Predicting high perceived stress in late adolescence: development and validation of a prognostic model

Stress levels have increased among adolescents, yet tools for identifying those at high risk remain limited. This study aimed to develop and internally validate a prognostic model estimating the probability of high perceived stress in young adulthood using data collected at age 15.

Data were from the West Jutland Cohort Study including 2,108 Danish adolescents with complete information on the 4-item Perceived Stress Scale at ages 15 and 18. High stress at age 18 was defined as a total score of 9 or higher. Twenty-three candidate predictors were examined using Least Absolute Shrinkage and Selection Operator (LASSO) regression across 100 imputed datasets. Predictors selected in ≥ 80% of imputations were entered into a pooled logistic regression model. Model performance was evaluated using 10-fold cross-validation under multiple imputation. Discrimination was assessed with the area under the receiver operating characteristic (ROC) curve (AUC), and calibration with calibration-in-the-large (CITL) and slope.

Three predictors at age 15 were retained: perceived stress, depressive symptoms, and self-esteem. The model showed good calibration (CITL = 0.003; slope = 1.008; expected-to-observed ratio = 0.998) and acceptable discrimination with an AUC = 0.72 (0.68;0.76). The Brier score was 0.081, indicating satisfactory overall predictive accuracy. The model showed high negative predictive values but modest positive predictive values.

A simple model based on self-reported psychosocial measures in mid-adolescence predicted later high stress with good calibration and moderate discrimination. External validation is needed to assess generalisability and inform potential use as a low-burden decision-support tool in school health settings.