Introduction and objectives Community acquired pneumonia (CAP) is a leading cause of death worldwide. Early recognition of patients with CAP at risk of poor outcomes improves survival. Identification of factors independently associated with poor outcomes could inform the design of novel strategies for detecting those patients at risk of decline.

We retrospectively collected data on patients with CAP in order to ascertain which physiological, biochemical and haematological parameters at the point of admission are independently associated with 30 day mortality.

Methods Adults admitted to a large tertiary centre hospital, with CAP, between 10/2014 and 01/2016 were included. Radiology, admission clerkings and electronic patient records were reviewed to confirm pneumonia and collect comorbidity, observation and laboratory investigation data from the point of presentation to hospital. Cases were excluded if there was no radiological confirmation of CAP, or if they had hospital-acquired pneumonia. Optimal cut-off points were calculated for continuous variables, and univariate analysis performed to identify factors associated with 30 day mortality. Those found to be associated were then included in a multivariate logistic regression analysis to identify those independently associated with 30 day mortality. Statistical significance was pre-defined with a p value<0.05.

Results 1545 patients were included in the final analysis (median age 76, 30 day mortality 19.0%, 49.2% female). Univariate analysis identified age, aetiology, tachypnoea, hypotension, need for supplementary oxygen, hypothermia/hyperthermia, tachycardia, confusion, uraemia, elevated creatinine, hypoproteinaemia, hypoalbuminaemia, anaemia, neutrophilia, lymphopaenia/lymphocytosis, raised neutrophil-lymphocyte ratio, raised CRP, acidosis, hyperlactataemia, dementia, cerebrovascular disease, solid tumour malignancy and haematological diagnosis as associated with increased risk of 30 day mortality. Following stepwise, forward, logistic regression analysis, factors independently associated with 30 day mortality included age, tachypnoea, need for supplementary oxygen, hypothermia, uraemia, hypoalbuminaemia, high neutrophil-lymphocyte ratio, acidosis, hyperlactataemia and solid tumour malignancy. Presence of metabolic disease (obesity and dyslipidaemia) had a protective effect (table 1). There was no evidence of co-linearity between these variables.

Conclusions We present factors independently associated with 30 day mortality in a large cohort of CAP patients. This analysis may be helpful to inform the creation of novel severity assessment tools for CAP.