomes and 112 alert-positive patients with infection and ED mSOFA score <2 went on to die in hospital. CONCLUSION: The hybrid alert performed modestly in identifying ED sepsis and secondary outcomes from infection. Not all infectedpatients with a secondary outcome were identified by the alert or mSOFA score ≥2 threshold. Augmenting clinical practice with auto-alerts rather than pure automation should be considered as a potential for sepsis alerting until more reliable algorithms are available for safe use in clinical practice.
Authors: Khalia Ackermann; Jannah Baker; Malcolm Green; Mary Fullick; Hilal Varinli; Johanna Westbrook; Ling Li Journal: J Med Internet Res Date: 2022-02-23 Impact factor: 7.076