Contents · 9 sections
- What are the Russian Ministry of Health clinical guidelines and why you need to know them
- Clinical guidelines aggregator: where to find the official text
- How clinical guidelines are structured: sections and where to look for what
- Levels of recommendation conviction (LRC) and levels of evidence credibility (LEC)
- How to quickly find an answer in clinical guidelines
- AI-powered answers based on clinical guidelines: example questions
- How the clinical guidelines neural network differs from ChatGPT
- Limitations: where the clinical guidelines AI can make mistakes
- FAQ
The clinical guidelines neural network answers questions not from "memory", but using the text of specific Ministry of Health guidelines, and indicates the section from which the answer was taken. On WhiteCoat Pro, this is the clinical guidelines section: all 756 active guidelines, search by title and ICD-10 code, full text open without registration, and the "Ask AI" button provides answers based on the document text. Below, we break down how clinical guidelines are structured, what LRC and LEC mean, how to phrase questions, and where the AI can make mistakes.
What are the Russian Ministry of Health clinical guidelines and why you need to know them
Clinical guidelines (CG), pursuant to Federal Law No. 323-FZ, are documents containing evidence-based structured information on prevention, diagnosis, treatment, and rehabilitation: what examinations to prescribe, which drugs and in what situations, and how to manage a patient taking into account complications and comorbidities. They are developed by medical professional non-profit organizations, approved by the Scientific and Practical Council of the Ministry of Health, and reviewed at least once every three years.
Article 37 of the same law establishes that medical care is provided on the basis of clinical guidelines. Therefore, physicians and medical care quality assessments rely on them, and for students and residents, they form the basis for answers on exams, in clinical case studies, and during the accreditation exam. We discussed how to prepare for it using AI in the article neural network for accreditation preparation.
Clinical guidelines aggregator: where to find the official text
The official source is the Clinical Guidelines Aggregator of the Russian Ministry of Health (cr.minzdrav.gov.ru). You can search it by title, ICD-10 code, and age category. A few rules that save time:
- Check the age category. There are often separate guidelines for adults and children for the same disease, such as community-acquired pneumonia.
- Check the approval year and status. Textbooks and old methodological guidelines often contain regimens that have already been revised.
- Search by ICD-10 code if the title cannot be found. Some guidelines cover multiple codes; for example, arterial hypertension encompasses the I10-I15 group, and the wording in the title may not match the usual terms.
The WhiteCoat Pro clinical guidelines section contains the same 756 active guidelines with search by title and ICD-10, and full texts can be opened without logging in. Additionally, there are clinical scales: CURB-65, CHA2DS2-VASc, HAS-BLED, SCORE2, the Glasgow Coma Scale, and other tools found in the guidelines.
How clinical guidelines are structured: sections and where to look for what
All Ministry of Health clinical guidelines have a unified structure, and knowing it helps you find the right answer in a minute.
| Section | What to look for there |
|---|---|
| Title page | ICD-10 codes, age category, approval year, developers |
| 1. Brief Information | Definition, etiology and pathogenesis, epidemiology, coding features, classification, clinical picture |
| 2. Diagnostics | Complaints and anamnesis, physical examination, laboratory and instrumental examinations, diagnostic criteria |
| 3. Treatment | Conservative and surgical treatment, drug groups, pain management, diet therapy |
| 4. Rehabilitation | Medical rehabilitation, including sanatorium-treatment |
| 5. Prevention and Dispensary Observation | Visit frequency, parameters to monitor |
| 6. Organization of medical care | Indications for hospitalization and discharge |
| Quality assessment criteria | What must be performed for the patient: convenient as a checklist |
| Appendices | A3: reference materials, including drug dosages; B: physician action algorithms; C: patient information; D: scales and questionnaires |
Study tip: take the diagnosis wording from the "Classification", the examination plan from Section 2, treatment from Section 3 and Appendix A3, and severity and hospitalization indications from Section 6 and Appendix D. This is the exact way to cross-check a student case history; more details are in the article case history with a neural network.
Levels of recommendation convencity and evidence reliability (LoR and LoE)
Each thesis in the guidelines is marked with two levels, for example: "Recommended... (LRC A, LEC 1)". The LRC shows how strong the recommendation itself is, while the LEC shows what kind of studies it is based on.
| GoR | Meaning |
|---|---|
| A | Strong recommendation: outcomes are important, studies are of high or moderate quality, and conclusions are consistent |
| B | Conditional recommendation: not all outcomes are important, not all studies are of high quality, or conclusions are inconsistent |
| C | Weak recommendation: no proper quality evidence is available |
| LEC (for treatment and prevention) | What it is based on |
|---|---|
| 1 | Systematic review of RCTs with meta-analysis |
| 2 | Individual RCTs and systematic reviews of studies of any design other than RCTs, with meta-analysis |
| 3 | Non-randomized comparative studies, including cohort studies |
| 4 | Non-comparative studies, case reports and case series, case-control studies |
| 5 | Justification of mechanism of action or expert opinion |
For diagnostic interventions, the LEC scale is also five-tiered, but based on studies with a reference standard control. What to do with this in practice: "LRC C, LEC 5" does not mean "do not do it"; it means that data are scarce and the decision depends more on the clinical situation. On an exam, answering "according to clinical guidelines" means exactly what is written in the text, along with the conditions of use.
How to quickly find an answer in clinical guidelines
Manual search through clinical guidelines
Open the required guidelines, go to the section according to the table above, and search the page (Ctrl+F or "Find on page"). Difficulties begin when a document is over a hundred pages long and the terms in the text differ from yours: you search for "ACEI", but the text says "angiotensin-converting enzyme inhibitors".
AI-powered answer based on clinical guidelines in /cr
In the clinical guidelines section, there is an "Ask AI" button. You type your question in plain language, and the neural network answers based on the guidelines texts and shows the source sections from which the answer was compiled. Without logging in, 3 questions per day are available, and 15 after logging in. This is how a clinical guidelines neural network should work: the answer is tied to the document text, not the model's memory. The main rule: open the source section and make sure the answer is conveyed accurately, with all conditions.
AI tutor in your dashboard
If the question is broader than a single guideline—for example, if you need to connect the pathogenesis from a textbook with the management tactics from the guidelines—the AI tutor is suitable. It answers based on the platform's methodological materials (31 subjects, 3,400+ video lessons) and clinical guidelines, adds footnotes [1], [2] to the textbook pages, and shows sources and a video review on the topic below the answer. In "Precise" mode, the tutor searches through the materials; in "Fast" mode, it answers quicker. It understands photos, PDFs, and Word documents. On any account, even a free one, up to 30 messages per hour are available.
AI-powered answers based on clinical guidelines: example questions
A good question includes age (adult or child), setting (outpatient or inpatient), severity, and important comorbidities. One question, one topic.
| Vague | Specific |
|---|---|
| How to treat pneumonia? | What initial antibiotic therapy is recommended for non-severe community-acquired pneumonia in an adult outpatient without comorbidities? |
| Treatment of CHF | Which drug classes are included in the baseline therapy for heart failure with reduced ejection fraction and with what levels of reliability and conviction of evidence (LoR/LoE)? |
| Blood pressure in diabetes | What is the target blood pressure level recommended for a 60-year-old patient with arterial hypertension and type 2 diabetes mellitus? |
| When to hospitalize? | What are the indications for hospitalization in community-acquired pneumonia in adults and what scales are used for this? |
| PE diagnosis | What clinical probability scales for pulmonary embolism are given in the clinical guidelines and in what order should studies be prescribed? |
Could you please provide the LoR and LoE levels: they are in the text of the guidelines themselves and can be easily verified against the source section.
How does a clinical-guidelines-based neural network differ from ChatGPT
A general-purpose model answers based on what it remembered during training. It does not know which Russian clinical guidelines are currently in effect, may confuse foreign and Russian approaches, and sometimes "hallucinates", i.e., confidently invents dosages and references. Furthermore, ChatGPT is officially unavailable from Russia. GigaChat and YandexGPT operate in Russia, but these are general models lacking a medical database with sources. Read more in the article about ChatGPT for medical professionals.
The American platform OpenEvidence answers with references to medical literature, but it is in English, focused on US physicians, and does not contain Russian clinical guidelines. Therefore, working with the guidelines of the Russian Ministry of Health requires a medical neural network in Russian that answers based on their text and shows the source. A comparison of approaches is available in the article OpenEvidence and the Russian equivalent.
Limitations: where an AI based on clinical guidelines may make a mistake
- The neural network does not diagnose or prescribe treatment for a specific individual. It is a tool for study and information retrieval.
- It may retrieve the wrong clinical guidelines: adult guidelines instead of pediatric ones, or a document with a similar title. Verify the title and age category of the source.
- Summarization simplifies things. Conditions like "in case of intolerance", "if GFR is below 30 mL/min/1.73 m²", or "in patients older than 75 years" may be lost. Make decisions based on the text of the section.
- Clinical guidelines are not everything. For a specific drug, the package insert is important; for care organization, operational procedures and standards apply.
- Do not paste patients' personal data: full name, date of birth, address, insurance policy number. This is medical confidentiality (Article 13 of Federal Law 323-FZ) and personal data (Federal Law 152-FZ). Ask anonymously: "60-year-old male, heart failure, GFR 40".
Frequently asked questions
Where to find the clinical guidelines of the Russian Ministry of Health online?
Officially in the registry of the Russian Ministry of Health. On WhiteCoat Pro in the /cr section, all 756 active clinical guidelines are collected with search by title and ICD-10, and the full text is open without registration.
Is there a neural network based on the clinical guidelines of the Russian Ministry of Health?
Yes. The "Ask AI" button in /cr answers based on the texts of clinical guidelines and shows the source sections. The AI tutor in the personal account answers based on the guidelines and platform study guides with footnotes.
How many questions about clinical guidelines can be asked for free?
3 questions per day without logging in, 15 after logging in. The text of the clinical guidelines itself is open without restrictions.
What do LoR A and LoE 1 mean?
A strong recommendation based on a systematic review of RCTs with meta-analysis. This is the highest level in the system adopted in Russian clinical guidelines.
How to find clinical guidelines by ICD-10 code?
Enter the code into the search bar of the registry or the /cr section. If there are no results, try the category code without the refinement after the dot: some clinical guidelines cover an entire group of codes.
Is it permissible to treat a patient based on an AI response?
No. The AI response helps quickly find the necessary fragment of the clinical guidelines, but the decision on diagnosis and treatment is made by the physician based on the full text of the guidelines and taking into account the specific patient.