# Identification of early predictors and model for bacterial infection in diabetic ketoacidosis patients: A retrospective study

## TL;DR

This study identifies CRP and cortisol as early predictors of bacterial infection in diabetic ketoacidosis patients and develops a diagnostic model combining these markers for improved accuracy.

## Contribution

A novel diagnostic model combining CRP and cortisol for early detection of bacterial infection in DKA patients is proposed.

## Key findings

- CRP and cortisol independently predict bacterial infection in DKA patients with high sensitivity and specificity.
- A combined diagnostic model using CRP and cortisol achieves an AUC of 0.930, significantly improving diagnostic accuracy.
- The model demonstrates 93.5% sensitivity and 80.0% specificity for bacterial infection detection in DKA.

## Abstract

The purpose of this report was to identify effective indicators capable of predicting bacterial infection during the early stages of diabetic ketoacidosis (DKA) and to establish a diagnostic model suitable for clinical application.

This was a retrospective cross-sectional study. Between February 2018 and May 2023, Yuhuangding Hospital admitted 101 DKA patients, of whom 45 were diagnosed with bacterial infections. A confirmed bacterial infection was defined as documented bacteriological evidence in any bacterial sample. Clinical parameters and biological markers (including cortisol, C-reactive protein (CRP), procalcitonin, etc.) were recorded during the initial DKA phase. Multivariate regression analysis was employed to construct a diagnostic model.

CRP (OR = 1.014, 95% CI: 1.002–1.026, p = 0.017) and cortisol (OR = 1.007, 95% CI: 1.002–1.012, p = 0.003) were found to have an independent association with bacterial infection in DKA patients. The area under the receiver operating characteristic curve (AUC) for CRP in identifying bacterial infection was 0.855 (95% CI, 0.771–0.917), with a sensitivity of 76.1% and a specificity of 83.6%. The AUC for cortisol in identifying bacterial infection was 0.847 (95% CI, 0.761–0.911), with a sensitivity of 71.7% and a specificity of 89.1%. A joint diagnostic model based on cortisol and CRP was developed through multifactor regression analysis. The AUC of this diagnostic model was 0.930 (95% CI, 0.862–0.972), resulting in a sensitivity of 93.5% and a specificity of 80.0%.

CRP and cortisol are early indicators of bacterial infection in DKA patients. Furthermore, based on their combination, the regression diagnostic model exhibits enhanced diagnostic performance.