ABSTRACT
Respiratory failure (RF) is frequent in hospitalized older patients, but was never systematically investigated in large populations of older hospitalized patients. We conducted a retrospective administrative study based on hospitalizations of a Geriatrics Unit regarding 2014, 2015, and 2016. Patients underwent daily screening for hypoxia. Hospital discharge records were coded through a standardized methodology. RF, defined as documented hypoxia on room air, was always coded, whenever present. We investigated how RF affected clinical outcomes, whether RF grouped into specific comorbidity phenotypes, and how phenotypes associated with the outcomes. RF was coded in 48.6% of the 1,810 hospitalizations. RF patients were older and more frequently had congestive heart failure (CHF: 49 vs 23%), chronic obstructive pulmonary disease (COPD: 27 vs 6%), pneumonia (14 vs 4%), sepsis (12 vs 7%), and pleural effusion (6 vs 3%), than non-RF patients. RF predicted longer length of stay (a-Beta 2.05, 95% CI 1.4-2.69; p < 0.001) and higher in-hospital death/intensive care units (ICU) need (aRR 7.12, 5-10.15; p < 0.001) after adjustment for confounders (linear and Poisson regression with robust error variance). Among RF patients, cerebrovascular disease, cancer, electrolyte disturbances, sepsis, and non-invasive ventilation predicted increased, while CHF and COPD predicted decreased in-hospital death/ICU need. The ONCO (cancer) and Mixed (cerebrovascular disease, dementia, pneumonia, sepsis, electrolyte disturbances, bedsores) phenotypes displayed higher in-hospital death/ICU need than CARDIO (CHF) and COPD phenotypes. In this study, RF predicted increased hospital death/ICU need and longer hospital stay, but also reflected diverse underlying conditions and clinical phenotypes that accounted for different clinical courses.