The prediction of early mortality by the ROX index of oxygenation and respiratory rate in diverse Canadian and Ugandan cohorts of unselected patient: a post-hoc retrospective analysis of 80,558 patient observations.

AIM To investigate the association between in-hospital mortality and the ROX index of respiratory rate and oxygenation in diverse cohorts of unselected patient at different prediction windows. METHODS A retrospective post-hoc analysis of data from a major regional referral Canadian hospital and a low-resource hospital in sub-Saharan Africa. RESULTS Four patient cohorts were examined: Canadian medical, surgical and intensive care unit (ICU) patients, and all patients admitted to an African hospital. In all patients in-hospital mortality rose as ROX declined. Apart from ICU patients, ROX had a high discrimination for death within 72 hours. For non-ICU patients the negative predictive value of death within 72 hours for a ROX value <22 ranged from 0.994 to 1.000 Conclusion: In diverse cohorts of unselected patients, the ROX index has a high discrimination for death within 72 hours. However, the index has little or no prognostic value for patient admitted to ICU.