Hemoglobin - to - red cell distribution width ratio in depression symptoms: Threshold effects and metabolic - inflammatory mediation revealed by multimodal machine learning and symptom network analysis in 196,260 adults.

OBJECTIVE

To investigate the association between hemoglobin to red blood cell distribution width ratio (HRR) and depression symptoms.

METHODS

This study enrolled 196,260 participants from Nanjing Gaochun People's Hospital between 2020 and 2024, using latent class analysis (LCA) to classify depression symptoms patterns, employing six machine learning (ML) algorithms to evaluate HRR's predictive value, and implementing symptom network (SN) analysis to explore the pathways of HRR's mechanism of action.

RESULTS

Following the integration of key metabolic indicators including body mass index (BMI) and dyslipidemia, the gradient boosting decision tree model demonstrated the optimal predictive performance (The accuracy of the internal validation reached 0.6988, with a sensitivity of 0.8952, a specificity of 0.6824, an F1 score of 0.7540, and an AUC of 0.824. The accuracy of the external validation reached 0.6702, with a sensitivity of 0.6060, a specificity of 0.8743, an F1 score of 0.6054, and an AUC of 0.747), achieving a contribution score of 0.7 under HHR ≤ 8 conditions. In the metabolic-inflammatory pathway analysis, BMI exhibited significant bridge strength (1.08285495) while dyslipidemia showed prominent mediation centrality, indicating their critical mediating roles in the biological pathway.

CONCLUSIONS

HRR was significantly correlated with the risk of depression symptoms, suggesting that it may have the value of serving as a potential predictive biomarker. BMI and dyslipidemia played a role in the depression-inflammation/metabolic pathway, indicating the possibility of them being potential intervention targets.