Novel screening tool for psychiatric samples with major depressive disorder: a two-centre study

Major depressive disorder (MDD) poses a significant global public health challenge. Although the Patient Health Questionnaire-9 (PHQ-9) is widely used for MDD screening, its validation in psychiatric samples is limited. Additionally, variability in cut-off values across populations may lead to diagnostic inaccuracies. This study aims to develop a highly accurate nomogram tailored to improve MDD screening precision in psychiatric samples.

From March 2024 to February 2025, participants were consecutively recruited from psychiatric outpatient clinics at Xijing Hospital and the 904 Hospital, assigned to training and validation cohorts. Using Least Absolute Shrinkage and Selection Operator regression for dimensionality reduction and predictor identification, we then weighted key factors via multivariable logistic regression to construct a clinically interpretable nomogram. The model's discriminative ability, calibration, clinical utility and improvement metrics underwent rigorous evaluation.

In the training cohort (n=519) and validation cohort (n=225), 53.95% and 53.33% of participants were diagnosed with MDD. We constructed a nomogram using 11 predictors selected from 20 variables. In internal validation, the nomogram's area under the curve (AUC) (95% CI) surpassed the PHQ-9's (0.932 (0.910-0.954) vs 0.874 (0.844-0.905), p<0.001). Similarly, in external validation, the nomogram's AUC (95% CI) was higher than the PHQ-9's (0.935 (0.902-0.968) vs 0.888 (0.844-0.932), p<0.001). Moreover, the nomogram demonstrated better performance than the PHQ-9 in calibration and clinical utility during both internal and external validation. These improvements suggest that the nomogram has the potential to reduce false-positive and false-negative rates in MDD screening.

This study developed a novel tool for MDD screening in psychiatric samples by integrating weighted relevant factors, which demonstrates significantly improved accuracy in identifying MDD.