Introduction Doctors, patients and carers need realistic expectations of outcome when considering percutaneous endoscopic gastrostomy (PEG) placement but large-scale studies of patient factors associated with mortality risk are lacking. Hospital Episode Statistics (HES) contain details of all episodes of care undertaken in NHS hospitals in England. We linked this data set to death registrations to perform a national-level analysis.

Methods Aims:(a) Determine feasibility of analysing post-PEG mortality (at 7 d and 30 d) using HES data linked to Office of National Statistics (ONS) death certification; b) Identify factors associated with mortality. We linked HES data for 2006/7 and 2007/8 to ONS status. Care episodes (incl. day cases) containing OPCS-4 codes for gastrostomy-related interventions were extracted and deaths identified. Diagnostic fields (ICD-10 codes) were analysed for indication and comorbidity. Factors associated with death were identified by univariate and multiple logistic analyses. Crude mortality rate at each hospital Trust was analysed vs PEG number to explore “volume” effect.

Results 30 781 patients had gastrostomy-related procedure codes of which 14 055 were coded with G44.5 (“fibreoptic endoscopic percutaneous insertion of gastrostomy (PEG)”). Excluding cancer (eg, PEG pre-op for ENT surgery) yielded 10 952 PEG patients (mean age 68.4, range 16–99 years); male, 51%; emergency admission, 72%; stroke, 40%; motor neurone disease (MND), 7.2%; Parkinson's disease, 5.4%; multiple sclerosis, 4.9%; dementia, 7.2%). All-cause mortality rate: 4.2% (7 d) and 14.6% (30 d). Binary logistic regression identified predictors of 30 d death (OR with 95% CI, p<0.03 for all): age, 6.8 (5.3 to 8.7) for 85+ years vs <55 years; male sex, 1.3 (1.1 to 1.4) vs female; emergency admission, 2.2 (1.9 to 2.6) vs elective; MND diagnosis, 1.3 (1.1 to 1.7) vs not MND; dementia diagnosis, 1.2 (1.03 to 1.5) vs not dementia. In elderly patients (85+) during emergency admissions, dementia was associated with 7 d mortality of 14.6% compared to 7.8% in stroke. No correlation for 30 d death vs PEG volume at each NHS trust (Pearson r=0.04).

Conclusion This national-scale linkage analysis provides real-world estimates of prognosis in subjects selected for PEG in England and identifies factors associated with all-cause death. Elderly subjects with dementia are a high-risk group and present an ethical dilemma for PEG placement. This is the largest-scale UK-based study available for all-cause mortality after PEG and provides a model for predicting outcome using simple variables. Hospital-level data will be shared with front-line teams to seek local interpretation of the data.