Abstract:Background: Early screening of adolescent depression is a high public health priority worldwide. Few existing methods are available to provide targeted estimation of depression risk in subgroups with different characteristics. The current study aimed to identify characteristics of adolescents at high risk of depression, and define subgroups to build group-specific models that are optimized to detect adolescents with a current, high risk of depression. Methods: We developed the Interpretable Search Tree family of algorithms (ISTs) that apply subgroup analysis to predict the absence or presence of depression risk (Patient Health Questionnaire-2 [PHQ-2] score <3 or PHQ-2 ≥3, respectively) for a large-scale population of Chinese adolescents (n=414188) from the 2021 Tsinghua Adolescence Health Survey (TAHS). In addition to identifying subgroups with the highest risk of depression using IST-Prevalence (IST-P), we were able to segment the whole population of adolescents with meaningful sets of feature-value pairs using IST-Learnability (IST-L) to build depression prediction models that are subgroup-specific. Findings: We have identified a large number of subgroups with elevated risk of depression and retrieved the best features for subgroup divisions. We also found that it is possible to retain only the top 10% to 30% of best predictors (N = 4 and N = 10, respectively) for most subgroups without sacrificing model performance. Models developed for the subgroups improved substantially from the baseline models developed for the whole population. Interpretation: Our findings supported that subgroups of adolescents with divergent individual, family and social backgrounds are at remarkably different levels of depression risk, and that developing subgroup-specific models with easily measurable features can reduce the costs of screening implementation while providing more reliable estimations of depression risk in adolescents. Funding: The study was supported by Tsinghua University Spring Breeze Fund (Project ID: 2020Z99CFG013).Declaration of Interest: We declare no competing interests.Ethical Approval: TAHS was approved by the Ethics Committee from the Department of Psychology at Tsinghua University.

Early Screening of Adolescents at Risk for Depression: an Efficient Machine-Learning-Based Identification and Subgroup Discovery System

Validity Study of Patient Health Questionnaire-9 Items for Internet Screening in Depression among Chinese University Students

Subphenotyping Depression Using Machine Learning and Electronic Health Records

Comparison of three machine learning models to predict suicidal ideation and depression among Chinese adolescents: A cross-sectional study

Exploration of Adolescent Depression Risk Prediction Based on Census Surveys and General Life Issues

Prediction of adolescent depression from prenatal and childhood data from ALSPAC using machine learning

A novel machine learning approach to shorten depression risk assessment for convenient uses

Prediction of the trajectories of depressive symptoms among children in the adolescent brain cognitive development (ABCD) study using machine learning approach

Identifying the risk of depression in a large sample of adolescents: An artificial neural network based on random forest

Research on a New Intelligent and Rapid Screening Method for Depression Risk in Young People Based on Eye Tracking Technology

Risk Prediction Model for Non-Suicidal Self-Injury in Chinese Adolescents with Major Depressive Disorder Based on Machine Learning

Machine learning-based discrimination of unipolar depression and bipolar disorder with streamlined shortlist in adolescents of different ages

Federated learning-based prediction of depression among adolescents across multiple districts in China

Design and Preliminary Realization of a Screening and Early Warning Health Management System for Populations at High Risk for Depression

Predictive factors for the development of depression in children and adolescents: a clinical study

Machine learning identifies different related factors associated with depression and suicidal ideation in Chinese children and adolescents

Prediction of non-suicidal self-injury (NSSI) among rural Chinese junior high school students: a machine learning approach

A case for the use of deep learning algorithms for individual and population level assessments of mental health disorders: Predicting depression among China's elderly

Predicting suicidal thoughts and behavior among adolescents using the risk and protective factor framework: A large-scale machine learning approach

The development and validation of an algorithm to predict future depression onset in unselected youth

Predicting non-suicidal self-injury among Chinese adolescents: The application of ten algorithms of machine learning