Canonical text (the exact bytes fingerprinted)

TRUVACE RECORD VERSION
record: TRV-2026-1810
version: 1
kind: certified
reason: Certified into the record
timestamp: 2026-10-18T10:42:03.946840Z
status: published
lens: g_space
sector: health
headline: Development and Preliminary Evaluation of a Conversational Agent Delivering Problem-Solving Therapy for Family Caregivers of Children With a Chronic Health Condition: Multiphase Mixed Methods Study
dek: Background Family caregivers of children with chronic health conditions experience substantial physical and mental health burdens, including burnout, anxiety, depression, fatigue, and sleep disturbances. Despite this need, validated digital mental health tools tailored to family caregivers remain limited. AI-powered conversational agents offer a promising approach for delivering on-demand, personalized mental health support, yet development and evaluation frameworks for this population are lacking. Objective Thi…
gain_title: In a single-session test with 38 caregivers, a Wizard-of-Oz version of COCO was rated acceptably usable and was followed by lower negative affect, sadness, guilt, and fatigue.
problem_title: (none)
trace_subject: (none)
gain_reading: In a single-session test with 38 caregivers, a Wizard-of-Oz version of COCO was rated acceptably usable and was followed by lower negative affect, sadness, guilt, and fatigue.
gain_evidence: Participants demonstrated significant improvement in negative affect, sadness, guilt, and fatigue following PST sessions
problem_reading: (none)
problem_evidence: (none)
quick_read: Researchers built COCO, a conversational agent that delivers problem-solving therapy and motivational interviewing to family caregivers of children with chronic health conditions. Development moved through four phases, from scripted dialogues and prototype testing with standardized patients and caregivers to a GPT-4-powered version.

The work matters because caregivers carry heavy physical and mental burdens, and validated digital tools for them are scarce. The evidence is early, though: the usability and mood results come from a single-session test with a human-operated Wizard-of-Oz version, and the GPT-4 version was only evaluated by research team members using personas.
limitation: The user-facing results come from a single therapy session with a Wizard-of-Oz prototype, and the LLM version was evaluated only by research team members using personas, so real-world performance of the final AI agent is untested.
tag: Evidence-backed gain
key_points: Caregivers of children with chronic conditions face burnout, anxiety, depression, fatigue, and sleep disturbances, and validated digital tools for them are limited. | COCO was developed in four phases, beginning with four gold-standard dialogues built from caregiver personas. | Standardized patients in phase 2 called COCO validating and found its problem-solving and on-demand support helpful. | Phase 3 with 38 caregivers produced a mean SUS score of 75.6% (SD 12.9%), which the authors describe as acceptable usability. | In phase 4, GPT-4 was integrated using few-shot learning and checked by research team members against caregiver personas.
rundown: The design combined problem-solving therapy with motivational interviewing inside a human-in-the-loop framework. The Wizard-of-Oz method collected naturalistic dialogues in phases 2 and 3 to refine the conversational design before an LLM was introduced.

Phase 3 measured caregiver emotions before and after a session using six PANAS-X subscales, and the authors report significant improvement in negative affect, sadness, guilt, and fatigue. Motivational interviewing techniques used by COCO across phases were coded with the Motivational Interviewing Treatment Integrity Coding Manual.

The authors conclude COCO showed feasibility and usability for delivering protocolized therapeutic support, and offer a structured framework for testing evidence-based therapy approaches iteratively before LLM deployment.
sources:
- peer_reviewed | Journal of Medical Internet Research | https://doi.org/10.2196/94175 | 2026-10-06
prev: 0000000000000000000000000000000000000000000000000000000000000000

sha256

838e2e671128f5ce8c728804c600acd034956f00a199f0c66bad1697c43ae830

previous

0000000000000000000000000000000000000000000000000000000000000000