To determine whether the alveolar-arterial oxygen gradient (Grad[A-a]O<sub>2sub>) helps confirm the influence of PEEP on PaFi (PaO<sub>2sub>/FiO<sub>2sub>).
Observational study; we used linear regression to perform a multivariate study to improve the PaFi formula by taking PEEP into account.
Tertiary hospital.
We included all patients who were admitted to the intensive care unit, regardless of pulmonary damage.
We recorded personal history, clinical judgment, intensive care data, severity scores on the first day and progression. Two calculated variables: PaFi and Grad(A-a)O<sub>2sub>.
A total of 956 patients were included: 63.9%men; median age 68 years. On the first day, 31.8%did not have mechanical ventilation (MV), 13.1%had non-invasive MV and 55.1%had invasive MV. PaFi values: 32.9%0-200, 32.2%201-300, and 34.8%>300. PEEP values: 0-5 69.8%, 6-10 27.5%and >10 2.6%. We observed a correlation (Pearson) between Grad(A-a)O<sub>2sub> and PaFi of -0.84 (p < 0.001). On performing multiple regression (dependent variable: Grad[A-a]O<sub>2sub>), the following variables were included in the model: PaFi, PEEP, APACHE IV and SOFA; coefficient of determination (R<sup>2sup>) of 0.62 without PEEP and 0.72 with PEEP. We changed the PaFi formula, referring to it as PaFip (PaFi plus PEEP): Ln (PaFi/[PEEP + 12]). Correlation index between PaFip and Grad(A-a)O<sub>2sub>: -0.9 (p < 0.001). We performed linear regression (dependent variable: Grad[A-a]O<sub>2sub>) and used PaFip instead of PaFi. Only PaFi remained in the model, and was discretely complemented by APACHE IV; R<sup>2sup> = 0.8.
By adding PEEP to the PaFi model (PaFip), we clearly improve the latter, as reflected by a better goodness of fit.