MedXY
Evidence-based Medical AI

How to Ask MedXY the Right Way?

MedXY is an evidence-based medical AI designed for clinicians and medical researchers. Understanding how to ask the right question helps you get the highest-quality evidence-based answer every time.

RAG: Retrieval-Augmented Generation High-quality medical literature only No open-web browsing for hallucination control Evidence-based, not generative guesswork

1. Reasonable Questions vs Unreasonable Questions

MedXY needs a complete question sentence to work effectively. A single medical term by itself—such as a disease name, drug name, or lab marker—is not a real question. The system cannot infer what you actually want to know.

No. Example Input Rating Why
01 Aspirin ✗ Poor This is just a drug name, not a question. AI cannot know whether you are asking about dosage, indication, side effects, interactions, or something else.
02 Can aspirin be used together with PD-1 therapy? ✓ Good This clearly identifies two drugs and the core issue—potential interaction—so the system can retrieve relevant evidence on co-administration and safety.
03 A patient is taking aspirin and is scheduled for surgery next week. When should the drug be stopped? ✓ Good This provides a clinical scenario, current medication context, and a specific decision target: preoperative discontinuation timing.
04 Diabetes ✗ Poor A disease name alone contains no question. Diagnosis? Treatment? Complications? Monitoring? The system cannot guess your real intent.
05 In a patient with type 2 diabetes and heart failure, should an SGLT-2 inhibitor be prioritized over metformin? ✓ Good This specifies the patient profile, the compared therapies, and the clinical decision being made. It is a standard clinical question format.
06 Troponin ✗ Poor This is only a lab marker name. Normal range? Clinical significance of elevation? Dynamic monitoring? You need to state your actual question.
07 A patient with acute chest pain had a normal first troponin, but the repeat level rose after 3 hours. Should the cath lab be activated immediately? ✓ Good This includes the full clinical setting, dynamic lab changes, and a concrete decision point. The system can accurately match NSTEMI-related guideline content.
Core principle: a noun is not a question — a sentence is. Each input should include: who (patient characteristics), what is happening (background / medication / tests), and what exactly you want to know.

2. Clear Questions vs Vague Questions

Even a complete sentence can still be too vague if it lacks key information. A clear medical question should include age, comorbidities, specific values, or other relevant background so the system can target the most relevant evidence.

✗ Vague Questions (answers will stay generic)

How should aspirin be taken?

Missing: age, indication (primary prevention? secondary prevention? anti-inflammatory use?), comorbidities, current medications.

How should I choose an antibiotic?

Missing: infection site, likely pathogen, inpatient vs outpatient setting, allergy history, severity, resistance concerns.

How should hypertension be treated?

Missing: blood pressure values, cardiovascular risk, comorbidities such as diabetes, CKD, or heart failure, and current treatment.

✓ Clear Questions (the system can answer precisely)

For a 65-year-old patient using aspirin for secondary cardiovascular prevention, what is the recommended dose?

Age is clear. Indication is clear. The question is tightly focused on dosing.

For a 65-year-old patient using aspirin for secondary cardiovascular prevention with chronic kidney disease (eGFR 45 mL/min/1.73m²), does the dose need adjustment?

Age is clear. Comorbidity and a specific value are provided. The decision point is dose adjustment.

In hospitalized community-acquired pneumonia, PORT class III, no penicillin allergy, what is the preferred initial antibiotic regimen?

Infection site, severity level, and allergy history are clearly stated.

PICO Framework — the gold standard for building clear clinical questions

P

Patient

Patient characteristics: age, sex, comorbidities, lab values, disease stage, risk level.

I

Intervention

Intervention details: medication, surgery, dose, route, frequency, duration, device, or strategy.

C

Comparison

Optional comparator: another drug, another procedure, standard care, or no treatment.

O

Outcome

Outcome of interest: survival, symptom control, adverse events, recurrence, or dose adjustment.

3. MedXY ≠ a General LLM

Do not use MedXY the same way you use a general AI assistant such as ChatGPT. The design goals, working method, and ideal use cases are fundamentally different.

No role prompting needed

“You are a professor-level medical expert, please help me...”

Ask directly: “A 65-year-old patient, ASA class III, coronary artery disease, planned knee replacement...”

MedXY already has an evidence-based medical knowledge system built in. Role prompts do not improve answer quality.

No open-web literature search

“Search PubMed for the latest papers on...”

Ask the clinical question directly and let the system retrieve from a curated high-quality medical knowledge base.

The system intentionally avoids low-quality sources to reduce misinformation and hallucination risk.

No fabricated references

A general LLM may generate plausible-sounding but nonexistent DOI numbers, author names, or journals.

MedXY cites only real guidelines and literature that exist in the knowledge base.

Answers are grounded in retrieved evidence fragments rather than invented by free generation.

Focused on medicine, not general-purpose conversation

“Write a poem about the heart” / “Draft a meeting invitation email”

Clinical decision support, medication questions, diagnostic reasoning, guideline interpretation, and medical knowledge queries

Non-medical tasks are outside the product’s primary design goal.

Quick comparison

Dimension General LLM (e.g. ChatGPT) MedXY
Knowledge source Static training data with a cutoff Retrieval from a medical evidence knowledge base
Role prompting needed Sometimes useful No — unnecessary
Web search Available in some versions No open-web browsing by design
Reference citation May hallucinate Grounded in real sources
Clinical reasoning depth General-purpose, often shallow for evidence work Specialized and optimized for medical evidence
Medical knowledge quality control Mixed and unfiltered Curated high-quality sources only
General writing / conversation Strong Not the primary design goal

4. What Can MedXY Do?

You can ask directly in all of the following areas. No role setup or identity prompt is needed.

Clinical diagnosis & differential diagnosis

Input symptoms, signs, imaging, or lab results to obtain a structured differential diagnosis and reasoning pathway.

Example: A 38-year-old woman with arthralgia, malar rash, and positive ANA — what diagnoses should be considered?

Medication use & dosing

Ask about indications, contraindications, renal or hepatic dose adjustment, route, timing, and optimization.

Example: In a kidney transplant recipient on tacrolimus, how should the dose be adjusted based on trough concentration?

Drug-drug interactions

Review mechanisms, clinical significance, and management recommendations for combined medications.

Example: What changes in PT/INR are expected when warfarin is combined with fluconazole?

Treatment strategy & guideline recommendations

Review first-line and second-line treatment, escalation strategy, and guideline-based sequencing.

Example: What is the standard medication regimen for HFrEF with an EF of 35%?

Medical scores & clinical formulas

Interpret CHA₂DS₂-VASc, MELD, APACHE, Wells, CURB-65, and many other clinical scoring systems.

Example: For atrial fibrillation with a CHA₂DS₂-VASc score of 3, is anticoagulation indicated?

Rare diseases & genetic conditions

Ask about diagnostic criteria, inheritance patterns, genetic testing suggestions, and current treatment options.

Example: A child has episodic ataxia — which rare diseases should be considered?

Adverse effects & toxicity

Review common or rare adverse reactions, toxicity monitoring, and evidence-based management approaches.

Example: How should immune checkpoint inhibitor-related pneumonitis be graded and managed?

Interpretation of tests and laboratory markers

Explore the clinical significance, reference interpretation, confounding factors, and next-step management.

Example: BNP is markedly elevated but echocardiography is normal — what should be considered?

Medical history & basic mechanisms

Understand disease pathophysiology, mechanism, and the historical background of important medical concepts.

Example: What were the major contributions of the Framingham Study to cardiovascular prevention?

5. Advanced Tips: How to Get More Precise Answers

1. Give specific values instead of vague descriptions

Kidney function is poor

eGFR 28 mL/min/1.73m² (CKD stage 4)

2. List all current medications

The patient is taking several drugs

Current medications: warfarin 3 mg/day, digoxin 0.125 mg/day, furosemide 20 mg/day

3. State your decision target explicitly

About this patient...

I need to decide whether to start anticoagulation now or wait until repeat testing.

4. Split complex problems into steps

If your case includes multiple decision points, ask the main question first and then follow up step by step.

First ask: What is the preferred drug? Then ask: How should the dose be adjusted in renal impairment?

5. Tell the system if this is a special population

Pregnancy, lactation, advanced age, pediatrics, or immunosuppression can all change the recommended strategy.

The patient is 28 weeks pregnant with hyperthyroidism — which antithyroid drugs are appropriate?

6. Mention regional or resource context when relevant

If you want the answer aligned with a specific national context or a limited-resource setting, say so explicitly.

In a primary hospital without CRRT capability, how should this patient with acute renal failure be managed?

6. Common Mistakes

Mistake 1: Using MedXY like a search engine

A search engine returns a list of links. MedXY returns an integrated, evidence-based answer. Use full questions rather than keyword-style searches.

Mistake 2: Expecting it to replace clinical judgment

MedXY provides evidence-based reference support. Final clinical decisions must be made by licensed physicians based on the specific patient context. The system does not assume medical responsibility.

Mistake 3: Assuming longer prompts are always better

You do not need long decorative prompting. Two or three sentences are usually enough if they contain the key patient information and a clear question.