🧭 ChatVTE emotional support rated 1.92/5 by patients
🧭 ChatVTE emotional support rated 1.92/5 by patients
In a prospective cohort of 25 adult inpatients with Venous Thromboembolism (VTE), an immersive retrieval-augmented LLM education tool (ChatVTE; Qwen1.5-7B) scored >4/5 in 8 of 9 patient experience domains—but emotional support lagged at 1.92/5. In the same study, 4 VTE experts rated ChatVTE higher than ChatGPT on accuracy, completeness, consistency, and safety (safety P=.01; r=0.327).
Why It Matters To Your Practice
Patient-facing AI can deliver high-quality, actionable VTE education at scale, potentially reducing clinician time spent on repetitive counseling.
However, the lowest-rated domain was emotional support (1.92/5), signaling a likely gap in bedside acceptability even when informational quality is strong.
For clinicians, this reinforces that “good answers” may not translate into “good patient experience,” especially during acute hospitalization.
Clinical Implications
Position patient-education chatbots as adjuncts for teaching (e.g., anticoagulation basics, warning signs, mobility guidance), not substitutes for empathic conversations.
If deploying similar tools, add guardrails and workflows that route distress, fear, or uncertainty to humans (nursing, pharmacy, care management) rather than attempting automated reassurance.
Consider pairing AI education with structured clinician follow-up prompts (e.g., “What worries you most about your clot or blood thinner?”) to close the emotional-support gap.
Insights
The system combined retrieval augmentation with speech and lip-synch—suggesting that “immersive” UX alone doesn’t ensure perceived emotional support.
Expert ratings favored ChatVTE over ChatGPT across multiple quality dimensions, but patient ratings revealed a different failure mode: relational needs.
The evaluation spanned standardized VTE questions (Dec 10–31, 2024) plus real inpatient use with discharge surveys (Mar 1–May 31, 2025), highlighting the importance of testing AI in real clinical contexts—not just benchmarks.
The Bottom Line
AI can raise the floor on patient education quality for VTE, but this study’s standout signal is that emotional support may remain a weak point (1.92/5) even when other experience scores are high.
Adoption strategies should treat empathy as a clinical workflow problem (handoffs, escalation, scripting), not a feature checkbox.