Predictive modeling of the ROX index in high-flow nasal cannula use for COVID-19 ARDS: Panel and functional data
V Diana Sotelo1, Henry Mauricio Parada-Gereda2, M Pilar Mogollón1
1Clínica Infantil Santa María del Lago, Clínica Colsanitas. Grupo de investigación en Nutrición Clínica y Rehabilitación, Grupo Keralty, Bogotá, Colombia.
Background:
Acute respiratory distress syndrome (ARDS) associated with COVID-19 has led to an increased use of high-flow nasal cannula (HFNC) as a noninvasive respiratory support strategy. However, identifying factors of HFNC failure remains a clinical challenge. This study integrates panel data analysis and functional data modeling to assess predictors associated with HFNC failure and to optimize its use in COVID-19-related ARDS.
Methods:
A retrospective cohort study was conducted in patients with COVID-19-related ARDS treated with HFNC at a high-complexity institution in Bogotá. Clinical and physiological variables were analyzed using logistic regression with panel data and functional data models to identify predictors of HFNC failure. Model discrimination was evaluated through ROC curves, while performance was validated using accuracy and concordance metrics.
Results:
A total of 361 patients with 1900 repeated measurements were analyzed. In the panel data model, higher FiO2 (OR: 1.03; 95 % CI: 1.01-1.05; p = 0.01), APACHE II score (OR: 1.09; 95 % CI: 1.05-1.13; p = 0.001), and obesity (OR: 1.60; 95 % CI: 1.09-2.34; p < 0.05) were associated with HFNC failure. Conversely, the ROX index (OR: 0.85; 95 % CI: 0.76-0.96; p = 0.01) and prone positioning (OR: 0.30; 95 % CI: 0.20-0.46; p = 0.001) were protective factors (AUC: 0.78). The functional data model confirmed the significant influence of FiO2, PaO2/FiO2, prone positioning, and the ROX index on classification performance (AUC: 0.81). Subgroup analysis revealed that, among obese patients, diabetes and higher APACHE II scores were associated with HFNC failure, whereas a higher ROX index and prone positioning were protective factors (AUC: 0.83). In COPD patients, older age and lower PaO2/FiO2 increased the risk of failure (AUC: 0.97).
Conclusions:
Key factors associated with HFNC failure in COVID-19 ARDS were FiO2, PaO2/FiO2, prone positioning, and ROX index. Combining panel data and functional data analysis improved predictive accuracy, highlighting the potential role of obesity in modifying the response to HFNC therapy.
