Vindicate

VINDICATE: Differential Diagnosis Trainer A mobile-friendly medical education web application designed to help learners rapidly practice diagnostic reasoning using the VINDICATE framework.

Comment

VINDICATE

🥉 3rd Place at Hack4Health Hackathon hosted by the Artificial Intelligence in Medicine Society (AIMS) and supported by the Alberta Medical Association (AMA) and the Centre for Artificial Intelligence Ethics, Literacy and Integrity (CAIELI).

UCalgary Article on Hack4Health: Hack4Health Hackathon Challenges Students to Design Solutions for Health Systems

Inspiration

Diagnostic errors are one of the leading causes of preventable patient harm. The single most common cognitive error is premature closure — when clinicians stop considering reasonable alternatives after an initial diagnosis is made.

A patient with chest pain could have a heart attack, pulmonary embolism, pneumonia, or something benign — but missing the wrong one can be fatal. Patient safety depends on thinking broadly.

Medical students are the future of medicine. When they are trained properly, you are safer as a patient.

What it does

VINDICATE is an AI-powered clinical reasoning tool that helps medical students build complete differential diagnoses using the VINDICATE framework.

VINDICATE stands for:

Vascular Infectious Neoplastic Degenerative Iatrogenic Congenital Autoimmune Traumatic Endocrine

Students enter a patient presentation (for example: chest pain, fever, shortness of breath), and VINDICATE generates a structured, safe list of possible diagnoses across all nine categories, ensuring nothing dangerous is missed.

How it works

Current medical studying relies on:

  • Manually written cue cards
  • Passive reading
  • Incomplete or non–fact-checked resources
  • Memorization instead of reasoning

VINDICATE replaces this with:

  • Validated differentials
  • Active recall
  • Spaced repetition
  • Fact-checked medical knowledge
  • Structured VINDICATE framework

This trains students to think the way real doctors must think — systematically and safely.

Technical Stack & AI Integration

VINDICATE was built using TypeScript across the frontend and backend, enabling strong type safety, scalable architecture, and rapid iteration during the hackathon. The platform included an optional AI assistant powered by the Gemini API, allowing clinicians to ask free-text clinical questions and receive timely, context-aware responses during case review. All symptoms, diagnoses, and clinical pathways were validated against DynaMed differential diagnosis references and a curated MCQ-based list of patient presentations, ensuring that every output was medically accurate, evidence-based, and aligned with real exam and clinical standards.

What we achieved

At the Hack4Health Hackathon, over 100 students formed interdisciplinary teams to build digital health tools.

Our project, VINDICATE, placed:

🥉 3rd Place Overall

We were recognized for:

  • Strong patient-safety impact
  • Clinical relevance
  • Educational innovation
  • Clear real-world application

Team

Muhammad Saim, Ramya Sridhar, Palak Patel, Sumaiya Rizwan, Evezi Esiehor, Tinashe Chigwida

Why this matters

Studies show that diagnostic error affects 10–15% of patients, and premature closure is the #1 cause of those errors.

VINDICATE directly targets this failure point by forcing future doctors to consider all major disease categories before committing to a diagnosis, improving safety, accuracy, and patient outcomes.

Disclaimer

The VINDICATE framework is an educational tool designed solely for the purpose of helping medical learners practice and strengthen their recall of differential diagnoses. It is not intended to replace professional medical judgment, clinical experience, or institutional guidelines. While every effort has been made to include a wide range of possible differentials, the lists provided may not be exhaustive or applicable to every clinical scenario.

Users must always apply independent clinical reasoning, verify information against current evidence-based sources, and prioritize patient safety when making diagnostic or management decisions. The creators of this material assume no liability for decisions made based on its content.

References

Cook CE, Décary S. Higher order thinking about differential diagnosis. Braz J Phys Ther. 2020;24(1):1-7. doi:10.1016/j.bjpt.2019.01.010

Kämmer JE, Schauber SK, Hautz SC, Stroben F, Hautz WE. Differential diagnosis checklists reduce diagnostic error differentially: A randomised experiment. Med Educ. 2021;55(10):1172-1182. doi:10.1111/medu.14596

What We Learned

One of the biggest takeaways from Hack4Health was that VINDICATE should be evaluated as a clinical education tool, not just a tech demo. Our next step is to run a pilot research study with medical students to measure whether using VINDICATE actually improves diagnostic breadth, retention, and clinical reasoning. If successful, this would support integrating VINDICATE into the Canadian medical curriculum as a formal learning and assessment tool.

We also learned that the most sustainable funding model is institutional licensing — selling VINDICATE directly to medical schools, residency programs, and teaching hospitals. These institutions already pay for clinical decision tools, exam prep platforms, and digital curricula, making this a natural fit for adoption and scale.

Why Not Just Use ChatGPT?

Large language models like ChatGPT are powerful, but they are not designed for safe clinical reasoning:

  • They hallucinate and may fabricate facts
  • Their sources are opaque and cannot be controlled
  • They decide what is important, which can cause clinically critical diagnoses to be omitted
  • They do not follow a systematic diagnostic framework

VINDICATE solves this by:

  • Using a fixed medical framework (VINDICATE mnemonic)
  • Constraining AI output to validated differential categories
  • Having content reviewed by medical students → residents → physicians
  • Ensuring outputs remain structured, complete, and safety-focused

This turns AI from a risky black box into a controlled clinical reasoning engine.

How We Compare to Existing Tools

We respect and build upon existing clinical tools — but VINDICATE fills a unique gap.

Diagnosaurus DDx

Diagnosaurus is an excellent quick-lookup differential list, but it is:

  • Static
  • Not personalized by age or population
  • Not designed for learning or cognitive training

VINDICATE is interactive, structured, and educational, helping users practice how to think, not just what to look up.

Anki

Anki is powerful for memorization, but:

  • Requires users to manually create or find cards
  • Does not enforce clinical frameworks
  • Does not adapt to real patient presentations

VINDICATE automatically generates case-based, framework-driven differentials, turning real clinical reasoning into an active learning system.

UpToDate

UpToDate is the gold standard for expert-authored clinical reference, but it is:

  • Designed for reading, not thinking
  • Not built to train diagnostic breadth
  • Not optimized for students or exams

VINDICATE complements UpToDate by acting as the diagnostic thinking layer

What Makes VINDICATE Different

  • Evidence-based differential sources
  • Exam- and clinic-aligned patient cases
  • Human-reviewed medical content
  • A safety-first design philosophy

Built With

Share this project:

Updates