evidoria

← Back to browse

Good practice Imported

HomeDOCtor — Slovenia's Nationally Localised AI Chatbot for Primary-Care Guidance

Slovenia · Ljubljana · See the Slovenia profile · See the Ljubljana profile

Evidence: Quasi-experimental Top 67% 53/100 · Ask Evidence Copilot about this practice

HomeDOCtor gives Slovenian citizens 24/7 AI primary-care guidance grounded in national protocols. In peer-reviewed testing it scored 90.7% on national medical-exam questions vs 80.7% for generic ChatGPT-4o. Authors call it a research prototype, not a permanent service.

up to 99% Top-1
Diagnostic accuracy, international clinical vignettes
90.7% vs 80.7% p=0.0135
National licensing-exam score, HomeDOCtor vs baseline
<3 seconds
Average response time
HomeDOCtor — Slovenia's Nationally Localised AI Chatbot for Primary-Care Guidance

Details

Maturity
Pilot
Promoter
Jožef Stefan Institute, Department of Intelligent Systems, funded by the EU Horizon Europe ChatMED project
Period
Nationwide public testing end of 2024 through February 2025; peer-reviewed evaluation published 2025; remains a research prototype
Keywords
digital health, generative AI, primary care, public research

Context

HomeDOCtor is a conversational AI system built by the Jozef Stefan Institute's Department of Intelligent Systems, funded through the EU Horizon Europe 'ChatMED' project, combining retrieval-augmented generation with a curated database of Slovenian medical guidelines to give round-the-clock primary-care guidance.

Results

Opened for nationwide public testing end of 2024 through February 2025 and evaluated in a peer-reviewed 2025 study. On 100 international clinical vignettes, HomeDOCtor variants reached up to 99% Top-1 diagnostic accuracy. On 150 questions from Slovenia's national medical licensing exam, HomeDOCtor (GPT-4o-based) scored 90.7% versus 80.7% for a generic ChatGPT-4o baseline (p=0.0135), with average response time under three seconds.

Conclusions

The authors are explicit that HomeDOCtor remains a research prototype: not integrated with electronic health records, scope limited to internal medicine, usability testing informal and limited to expert reviewers, and by design it retains no session data, which protects privacy but precludes personalisation or longitudinal outcome tracking.

Implementation
Indicative cost
Medium (€50k–€500k)
Time to results
Short (< 1 year)
Staffing & skills
Jozef Stefan Institute Department of Intelligent Systems researchers

Conditions for success

  • Retrieval-augmented generation grounded in national clinical protocols
  • Horizon Europe research funding (ChatMED project)

Common failure modes

  • Research prototype only, not integrated with electronic health records
  • No session retention precludes personalisation or longitudinal tracking

Commonly funded by

Horizon Europe

Indicative funding routes for practices of this type — always check each programme's current calls and eligibility rules.

Do you run this practice? Claim it — verified implementers get a public contact pathway and can propose corrections.

Data sources

Where this practice's information was retrieved from, and when.

Attachments

Similar practices you may find useful