Development and validation of a nomogram to predict diabetes ketoacidosis resolution time in a tertiary care hospital in the United Arab Emirates

Abstract

Aim

This study aimed to develop and validate a nomogram to predict prolonged diabetes ketoacidosis (DKA) resolution time (DRT).

Methods

We retrospectively extracted sociodemographic, clinical, and laboratory data from the electronic medical records of 394 adult patients with DKA admitted to Tawam Hospital between January 2017 and October 2022. Logistic regression stepwise model was developed to predict DRT ≥ 24 h. Model discrimination was evaluated using C-index and calibration was determined using calibration plot and Brier score.

Results

The patients’ average age was 34 years; 54 % were female. Using the stepwise model, the final variables including sex, diabetes mellitus type, loss of consciousness at presentation, presence of infection at presentation, body mass index, heart rate, and venous blood gas pH at presentation were used to generate a nomogram to predict DRT ≥ 24 h. The C-index was 0.76 in the stepwise model, indicating good discrimination. Despite the calibration curve of the stepwise model showing a slight overestimation of risk at higher predicted risk levels, the Brier score for the model was 0.17, indicating both good calibration and predictive accuracy.

Conclusion

An effective nomogram was established for estimating the likelihood of DRT ≥ 24 h, facilitating better resource allocation and personalized treatment strategy.