Objective  To determine the association between pulse oximetry-derived oxygen saturation(SpO₂) and all-cause in-hospital mortality of patients with pulmonary embolism during oxygen therapy.

  Methods  Clinical data of the patients with pulmonary embolism as the primary diagnosis in different American medical institutions from the eICU database during 2014 to 2015 were retrieved. Generalized additive model (GAM) was constructed and the graph was drawn to analyze the association between median SpO₂ and all-cause in-hospital mortality. The lowest and smooth area of the curve was the optimal SpO₂ range. Moreover, the mult- ivariate Cox regression model was applied to verify the association between SpO₂ level during oxygen therapy and all-cause in-hospital mortality of patients with pulmonary embolism.

  Results  A total of 422 patients with pulmonary embolism that met the inclusion and exclusion criteria were enrolled. The median score of SpO₂ was 97%(95%, 98%) during oxygen therapy.A total of 336 (79.6%) patients were discharged from the hospital and 86 (20.4%) patients died. The GAM indicated a U-shaped relationship between the median score of SpO₂ and all-cause in-hospital mortality in patients with pulmonary embolism. In addition, the lowest mortality was observed when the SpO₂ range was 96%-98%. Multivariable Cox regression analysis confirmed that the SpO₂ level was independently associated with decreased mortality. Taking patients with median SpO₂ rang of 96% to 98% as controls, the risk of all-cause in-hospital mortality was increased by 129.8% in patients with median SpO₂ < 96% (HR=2.298, 95% CI: 1.268-4.163, P=0.006) and 77.3% in patients with median SpO₂ > 98% (HR=1.773, 95% CI: 1.068-2.942, P=0.027).

  Conclusions  The relationship between SpO₂ levels and all-cause in-hospital mortality followed a U-shaped curve in patients with pulmonary embolism. The risk of all-cause in-hospital mortality was lowest when SpO₂ was between 96% to 98% during oxygen therapy, which may be the target range of oxygenation.