Glucose levels at hospital admission are associated with 5 year mortality

Abstract

Aim

We aimed to determine if hospital admission hyperglycaemia and hypoglycaemia are associated with increased long-term mortality.

Methods

A post-hoc analysis of data from a trial of glucose screening in the emergency department was conducted. Data were linked with a death registry up to 5 years after admission. The relationship between admission glucose and mortality was examined by cox regression. Further analyses of people who survived the admission and subsequent 28 days was performed.

Results

There were 131,322 patients, of whom 38,712 (29.5 %) died. Mean follow-up was 3·3 ± 1·5 years. Compared to the reference glucose band of 6·1-8·0 mmol/L, there was increased mortality in higher bands, reaching a hazard ratio (HR) of 1·44 (95 %CI 1·34-1·55, p < 0·001) for people with glucose > 20·0 mmol/L. The HR was 1·56 (95 %CI 1·46-1·68, p < 0·001) for people with glucose ≤ 4·0 mmol/L. Similar relationships were observed among 28-day survivors. The relationships were attenuated among people with known diabetes.

Among 4867 subjects with glucose ≥ 14·0 mmol/L, those diagnosed with diabetes during the admission had lower mortality compared to subjects where the diagnosis was not made (HR 0·53, 95 %CI 0·40-0·72, p < 0·001). This was attenuated among 28-day survivors.

Conclusion

Hyperglycaemia and hypoglycaemia on hospital admission are associated with increased long-term mortality.