New Nomogram Predicts Early Intubation Risk in COVID-19 Patients on Awake Prone Positioning

Patients with COVID-19-related acute hypoxemic respiratory failure who are treated with high-flow nasal oxygen and awake prone positioning still sometimes deteriorate and need a breathing tube, and mortality among those who are intubated remains high. A study published in Annals of Intensive Care now reports a predictive model, built from routinely measured bedside variables, designed to identify within the first 24 hours of awake prone positioning which patients are most likely to require intubation within 72 hours. The work, a secondary analysis of a prospective multicenter cohort from six intensive care units in Argentina, was conducted between June 2020 and January 2021 during the first and second waves of the pandemic.

The research team enrolled consecutive adults admitted to the ICU with COVID-19-related acute hypoxemic respiratory failure who were receiving high-flow nasal oxygen. Acute hypoxemic respiratory failure was defined by at least one of three criteria: oxygen saturation below 92 percent despite at least 4 liters per minute of supplemental oxygen, signs of increased work of breathing with a respiratory rate above 30 breaths per minute, or a PaO2/FiO2 ratio below 200 mmHg. Patients were excluded if they had respiratory failure from a non-COVID cause, contraindications to awake prone positioning such as a Glasgow Coma Scale score below 9, hemodynamic instability requiring vasopressors, pregnancy, or inability to tolerate at least 6 hours of prone positioning per day, as well as those with prior positive-pressure ventilation or a need for immediate intubation on ICU admission.

Of 1,263 patients with COVID-19 pneumonia and acute respiratory failure admitted during the study period, 728 met the eligibility criteria for high-flow nasal oxygen and awake prone positioning. After excluding patients who spent less than 6 hours per day in the prone position or had missing data, 400 patients entered the final analysis. Physiological variables were collected at two time points: at ICU admission, serving as baseline, and 24 hours after awake prone positioning began. These included respiratory rate, arterial oxygen pressure, peripheral and arterial oxygen saturations, and the fraction of inspired oxygen, from which derived indices such as the PaO2/FiO2 ratio, the SaO2/FiO2 ratio, and the ROX index were calculated.

The primary outcome was early intubation, defined as the need for invasive mechanical ventilation within 72 hours of awake prone positioning initiation. The decision to intubate rested with the attending clinical team, but standardized recommendations were provided, including deterioration of neurologic status, hemodynamic instability, or meeting two or more criteria such as a decline in oxygen saturation below 90 percent for more than five minutes, lack of improvement in respiratory muscle fatigue, inability to control airway secretions, or respiratory acidosis with a pH below 7.30. Among the 400 patients, 136, or 34 percent, required intubation within the first 72 hours, while 264, or 66 percent, remained non-intubated.

The comparison of the two groups revealed clear physiological differences. Patients who required intubation were significantly older, with a mean age of 61.5 years versus 53.4 years, and had lower baseline PaO2, at 71 versus 81 mmHg, a lower PaO2/FiO2 ratio of 125 versus 145, and a lower ROX index of 6.6 versus 7.5. The time from ICU admission to the first prone positioning session was longer in the intubated group, six hours versus two hours, and the total number of days spent in the prone position was shorter, three versus five days. After 24 hours of awake prone positioning, the intubated group showed significantly lower PaO2, at 87 versus 104 mmHg, lower PaO2/FiO2 and SaO2/FiO2 ratios, lower SpO2, a lower ROX index, and a higher FiO2 of 0.72 versus 0.63, with p values below 0.001 for these comparisons.

Mortality figures reported in the study underscored the stakes involved. By day 28, 48 of 151 intubated patients, or 31.8 percent, had died, compared with 2 of 249 non-intubated patients, or 1.0 percent. The authors note that while awake prone positioning has been shown in prior research to reduce the risk of intubation and mortality in COVID-19-related acute respiratory failure, particularly among patients receiving non-invasive ventilatory support with sufficient exposure to prone positioning, a substantial proportion of patients undergoing the maneuver are ultimately intubated anyway, and outcomes in that subgroup remain poor.

To build the predictive model, the investigators constructed two multivariable logistic regression models, one using baseline variables and another using variables measured 24 hours after awake prone positioning began. Candidate predictors were selected based on clinical relevance and univariable associations, then retained through backward elimination. At baseline, older age, with an odds ratio of 1.05 per year, lower PaO2, with an odds ratio of 0.98 per mmHg, and higher respiratory rate, with an odds ratio of 1.08 per breath per minute, were independently associated with early intubation. At the 24-hour mark, the same variables plus FiO2, with an odds ratio of 10.2, and the SaO2/FiO2 ratio were independently associated with the outcome.

The final model integrated five predictors, all measured 24 hours after prone positioning initiation: age, respiratory rate, PaO2, FiO2, and the SaO2/FiO2 ratio. From this model the team constructed a nomogram, a graphical scoring tool in which each predictor receives a weighted score proportional to its regression coefficient and the total score maps to a predicted probability of intubation through the logistic equation. An illustrative case in the paper describes a 55-year-old patient on high-flow oxygen at 60 liters per minute with an FiO2 of 0.7, a respiratory rate of 25 breaths per minute, an SaO2/FiO2 of 128, and a PaO2 of 65 mmHg; this patient’s score of 24.5 corresponded to a predicted probability of intubation above 95 percent by day 3. Discrimination was assessed with the C-statistic, calibration with calibration-in-the-large and calibration slope, and internal validation used bootstrap resampling with 1,000 iterations.

Sensitivity analyses tested alternative models suited to different monitoring scenarios, including a PaO2/FiO2-based model for settings where only arterial blood gases are available, an SaO2/FiO2-based model relying exclusively on oxygen saturation, and a model capturing 24-hour changes from baseline. Notably, a multivariable model evaluating the dynamic changes in physiological variables between baseline and 24 hours did not add predictive value beyond the models based on absolute values at each time point and was not retained. The main model showed robust and consistent discriminative performance across these sensitivity analyses, according to the authors.

The study carries important limitations that the authors themselves emphasize. The model underwent internal validation by bootstrap resampling only, and external validation in independent cohorts, including patients with non-COVID acute respiratory failure, remains necessary before the tool can be implemented in routine clinical practice. Because patients who required immediate intubation or could not tolerate at least 6 hours of awake prone positioning were excluded, the study population may underrepresent the most severely ill patients, limiting generalizability to that subgroup. The authors also acknowledge that clinically relevant factors such as signs of respiratory muscle fatigue, frailty, and nuanced clinician judgment were not captured, and that the decision to intubate is inherently variable, influenced by admission thresholds, resource availability, and clinician perception. Even so, they argue that identifying early those patients unlikely to benefit from prolonged awake prone positioning is a clinically relevant approach, given that randomized trials testing earlier versus later intubation directly would be neither feasible nor ethically appropriate, and that approximately two-thirds of patients who ultimately require intubation do so within the first 72 hours.

Subject of Research: A predictive model for early intubation in COVID-19 patients receiving awake prone positioning

Article Title: A predictive model for early intubation in patients with COVID–19–induced acute hypoxemic respiratory failure under awake prone position

Article References: Morales–Quinteros, L., Fuentes, N. A., Muriel, A., Olmos, M., Busico, M., Vitali, A., Gallardo, A., Plata–Menchaca, E. P., Ferrer, R., Artigas, A., & Esperatti, M. (2025). A predictive model for early intubation in patients with COVID–19–induced acute hypoxemic respiratory failure under awake prone position. Annals of Intensive Care, 15(1), Article 188. https://doi.org/10.1186/s13613-025-01602-4

Image Credits: AI Generated

DOI: 10.1186/s13613-025-01602-4

Keywords: COVID-19, awake prone positioning, high-flow nasal oxygen, intubation, predictive model, nomogram, acute respiratory failure, hypoxemia, intensive care, risk prediction, PaO2, respiratory rate