Symptom trajectories and their predictive factors in psychiatric inpatients with depression

Depression presents with heterogeneity in symptom trajectories, complicating individualized treatment. Identifying distinct classes of symptom trajectories and their predictors may enable earlier intervention for those at greatest risk of poor outcomes.

We analyzed 620 inpatients with depression drawn from two independent naturalistic samples. Using Growth Mixture Modeling, latent classes of symptom trajectories were identified. Using baseline clinical scores and anamnestic data, we trained a regression model to predict membership in the non-responding class in the main sample and applied the model to the replication sample.

Three distinct classes of symptom trajectories were identified across samples: Class 1 (46 % of the main and 75 % of the replication sample) included patients with moderate baseline symptoms that remitted by discharge. Class 2 (39 % and 8 %, respectively) consisted of patients with severe baseline symptoms that demonstrated substantial improvement. Class 3, comprising 15 % of the main and 17 % of the replication sample, included patients with persistently severe symptoms, suggesting non-response to treatment. The predictive model identified Class 3 patients in the main sample with 76 % accuracy and an AUC of 0.87. Neuroticism and somatization emerged as key predictors. Cross-dataset validation, limited to anamnestic predictors available across both samples, showed attenuated performance. Class 3 membership predicted a 1.40-fold increased risk of a symptomatic episode one-year post-discharge.

Our findings highlight reproducible inpatient symptom trajectories shaped by shared risk factors and the potential of longitudinal phenotyping to guide individualized care in depression.