A pragmatic calibration of the ROX index to predict outcome of nasal high-flow therapy in India.

ABSTRACT

PURPOSE: Identifying thresholds at which the ROX index would satisfactorily predict HFNC failure across heterogenous resourced contexts is necessary for clinical use. METHODS: Patients ≥18 years admitted to 30 diverse ICUs across 10 states in India who required HFNC for respiratory support were retrospectively included in this study. Patient data and hourly ROX indices were then analyzed and contextualized to clinical outcomes as well as with ROX index thresholds in other regions of the world. MEASUREMENTS AND MAIN RESULTS: Among the 614 patients included, 276 (44.9%) required respiratory escalation. Pneumonia was the most common diagnosis on admission. Receiver operating characteristic curve analysis revealed that a baseline ROX score of 7.86 was similar to 4.88 in other populations which was confirmed by Kaplan-Mier curves (hazard ratio,3.58 (2.72-4.69, p < 0.001)). ROX scores at 11.84 or 5.89 had roles in screening and confirming HFNC failure. The index performed poorly in a subset of patients who died without respiratory escalation. The ROX index was most predictive in the initial hours of ICU admission and a longer duration of HFNC was associated with more severe outcomes. CONCLUSIONS: When optimally calibrated this index can using a method that can reliably predict the risk of HFNC failure among ICU patients from diverse settings.