“…Overall, the results of the present analysis show that the model is both temporally and externally valid: it performed well in both more recent data from the original hospital (UHB) and in other hospitals with sociodemographically The strengths of the present study include the external validation across two distinctly different hospital settings, datasets with large numbers of patients and events, robust methodology including a number of sensitivity analyses, and being the first study to derive and validate a prediction model to identify adverse events in hospitalized patients with diabetes based on routinely collected data close to the time of admission. Only two other algorithms that aim to identify patients at risk of hypoglycaemia exist for inpatients with diabetes, but these were not externally validated [27,28]. Validation of the model is, however, limited to hospitals in the UK, and data from Ipswich Hospital had limitations that prevented us from optimally demonstrating the performance in a district general hospital setting.…”