The potential for utilising in‐hospital glucose measurements to detect individuals at high risk of previously undiagnosed diabetes: Retrospective cohort study

BACKGROUND: Many people with undiagnosed diabetes have hyperglycaemia when admitted to hospital. Inpatient hyperglycaemia can be an indication of diabetes mellitus but can also indicate a stress response. This study reports the extent to which an in‐hospital maximum observed random glucose measurement is an indicator of the need for in‐hospital (or subsequent) HbA1c measurement to look for undiagnosed diabetes. METHODS: Blood glucose, HbA1c, age and sex were collected for all adults following admission to a UK NHS trust hospital from 1 January 2019 to 31 December 2020. We restricted the analysis to those participants who were registered with a GP practice that uses the trust laboratory and who had at least some tests requested by those practices since 2008. We stratified individuals according to their maximum in‐hospital glucose measurement and report the number of these with HbA1c measurement ≥48 mmol/mol (6.5%) prior to the index admission, and during and after admission. We calculated an estimated proportion of individuals in each blood glucose stratum without a follow‐up HbA1c who could have undiagnosed diabetes. RESULTS: In toal, 764,241 glucose measurements were recorded for 81,763 individuals who were admitted to the Oxford University Hospitals Trust. The median (Q1, Q3) age was 70 (56, 81) years, and 53% were males. Of the population, 70.7% of individuals declared themselves to be of White ethnicity, 3.1% of Asian background, and 1.1% of Black background, with 23.1% unstated. Of those individuals, 22,375 (27.4%) had no previous HbA1c measurement recorded. A total of 1689 individuals had a diabetes‐range HbA1c during or after their hospital admission (2.5%) while we estimate an additional 1496 (2.2%) may have undiagnosed diabetes, with the greatest proportion of these having an in‐hospital glucose of ≥15 mmol/L. We estimate that the number needed to detect a possible new case of diabetes falls from 16 (in‐hospital glucose 8 mmol/L to <9 mmol/L) to 4 (14 mmol/L to <15 mmol/L). CONCLUSION: The number of people who need to be tested to identify an individual who may have diabetes decreases as a testing threshold based on maximum in‐hospital glucose concentration increases. Among those with hyperglycaemia and no previous HbA1c measurement in the diabetes range, there appears to be a lack of subsequent HbA1c measurement. This work identifies the potential for integrating the testing and follow‐up of people, with apparently unrecognised hospital hyperglycaemia across primary and secondary care.