Mortality, morbidity and temporal patterns of glycaemic measurements in populations with diabetes

People with diabetes mellitus account for up to 25% of a hospital population despite being only 5-8% of the general population. It has been demonstrated that the length of hospital stay and mortality is increased in patients with diabetes. We used a database of all inpatient capillary blood glucose testing performed in Greater Glasgow and Clyde Hospitals and the national database of all patients with diabetes (SCI-diabetes) to explore the relationship between inpatient glucose outcomes. This then led to a wider exploration of temporal patterns of markers of glucose control and clinical outcome. We analysed a large inpatient capillary blood glucose dataset to confirm that hypoglycaemia was common and particularly so overnight. We also inferred from analysis of our data that most hypoglycemia was being uncovered by routine testing and that there was a high chance of significant undiscovered hypoglycaemia. We then devised a novel metric of hypoglycaemia treatment quality (time to repeat blood glucose (TTR)). We used TTR as a marker of nationally agreed standards of care and showed poor adherence to guidelines and demonstrated that a quality improvement package could produce a sustained improvement in this metric. Our investigation of inpatient hypoglycaemia management led us to reveal that quality of hypoglycaemia care varied with the type of diabetes and diabetes treatment modality, with patient groups most at risk of harm having the least good treatment. We noticed that a highly cited paper was potentially overestimating the association between inpatient hypoglycemia and length of hospital stay as it did not take into account that the more measurement is made across a given distribution the more likely extreme results will be found. We, therefore, used a technique which compared a mathematically modelled inpatient dataset with a real-world dataset. This confirmed the association between hypoglycaemia and length of stay remained but had been overestimated. This work enables more accurate health economic modelling of the impact of diabetes in hospital systems. During our investigation of inpatient glucose patterns and hypoglycaemia we became interested in whether the variability of glucose per se rather than high or low glucose was clinically important. We demonstrated that inpatient glucose variability is independently associated with long term mortality. We have also confirmed the association between glucose variability patients following acute stroke and renal replacement therapy. Our work on inpatient glucose variability led us to explore whether the variability of other longer-term indices of glucose control might be associated with poor outcome. We demonstrated an independent association between HbA1c variability and mortality in type 1 diabetes. We have since also shown that this association is independent and additive to an association between mortality and variability in visit-to-visit systolic BP. These data could be used to infer that the finding is not simply a marker of poor treatment concordance or a chaotic lifestyle. We have also shown that variability of HbA1c can be reduced by attending a structured patient education course, showing that variability of HbA1c is also a potentially modifiable risk factor for poor outcome.