MC-TRCM: Observation-Aware Recursive Fusion for Incomplete Mobile and Wearable Mental-Health Feature Views

2026-10-08

AI Summary

This study addresses the challenges of incomplete feature views and missing values in mental health data collected from mobile and wearable devices. To this end, it proposes the Modality-Conditional Temporal Recursive Context Model (MC-TRCM), which treats each feature source as an independent token and introduces a learnable missing token to encode absent information. The model employs Feature-wise Linear Modulation (FiLM) to fuse dataset and task embeddings as conditioning signals, while leveraging a validation-guided recursive refinement mechanism to iteratively optimize predictions. Evaluated on benchmarks such as DepreST-CAT, MC-TRCM significantly reduces regression errors for PHQ-9 and GAD-7 scores and achieves strong classification performance, demonstrating efficient fusion of multi-source heterogeneous data. The implementation has been made publicly available.

Abstract

Public mobile and wearable mental-health datasets often provide summarized feature tables rather than synchronized raw sensor streams. In these releases, each anchor corresponds to a survey or label time and may combine phone or wearable summaries, prior symptom scores, demographics, clinical variables, and source-availability indicators. We propose the Modality-Conditioned Temporal Recursive Context Model (MC-TRCM), which preserves each feature source as a separate token and incorporates missingness as part of the input context. Observed sources are encoded with values and missingness summaries, absent sources use learned absence tokens, dataset and task embeddings condition fusion, and a recursive prediction head refines each output over validation-selected steps. We evaluated MC-TRCM on six predefined endpoints from DepreST-CAT and Prediction of Severity Change-Depression (PSYCHE-D) using participant-level splits and validation-only model selection. MC-TRCM achieved the lowest mean absolute error on DepreST-CAT Patient Health Questionnaire-9 (PHQ-9) and Generalized Anxiety Disorder-7 (GAD-7) severity, improving over the best tabular reference by 0.181 and 0.217 scale points. Classification endpoints showed task-dependent behavior: MC-TRCM matched the best rounded GAD-7 category balanced accuracy, was numerically highest by 0.002 balanced-accuracy points on PSYCHE-D multiclass prediction, and remained close to the strongest references on PHQ-9 category and PSYCHE-D binary prediction. Ablations support Feature-wise Linear Modulation, absence tokens, missingness projections, and recursive refinement, while calibration and feature-source controls characterize endpoint behavior. Our code is available at https://github.com/Botwwt/MC-TRCM.

Problem

• mental health prediction
• incomplete feature views
• multimodal fusion
• wearable data
• missingness handling

Innovation

• Modality-Conditioned Fusion
• Absence Tokens
• Recursive Refinement
• Missingness-Aware Modeling
• Feature-wise Linear Modulation