Comorbidity and intensive care outcome — a multivariable analysis

Decisions regarding admission to intensive care are made considering both the physiological state of the patient and the burden of comorbidity. Despite many retrospective cohort studies looking at isolated comorbidities, there has been little work to study multiple comorbidities and their effect upon intensive care outcome. In this retrospective cohort analysis, detailed comorbidity and demographic data were gathered on 1,029 patients from the West of Scotland and matched to both unit and hospital mortality at 30 days. Logistic regression was performed to investigate the factors associated with death within 30 days at both hospital and unit level. Variables with a p-value <0.25 at the univariable level were considered in a multivariable model. Variable selection for the multivariable modelling was carried out using backward selection and then replicated using forward selection to check for model stability. A modelling tool was constructed for both unit and hospital mortality at 30 days. This modelling has shown significant odds ratios for hospital death for alcoholic liver disease (OR 4.83), age (1.03), rheumatological diseases (1.93) and functional exercise tolerance prior to admission (3.08). Results from this work may inform a national prospective study to validate the modelling tool on a wider populatio