Predicting inpatient hypoglycaemia in hospitalized patients with diabetes: a retrospective analysis of 9584 admissions with diabetes

Abstract

The inclusion of routine biochemical data, available at the time of admission, can add prognostic value to demographic and medication history. The predictive performance of the constructed model indicates potential clinical utility for the identification of patients at risk of hypoglycaemia during their inpatient stay.

Importance of inpatient hypoglycaemia: impact, prediction and prevention

“…Hypoglycaemia is a common barrier to achieving euglycaemia in the inpatient setting and is associated with significant adverse clinical outcomes for both the person with diabetes and the healthcare system. Hypoglycaemia is important to predict, but existing prediction methods are often limited by prediction accuracy, with receiver‐operating curve analysis showing an area under the curve of 0.73 . In other studies, the failure rate to predict the hypoglycaemic events in the original analysis was as high as 40% .…”

“…Stuart et al . reported that confirmed inpatient hypoglycaemia could be predicted by medication (type of insulin and use of sulfonylureas) .…”

“…The model developed by Stuart et al . identified estimated GFR, C‐reactive protein, sodium and albumin as factors in predicting inpatient hypoglycaemia. In these studies, biochemical factors that were not found to be predictive included hyperkalaemia, elevated neutrophil count, haemoglobin level and anaemia.…”

Temporal and external validation of a prediction model for adverse outcomes among inpatients with diabetes

“…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]. 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].…”

“…However, we have shown that variables in the model can predict hypoglycaemia [28] and it is therefore reasonable to hypothesize that a care pathway with risk stratification can enable harm reduction and better clinical endpoints for inpatients with diabetes. Another criticism is that most covariates and the outcomes in the model are not diabetesspecific, and the model may therefore perform well in other chronic conditions to predict length of stay and mortality.…”