Establishment of multiple predictive models for early identification of the dengue cases in outpatient healthcare systems.

Dengue affected countries have responded with many surveillance strategies to detect outbreaks early. However, the early identification for the individual case before a confirmed dengue diagnosis is still lacking, limiting the effectiveness of early intervention strategies. This case-control study involved 206 dengue cases and 824 controls, which selected from patients who visited outpatient clinics at any medical institutions affiliated to the Futian District of Shenzhen city, China from January 1, 2018 to December 15, 2024. All medical visit information was obtained from the "Futian District Population Health Information Platform". Multiple predictive models for early identification of the dengue cases were established based on the logistic regression analyses. C-index was used to evaluate the accuracy. 1030 study subjects were randomly divided into a training set and a validation set at a ratio of 7:3. In the training set, three separate multivariable early case identification models were constructed across diverse data availability scenarios. Model 1 incorporated symptom descriptions with a C-index of 0.793. Model 2 incorporated symptom descriptions and epidemiological exposure factors (C-index: 0.844). Model 3 incorporated symptom descriptions and complete blood count parameters (C-index: 0.973). In the validation set, the C-index of the three models were 0.779, 0.867 and 0.957, respectively. Based on Chinese outpatient diagnostic data, this study established early identification models for dengue cases with high predictive accuracy. These models can be integrated into clinical decision support systems to alert high-risk patients and enable self-assessment tools for community members.