(1043) A 29-mRNA Host Response Test Reduces False Negatives and Accurately Predicts Infection Severity

Introduction: Current approaches for predicting severity of illness in patients with suspected infection, which include vital signs, biomarkers, and scoring systems have lower accuracy, limited generalizability across healthcare settings or lack prospective validation on a point-of-care platform. Here, we validated the severity classifier version 4 (SEV4) of the TriVerity™ Acute Infection and Sepsis Test, in development, on the Myrna™ Instrument, designed to predict the need for ICU-level care.

Methods: SEV4 was trained on a diverse dataset of 2,758 samples from 38 studies across various geographies, demographics, clinical settings, and platforms, exclusively using 28- or 30-day mortality outcomes. The prospective validation on the Myrna bench-top analyzer involved 850 samples across 7 cohorts to predict mortality and the need for 7-day ICU-level care, defined as the need for admission, mechanical ventilation, vasopressors, renal replacement therapy.

Results: In training and validation data, SEV4 had an AUROC of 0.89 and 0.79, respectively, for predicting mortality. In validation data, SEV4 had an AUROC of 0.77 for predicting the need for ICU-level care over 7 days. 15 patients in the validation set that presented in ED died without ICU care, which was explained by a weak association between ICU-level care and mortality endpoints in the validation set (Phi coefficient=0.28). SEV4 correctly identified 80% (12/15) of these patients as moderate-to-very-high severity, indicating its potential to identify patients at risk of severe outcomes that would otherwise be missed. SEV4 also had a higher AUROC than other markers for both mortality (SEV4:0.75 vs lactate:0.64; SEV4:0.81 vs qSOFA:0.69) and ICU level care (SEV4:0.70 vs lactate: 0.65; SEV4:0.74 vs qSOFA:0.70).

Conclusions: SEV4 accurately predicts both 30-day mortality and 7-day need for ICU-level care, highlighting its practical utility given the scarcity of high-quality ICU-level annotations compared to readily available mortality data. SEV4 reduced false negatives in the ED by correctly identifying patients who died but did not receive ICU level care. Finally, the higher accuracy of SEV4 over traditional markers such as lactate and qSOFA further underscores the clinical utility of the TriVerity test.