APPRAISE-HRI

The APPRAISE-HRI is a mobile health application designed to aid military healthcare providers in rapidly assessing hemorrhage risk in traumatically injured U.S. Service members by providing an adjunctive indicator. It employs an artificial intelligence linear regression model that analyzes continuously streamed vital signs (heart rate, systolic and diastolic blood pressure) from an external monitor, stratifying patients into distinct hemorrhage risk levels to inform situational awareness and clinical management priorities.

APPRAISE-HRI is a mobile health application developed for military healthcare providers to screen U.S. Service members for hemorrhage risk following a physically traumatic event. Its primary purpose is to stratify casualties, identifying those requiring immediate medical attention and emergency evacuation versus those who are injured but may not be at immediate risk of hemorrhage, thereby providing enhanced situational awareness to inform clinical management. The Android-based application continuously collects heart rate, systolic blood pressure, and diastolic blood pressure from a compatible external monitor, currently the ZOLL Propaq M, via Bluetooth, generating a hemorrhage risk score every minute. The core of APPRAISE-HRI is a three-module algorithm: it first pre-processes the raw vital sign data, then employs an artificial intelligence linear regression model, and finally stratifies the model's output into three Hemorrhage Risk Index (HRI) Levels – Low (I), Average (II), or High (III). Clinical validation, using retrospectively collected vital sign data from adult trauma patients, demonstrated its utility; for instance, patients with HRI Level III were nearly seven times more likely to be actual hemorrhage cases compared to controls, while those with HRI Level I were over five times less likely. While generally consistent across subgroups, performance in ruling in hemorrhage for penetrating injuries was noted as an area for further investigation due to limited data in that specific patient population.

AI Regulatory Concerns

Powered by

Jev reads the full FDA letter and returns calibrated concern probabilities per regulatory category. These are AI estimates, not regulatory advice.

• Q1. How does the APPRAISE-HRI's AI linear regression model specifically weigh each vital sign (HR, SBP, DBP) in its risk calculation, and are there plans to incorporate additional physiological parameters or battlefield-relevant inputs (e.g., mechanism of injury, estimated blood loss at scene) to enhance predictive accuracy?; What were the specific sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) for each HRI level in the clinical validation study, particularly for distinguishing true positives in the "High" risk category and true negatives in the "Low" risk category?; Given its continuous one-minute update cycle, how does the system address potential artifacts or transient fluctuations in vital sign data that might lead to temporary misclassification of hemorrhage risk, and are there smoothing algorithms or alert-fatigue mitigation strategies in place?; The clinical validation used retrospectively collected data; what are the plans for prospective, real-world operational testing within active military units to assess its performance, usability, and impact on patient outcomes in actual field conditions?; Considering the "Average (II)" risk category showed a likelihood ratio of 0.70 (nearly as likely for hemorrhage as for control), what is the recommended clinical pathway or further diagnostic steps for patients falling into this intermediate HRI level to ensure timely and appropriate care?