A10 An AI-enabled software as a medical device cleared by the FDA for assessing hemorrhage risk in trauma casualties

Abstract

Introduction Although haemorrhage is the leading cause of preventable death on the battlefield, combat medics lack clinical decision-support systems to help stratify haemorrhage risk in trauma casualties. Data from conflicts in Iraq and Afghanistan suggest that approximately 20% of trauma casualties with life-threatening hemorrhage who died before reaching a medical treatment facility could have potentially survived, had they been quickly identified and treated. Methods Using AI, we previously trained the APPRAISE-Haemorrhage Risk Index (HRI) software to associate patterns in vital signs (heart rate and blood pressures) collected from trauma patients with three HRI levels: I (low), II (average), or III (high). 1 To obtain FDA clearance, we independently and blindly validated the APPRAISE-HRI using trauma-registry and continuous vital-sign data collected from civilian patients in an emergency department or during pre-hospital ground- or air-ambulance transport from the point of injury to one of eight medical centres. The study outcome was haemorrhagic injury defined by documented injuries AND blood transfusion. Figure 1 shows the APPRAISE-HRI system, including a pictorial temporal representation of the device’s inputs (heart rate and blood pressure) and outputs (HRI). As part of the FDA clearance process, we performed formative and summative usability tests of the APPRAISE-HRI with U.S. Department of Defense (DoD) medics, the device’s intended users. Results We collected data from 5,895 trauma patients, 543 with haemorrhagic injuries. We used the likelihood ratio to assess whether APPRAISE-HRI could stratify haemorrhage risk and found that haemorrhagic patients were 6.88 times as likely as controls to be in level III, strongly suggesting the presence of haemorrhage in this level, and 0.18 times as likely as controls to be in level I, suggesting the absence of haemorrhage in this level. In addition, haemorrhagic patients were 0.70 times as likely as controls to be in level II. We enrolled five medics in the formative study, which allowed us to discover usability issues, and enrolled 15 medics in the summative study, where we assessed their ability to use and interpret the device results. Medics found that it was relatively easy to determine a patient’s haemorrhage risk and that the device was helpful, with 93% (14/15) correctly interpreting the outputs of the device. Discussion The DoD obtained FDA 510(k) clearance for the AI-enabled APPRAISE-HRI Class II device (K233249), 2 the first Software as a Medical Device (SaMD) approved for assessing haemorrhage risk in trauma patients. As an FDA-cleared SaMD, the APPRAISE-HRI is now available for combat medics to triage trauma casualties at the point of injury for haemorrhage risk after a physically traumatic event and stratify casualties who need immediate attention and emergency evacuation from those who may not be at risk for haemorrhage. References Stallings JD, Laxminarayan S, Yu C, et al . APPRAISE-HRI: an artificial intelligence algorithm for triage of hemorrhage casualties. Shock . 2023; 60 (2):199–205. United States Food and Drug Administration. K233249: APPRAISE-HRI [online]. 2024. https://www.accessdata.fda.gov/scripts/cdrh/cfdocs/cfpmn/pmn.cfm?ID=K233249 (accessed 27 Jan 2025). Disclaimer The opinions and assertions contained herein are the private views of the authors and are not to be construed as official or as reflecting the views of the Defense Health Agency, the U.S. Department of Defense, or The Henry M. Jackson Foundation for the Advancement of Military Medicine, Inc. This paper has been approved for public release with unlimited distribution. Disclosures A. T. Reisner has licensed software to the Nihon Kohden Corporation (Irvine, CA) for decision-support technology for the management of sepsis and circulatory shock. For the remaining authors, no conflicts are declared. Abstract A10 Figure 1 (A) The APPRAISE-Haemorrhage Risk Index (HRI) Software as a Medical Device (SaMD) cleared by the U.S. Food and Drug Administration (FDA) resides in a smartphone using Android operating system version 9 or higher. The SaMD continuously pulls and processes data from the FDA-cleared ZOLL Propaq M vital-sign monitor to generate an output every 1 minute. The SaMD displays two graphs as a function of time, one (top) shows the vital-sign values as displayed by the monitor and the other (bottom) shows the output of the device, i.e., the HRI levels (I and II in blue and III in red). (B) A pictorial temporal representation of the device’s inputs (vital signs) and outputs (HRI) during operation, with heart rates provided at 1-minute intervals (in this case) and blood pressures (systolic and diastolic) collected at varying multi-minute intervals. The software continuously processes the vital signs and generates a new HRI output whenever it receives a valid heart rate and a new pair of valid blood pressure measurements

Ähnliche Arbeiten