Prediction and Intervention Effect of Rehabilitation Status for Severe Mental Disorder Patients Based on Multimodal Analysis and AI Agents

Mental health issues represent a major public health and social problem that significantly impacts economic and social development. Compared to other diseases, mental disorders can impair various aspects of a patient' s life, including psychological, social, occupational, and educational functions, affecting their quality of life and daily living abilities. Particularly, severe mental disorders tend to have a chronic course, often resulting in diminished social functions and social withdrawal, making it difficult for patients to integrate into society. Repeated, systematic, and comprehensive rehabilitation training for patients with severe mental disorders can effectively control or delay disease recurrence, improve social functions, enhance quality of life, and facilitate patients' reintegration into society.

In recent years, the scope of mental disorder rehabilitation has expanded to include enhancing patients' social functions and promoting their integration into society. Vocational rehabilitation and social skills training are widely used in the rehabilitation treatment of patients with severe mental disorders, and some physical intervention methods, such as neurofeedback training, have also proven to be significantly effective in the rehabilitation process. However, traditional rehabilitation techniques often lack specificity and fail to meet individualized needs of patients. Additionally, the rehabilitation process lacks long-term monitoring, making it challenging to continuously assess and adjust patients' rehabilitation outcomes. Furthermore, the assessment of rehabilitation effectiveness mainly relies on patients' subjective feelings and clinical observations, lacking high-quality evidence. Therefore, there is an urgent need to introduce new rehabilitation technologies and scientifically evaluate their effectiveness to address the shortcomings of traditional methods and provide more personalized, precise, and effective rehabilitation support.

With the rise of digital health technologies, the field of mental health rehabilitation has encountered new opportunities. Compared to traditional therapies, digital health is revolutionizing the healthcare industry, moving away from traditional approaches to healthcare management to real-time personalized monitoring and therapeutic care.Technologies such as remote monitoring, virtual reality, and computer-assisted cognitive correction therapy are increasingly applied in rehabilitation. However, these methods still need improvements in data management and integration capabilities. A large amount of data accumulates in systems, recording only the training process and real-time effects of patients, without further evaluating their rehabilitation status, leading to resource waste. Therefore, there is an urgent need to develop a digital rehabilitation model that better meets the genuine needs of patients with severe mental disorders.

This study aims to integrate multimodal technology, reinforcement learning, and agent-based modeling (ABM) into the research of mental health rehabilitation to more accurately assess and predict the rehabilitation status of mental disorder patients and to more effectively guide and support decision-making in mental rehabilitation treatment.

Study Overview

• Status: Recruiting
• Conditions: Schizophrenia; Schizoaffective Disorder; Severe Mental Disorder; Bipolar Affective Disorder; Paranoid Psychosis; Mental Disorders Due to Epilepsy; Mental Retardation Accompanied by Mental Disorders
• Intervention / Treatment: Other: Gamified Digital Rehabilitation; Behavioral: Routine Care

Detailed Description

This study aims to integrate multimodal technology, reinforcement learning(RL), and agent-based modeling (ABM) into the research of mental health rehabilitation to more accurately assess and predict the rehabilitation status of mental disorder patients and to more effectively guide and support decision-making in mental rehabilitation treatment.

This research project is divided into three main phases: theoretical and experimental phase, multimodal analysis phase, and application and optimization phase.

Firstly, we will conduct in-depth research across 20 community mental health facilities in Shanghai. This will be combined with an analysis of existing literature and studies to understand the needs of potential users, providing theoretical support and design basis for subsequent gamification interventions. This phase of user feedback, literature review, and needs assessment will offer directional guidance for the entire research.

In the second phase, based on the user needs and literature analysis results from the previous phase, we will design and implement gamification interventions, followed by a randomized controlled trial. Simultaneously, we will collect and analyze game behavior data to systematically evaluate the actual effects of the gamification interventions. This phase, focusing on intervention design and effect evaluation, is the core part of the research, and user feedback will continuously guide us in optimizing the interventions.

In the third phase, using the Multimodal and Crossmodal AI framework(MMCRAI), we will analyze multimodal data including patients' game behavior, physiological indicators, and psychological health information to better understand the key factors and dynamic changes in the rehabilitation process. This will provide training signals for the subsequent modeling and optimization phase.

Finally, combining agent-based models and reinforcement learning algorithms, we will simulate and predict the effects of gamification interventions in actual community settings, thus translating theoretical and experimental results into practical guidelines. In this phase, we will validate the effectiveness of the agent-based model through real-world application scenarios and discuss potential limitations and assumptions.

Overall, these four phases are interrelated and logically coherent, progressively deepening our understanding of the mental health rehabilitation process and forming a complete research framework, thereby laying a solid foundation for practical applications. Throughout the study, we will uphold ethical principles and privacy policies, ensuring that all research activities comply with regulations.

• Study Type: Interventional
• Enrollment (Estimated): 82
• Phase: Not Applicable

Contacts and Locations

• Study Contact: Name: Weibo Zhang; Phone Number: 86-13764694223; Email: zhangweibo600@163.com

Study Locations

• China
  • Shanghai, China
    • Recruiting
    • Shanghai Mental Health Center
    • Contact: Jun Cai; Phone Number: 86-18017311058; Email: caijun533@163.com

Participation Criteria

Eligibility Criteria

• Ages Eligible for Study: Adult; Older Adult
• Accepts Healthy Volunteers: No

Description

Inclusion criteria:

• Registered in the Shanghai Mental Health Information Management System,

  • Diagnosed patients with one of the six severe mental disorders: schizophrenia, schizoaffective disorder, paranoid psychosis, bipolar (affective) disorder, mental disorder due to epilepsy, and mental retardation accompanied by mental disorder,

    • Aged between 18 and 65 years old, ④ Normal vision or hearing, or within the normal range after correction, ⑤ Patients or their families have provided informed consent for this study and signed the informed consent form.

Exclusion criteria:

Patients with severe physical illnesses or organic brain diseases.

Study Plan

How is the study designed?

Design Details

• Primary Purpose: Other
• Allocation: Randomized
• Interventional Model: Parallel Assignment
• Masking: None (Open Label)

Number of Arms

2

Arms and Interventions

Participant Group / Arm	Intervention / Treatment
Experimental: Intervention Group (Gamified Digital Rehabilitation); Participants in this arm will receive routine pharmacological treatment and standard community rehabilitation services, combined with a structured, story-based gamified digital rehabilitation intervention. The interventions are digital functional games five times a week (30 minutes each) for 3 months.	Other: Gamified Digital Rehabilitation; The gamification intervention measures are as follows: Currently, the initial version of the community mental health rehabilitation interactive game mainly focuses on six dimensions of medication management, including the importance of taking medication, identifying and dealing with adverse reactions to antipsychotic drugs, learning self-management of medication, assessing the effectiveness of medication treatment, long-term management of medication treatment, and discussing issues related to medication effects with medical staff. Next, we will continue to design a series of games themed on symptom management and psychological rehabilitation, and implement game-based digital rehabilitation interventions for patients in the intervention group based on these games.; Behavioral: Routine Care; Receive regular psychiatric medication treatment and regular community rehabilitation services, including regular follow-ups, rehabilitation guidance, and community education,etc.
Active Comparator: Control Group (Routine Care); Participants in this arm will receive routine pharmacological treatment and standard community rehabilitation services during the same period.	Behavioral: Routine Care; Receive regular psychiatric medication treatment and regular community rehabilitation services, including regular follow-ups, rehabilitation guidance, and community education,etc.

What is the study measuring?

Primary Outcome Measures

Outcome Measure	Measure Description	Time Frame
BPRS reduction rate	The Brief Psychiatric Rating Scale (BPRS) was used to measure the presence and severity of psychiatric symptoms entailing positive symptoms, general psychopathology, and affective symptoms (e.g., thought disturbance, emotional withdrawal, hostility, and suspiciousness) for patients with mental illness, particularly schizophrenia. Each of the 18 items are designed to represent a discrete symptom area. Items are rated on a 7-point Likert scale, from 1 = 'not present' to 7 = 'extremely severe', with scores ranging from 18 to 126 (achieved through summing the item scores). Higher scores indicated more severity of psychiatric symptoms. Reduction rate was calculated using the following formula: reduction rate= (Score before treatment-Score after treatment)/(Score before treatment-18) × 100%. A reduction rate of BPRS score > 25% was considered as a clinically meaningful improvement.	Baseline (pre-intervention), immediately post-intervention (3 months), 1-month follow-up (4 months), 3-month follow-up (6 months) and 6-month follow-up (9 months).

Secondary Outcome Measures

Outcome Measure	Measure Description	Time Frame
MMAS-8	The 8-item Morisky medicationadherence scale is the most widely used medication compliance assessment scale, which has been applied to patients with multiple chronic diseases and has been translated into different languages.	Baseline (pre-intervention), immediately post-intervention (3 months), 1-month follow-up (4 months), 3-month follow-up (6 months) and 6-month follow-up (9 months).
GAD-7	The Generalized Anxiety Disorder Screener (GAD-7) is a brief, 7-item self-report measure for anxiety symptoms in adolescents and adults. Although developed primarily as a screening tool for GAD, it can also be used as a diagnostic tool to monitor changes in symptoms over time, and as a screener for social anxiety, panic disorder, and post-traumatic stress disorder. Each item is rated on a likert-type scale from zero ('not at all') to three ('nearly every day'), with total scores ranging from 0 to 21.	Baseline (pre-intervention), immediately post-intervention (3 months), 1-month follow-up (4 months), 3-month follow-up (6 months) and 6-month follow-up (9 months).
PHQ-9	The Patient Health Questionnaire-9 (PHQ-9) is a multipurpose instrument for screening , diagnosing, monitoring and measuring the severity of depression. The tool rates the frequency of the symptoms which factors into the scoring severity index. The PHQ -9 is brief and useful in clinical practice. The PHQ -9 can be administered repeatedly , which can reflect improvement or worsening of depression in response to treatment .Each item is rated on a likert-type scale from zero ('not at all') to three ('nearly every day').	Baseline (pre-intervention), immediately post-intervention (3 months), 1-month follow-up (4 months), 3-month follow-up (6 months) and 6-month follow-up (9 months).
WHOQOL-BREF	The WHO Quality of Life Scale-Brief (WHOQOL-BREF) was developed by the Word Health Organisation as a quality of life assessment that would be applicable cross-culturally. It is a 26 item self report measure, especially useful for measuring outcomes with adults with a psychosocial disability. The WHOQOL-BREF measures four factors: physical health, psychological Health, social relationships, and environment. The total score is presented between 26 and 156, where higher scores represent higher levels of quality of life.	Baseline (pre-intervention), immediately post-intervention (3 months), 1-month follow-up (4 months), 3-month follow-up (6 months) and 6-month follow-up (9 months).
SDSS	The 10-item social disability screening schedule (SDSS) was used for measuring social function. It is a shortened version of psychiatric disability assessment schedule (DAS). SDSS was used to evaluate social disability, with 10 items in total, 0~2 points for each item.	Baseline (pre-intervention), immediately post-intervention (3 months), 1-month follow-up (4 months), 3-month follow-up (6 months) and 6-month follow-up (9 months).

Collaborators and Investigators

• Sponsor: Shanghai Mental Health Center
• Collaborators: Shanghai Jiao Tong University School of Medicine
• Investigators: Principal Investigator: Jun Cai, Shanghai Mental Health Center

Study record dates

Study Major Dates

• Study Start (Actual): March 12, 2024
• Primary Completion (Estimated): June 30, 2026
• Study Completion (Estimated): June 30, 2026

Study Registration Dates

• First Submitted: March 25, 2025
• First Submitted That Met QC Criteria: March 25, 2025
• First Posted (Actual): April 1, 2025

Study Record Updates

• Last Update Posted (Actual): August 17, 2025
• Last Update Submitted That Met QC Criteria: August 12, 2025
• Last Verified: August 1, 2025

More Information

Terms related to this study

• Keywords: Severe Mental Disorder; Reinforcement Learning; Agent-based Modeling; Multimodal and Crossmodal AI framework; Digital rehabilitation
• Additional Relevant MeSH Terms: Bipolar and Related Disorders; Neurologic Manifestations; Nervous System Diseases; Schizophrenia Spectrum and Other Psychotic Disorders; Neurobehavioral Manifestations; Neurodevelopmental Disorders; Schizophrenia; Psychotic Disorders; Bipolar Disorder; Mood Disorders; Mental Disorders; Intellectual Disability; Paranoid Disorders
• Other Study ID Numbers: Jun Cai

Plan for Individual participant data (IPD)

• Plan to Share Individual Participant Data (IPD)?: NO

Drug and device information, study documents

• Studies a U.S. FDA-regulated drug product: No
• Studies a U.S. FDA-regulated device product: No
• product manufactured in and exported from the U.S.: No