A Novel Depression Risk Prediction Model Using NHANES Data With Mendelian Randomization Validation

Despite depression's significant public health impact, efficient and accessible screening tools utilizing routine clinical indicators remain limited. This study aimed to develop and validate a practical depression risk prediction model based on commonly available biochemical markers, facilitating widespread early screening and timely intervention in general clinical settings.

We formulated a model for depression, scrutinizing an assortment of biochemical indicators and their bidirectional interrelationships with depression, employing data derived from the National Health and Nutrition Examination Survey (NHANES) and leveraging the Mendelian randomization (MR) approach, a method that utilizes genetic variants as instrumental proxies to ascertain causal nexus between risk determinants and diseases.

Using NHANES data (training cohort: n = 27,327; validation cohort: n = 4383), we developed two prediction models through LASSO and multivariate logistic regression. Both models demonstrated comparable performance in terms of discrimination (ROC curves), calibration (slope and Hosmer-Lemeshow test), Brier score, decision curve analysis, net reclassification improvement, and integrated discrimination improvement. Given the similar performance metrics and more parsimonious nature, Model 2, with 14 variables, was selected as the final model. MR analysis revealed bidirectional relationships between biomarkers and depression. Higher body mass index level was associated with increased depression risk (odds ratio [OR]: 1.061, p = 0.008). Depression itself showed significant associations with increased ALP (OR: 1.048, p = 0.010), decreased BUN (OR: 0.966, p = 0.032), and TB (OR: 0.963, p = 0.044) levels.

Model 2, selected for its predictive accuracy and streamlined complexity, presents a pragmatic instrument for large-scale population screenings, facilitating timely intervention and therapeutic strategies.