BastionGPT Healthcare AI Assistant Demonstrates 99.3% Accuracy in Peer-Reviewed PICU Study

HIPAA-compliant AI chatbot shows high accuracy and provider satisfaction in pediatric critical care setting

DALLAS, March 26, 2026 /PRNewswire/ -- Bastion Intelligence, the company behind BastionGPT, today announced the publication of a peer-reviewed feasibility study in Critical Care Explorations demonstrating that its HIPAA-compliant AI platform achieved 99.3% sentence-level accuracy when used to support parent communication in a pediatric intensive care unit (PICU).

The study, led by researchers at Baylor College of Medicine and Texas Children's Hospital, evaluated BastionGPT as a medical AI chat tool capable of answering parent questions using patient-specific electronic health record (EHR) data. Of 1,225 chatbot-generated sentences reviewed by PICU physicians, only eight errors were identified (all classified as minor) with no moderate or severe errors detected. Inter-rater reliability among physician reviewers was excellent (Gwet's AC2 = 0.98).

Healthcare providers rated the platform's response quality a median of 5.0 out of 6.0, and physicians expressed strong comfort with routine bedside use of the tool. Parents demonstrated sustained engagement throughout sessions and reported high satisfaction, with a Net Promoter Score of +57. The results have informed the design of a larger randomized controlled trial.

"This study validates what we've been building toward, a medical GPT platform that clinicians can trust at the point of care," said Sanjiv Mehta, MD, co-author of the study. "BastionGPT is purpose-built to meet the security and compliance demands of healthcare. While general-purpose AI tools like ChatGPT, Gemini, and Claude have expanded what's possible with large language models, clinical environments require a HIPAA-compliant AI solution designed specifically for the workflows and safety standards that physicians and care teams depend on" says Joshua Spencer, CEO of Bastion Intelligence.

BastionGPT served as the study's chatbot platform, integrating the latest healthcare AI models from OpenAI, Google and Anthropic within their secure, compliant infrastructure. The platform's architecture enabled researchers to incorporate real patient data into AI-assisted conversations, a critical requirement for clinical AI applications that general-purpose tools cannot safely support.

The study highlights the growing role of AI assistants in clinical settings, not as replacements for physician judgment, but as tools that help care teams communicate more effectively with patients and families. As the researchers note, the healthcare LLM is envisioned as "a supplement to rather than a substitute for bedside communication."

The full study, "Feasibility of a Large Language Model Chatbot to Support Parental Understanding in the PICU," is available in the April 2026 issue of Critical Care Explorations.

About Bastion Intelligence Bastion Intelligence is the maker of BastionGPT, the leading HIPAA-compliant AI platform designed exclusively for healthcare and mental health organizations. BastionGPT enables physicians, clinical staff, and care teams to leverage large language models securely within clinical workflows. Learn more at bastiongpt.com.

Media Contact
Amber Turrill, CMO
Bastion Intelligence
Email: [email protected]
Phone: +1 (214) 619-8696
Website: bastiongpt.com

SOURCE Bastion Intelligence