Dr. T Health Alert Mini

Inspiration Health concerns can become urgent before people know what to do next. Symptoms, personal health information, and uncertainty can make it difficult to communicate concerns clearly or recognize when professional help is needed. Dr. T Health Alert Mini grew from a simple goal: make AI-powered health guidance more accessible, actionable, and safety-conscious. We wanted to create an experience that helps people organize their concerns, identify potential warning signs, and prepare clearer information for healthcare professionals—without pretending that AI can replace a clinician. Our broader vision is to make useful health technology more approachable, inclusive, and practical for people across different communities and languages. What it does Dr. T Health Alert Mini is a health-alert and guided-intake experience within the Dr. T health AI platform. Its core capabilities include:

• Guided clinical intake: Helps users organize relevant information through a structured workflow.
• Safety-first prioritization: Applies explicit red-flag rules to help identify situations that may require urgent professional attention.
• Clinician handoff: Structures collected information into an SBAR-style summary—Situation, Background, Assessment, and Recommendation—to support communication with healthcare professionals.
• AI-assisted interaction: Uses conversational AI to make the experience more accessible and help users navigate the workflow.
• Structured health output: Supports generating a FHIR R4 RiskAssessment resource as a structured representation for further review and integration.
• Emergency awareness: Provides configured emergency contact guidance for supported regions. The experience is designed to support awareness and communication, not to diagnose conditions, prescribe treatment, or replace emergency services or professional medical judgment. How we built it We developed Dr. T Health Alert Mini as part of the existing Dr. T / Zenieverse platform rather than creating an isolated prototype. This approach allows the feature to build on shared application infrastructure while keeping its health-alert workflow distinct. The implementation combines a guided intake interface, application services, AI-assisted responses, and explicit safety logic. The AI integration uses the Google GenAI TypeScript SDK, with a configured model request timeout and a deterministic fallback path intended to keep the workflow useful when an AI request fails or takes too long. We also focused on responsible handling of health-related information, including authentication and data-ownership controls where supported by the application. Structured output and clinician-facing summaries help make information easier to review rather than presenting conversational AI as an unquestionable authority. Our design principle was to combine AI flexibility with predictable safety behavior: AI can help users navigate and organize information, while explicit rules handle defined warning signs and the interface communicates the limits of the system. Challenges we ran into One of the biggest challenges was balancing conversational AI with the predictability required in a health-related experience. A natural-language model can be helpful, but it must not be the sole decision-maker for urgent warning signs. We addressed this design challenge by separating defined red-flag checks from general AI interaction and by providing a fallback path for AI failures. Another challenge was translating a user's potentially unstructured concerns into information that can be reviewed and communicated efficiently by a healthcare professional. Integrating the feature into an existing multi-module platform also required care. We wanted to add a focused health-alert experience without disrupting established functionality, while keeping navigation, service behavior, and data access consistent. Finally, we recognize that software tests alone cannot establish clinical safety. Clinical validation, regional emergency-information verification, privacy review, and testing with representative users remain important steps toward broader deployment. Accomplishments that we're proud of We are proud to have developed a dedicated health-alert experience that brings guided intake, safety-oriented prioritization, clinician handoff, and structured health information into one workflow. We are especially proud of three design decisions:
• Safety is part of the workflow: Defined warning signs and escalation guidance are treated as important requirements rather than optional additions to a chatbot.
• AI output is meant to be reviewed: Structured summaries support communication and human oversight instead of encouraging users to treat AI responses as a final diagnosis.
• The feature builds on a broader platform: Dr. T Health Alert Mini contributes to a reusable health AI environment designed to support further development and improvement. We see this as a meaningful foundation for making health information easier to organize and communicate, while remaining transparent about the work still needed before clinical use at scale. What we learned We learned that building health AI responsibly involves much more than connecting a language model to a chat interface. Clear workflows, understandable limitations, predictable handling of urgent symptoms, privacy protections, and useful handoff information all matter. We also learned to design for failure: model requests can time out, information can be incomplete, and automated output always needs appropriate boundaries. Most importantly, we learned that AI should support people and healthcare professionals—not replace their judgment. A useful health assistant is not simply one that generates fluent answers; it is one that helps users take more informed next steps while communicating uncertainty honestly. What's next for Dr. T Health Alert Mini Our next priority is to strengthen the experience through systematic testing and real-user feedback. Planned improvements include:
• Better first-time guidance: Make it easier for users to understand what information to provide and what the system can and cannot do.
• Expanded safety testing: Evaluate red-flag scenarios, ambiguous symptoms, incomplete information, AI timeouts, and fallback behavior.
• Clinical review: Work with qualified healthcare professionals to review intake questions, escalation language, and handoff summaries before making stronger clinical claims.
• Privacy and interoperability: Further review data access, retention, consent, and the practical interoperability of structured health outputs.
• Accessibility and localization: Improve multilingual interaction and verify emergency guidance for each supported region.
• User-centered evaluation: Measure usability, clarity, reliability, and the quality of information prepared for professional review. Our long-term vision is for Dr. T Health Alert Mini to become a trustworthy bridge between a person's health concerns and the professional support they may need. We will prioritize safety, transparency, and evidence-based evaluation as the project evolves.

Built With

• and
• conversational-ai
• digital
• firestore
• generative-ai
• google-cloud-run
• google-gemini
• healthcare
• hl7-fhir
• react
• typescript