Bedtime prediction of nocturnal hypoglycemia in insulin treated type 2 diabetes patients

Research output: Contribution to journalJournal articleResearchpeer-review

1 Citation (Scopus)


Background and aims: Hypoglycemia may lead to anxiety, poor adherence, and hypoglycemia unawareness and is especially a threat during the night in patients with insulin-treated type 2 diabetes (T2D). It would therefore be beneficial to warn patients at risk of hypoglycemia at bedtime so they can react accordingly and avoid the episode. Hence, the aim of the present study was to develop a model for predicting nocturnal hypoglycemia.

Methods: Continuous glucose monitoring (CGM), mealtime, and insulin data were collected from 67 insulin-treated patients with T2D (NCT01819129). Data were structured into 24-hour periods and labeled as nocturnal hypoglycemia or not depending on whether 15 consecutive minutes were spent below 3.0 mmol/L (54 mg/dL) during the following night. Each period was divided into "last night," "morning," "day," and "evening" for feature extraction purposes, and 72 potential features were extracted for every period. A five-fold cross-validation was used to select features by forward selection and for training and validating a model based on logistic regression.

Results: The prediction model was based on 30 patients with 60/496 periods resulting in nocturnal hypoglycemia. Forward selection revealed that the best features were based on CGM and involved the last value and mean value during the evening, as well as the relative difference in maximum value during the day between the present period and previous periods. The model obtained a mean area under the receiver operating characteristics curve (AUC) of 0.82 with an accuracy of 0.79.

Conclusions: The model was able to predict nocturnal hypoglycemia with an acceptable accuracy and could therefore prevent such cases.
Original languageEnglish
JournalJournal of Diabetes Science and Technology
Issue number3
Pages (from-to)592-597
Number of pages6
Publication statusPublished - May 2024


  • blood glucose monitoring
  • hypoglycemia
  • prediction
  • type 2 diabetes


Dive into the research topics of 'Bedtime prediction of nocturnal hypoglycemia in insulin treated type 2 diabetes patients'. Together they form a unique fingerprint.

Cite this