Science & Methodology

1. How Blood Test Interpretation Works

Hema's interpretations are generated using a combination of a curated clinical knowledge base and AI-powered text generation:

• Knowledge Base: We maintain a comprehensive knowledge base compiled from authoritative clinical sources: Cleveland Clinic, Mayo Clinic, MedlinePlus (NIH), NCBI (PubMed), and peer-reviewed medical literature.
• AI Generation: Interpretations are generated by Claude (developed by Anthropic), a large language model. The AI is explicitly instructed to ground its outputs in our clinical knowledge base and NOT draw on general training knowledge for medical claims.
• Prompt Engineering: We use structured prompts that provide the AI with your specific biomarker values, demographic information, and the relevant clinical context from our knowledge base.

This approach combines the comprehensiveness and accessibility of AI with the rigor of curated clinical content.

2. Biomarker Classification System

Hema uses a three-tier classification system for each biomarker:

This classification is based on distance from the optimal range, not a binary in/out determination. A value that is 1% outside optimal is treated differently than one that is 50% outside.

3. Composite Health Score

Your overall health score is a weighted average of all submitted biomarker classifications:

For example, if you submit 10 biomarkers and 8 are optimal, 1 is borderline, and 1 is low, your score would be: ((8×1.0 + 1×0.65 + 1×0.15) / 10) × 100 = 88.

4. Cardiovascular Risk (ASCVD)

When you provide the required inputs (age, sex, race, cholesterol values, blood pressure, diabetes status, smoking status), Hema calculates your 10-year ASCVD (atherosclerotic cardiovascular disease) risk score.

Methodology

• Based on the 2013 ACC/AHA Pooled Cohort Equations
• This is a deterministic mathematical calculation — not AI-generated
• Valid for individuals aged 40-79
• Uses four separate equations based on sex and race (White Female, Black Female, White Male, Black Male)

Known Limitations

• May overestimate risk in contemporary low-risk populations compared to when the equations were developed
• Equations were derived from specific cohort studies and may not generalize to all populations
• Does not account for all cardiovascular risk factors (e.g., family history, Lp(a), inflammation markers)

5. 12-Week Protocol Generation

Your personalized 12-week health protocol is generated by Claude AI, grounded in the Hema knowledge base. The protocol covers nutrition, exercise, supplementation, and lifestyle modifications.

Exercise Recommendations

Exercise selections follow an evidence-based tier system based on your stated fitness level:

• Beginner: Low-impact, foundational movements with emphasis on form and consistency
• Intermediate: Progressive overload principles with moderate intensity
• Advanced: Higher intensity training with periodization

Nutrition Calculations

• Calorie targets are calculated using the Mifflin-St Jeor equation (basal metabolic rate) adjusted for activity level (TDEE)
• Protein recommendations are capped at 1.0g per pound of bodyweight, consistent with evidence-based guidelines
• Macronutrient ratios are adjusted based on your goals and biomarker profile

Supplementation

Supplement recommendations are based on your biomarker results and are graded by evidence quality. We distinguish between supplements with strong clinical evidence and those with emerging or limited evidence.

6. Reference Range Sources

For each biomarker, our optimal reference ranges are derived by averaging values from 3-5 authoritative clinical sources:

• Cleveland Clinic
• Mayo Clinic
• LabCorp reference ranges
• NIH MedlinePlus
• NCBI / PubMed clinical literature

Note: These ranges may differ slightly from your specific laboratory's ranges due to differences in testing methodology and population calibration. Always review your results alongside your lab's provided ranges.

7. Limitations

We believe in being transparent about what Hema can and cannot do:

• Not a diagnostic tool: Hema cannot diagnose medical conditions. Only a qualified healthcare provider can make diagnoses.
• Limited context: We only know what you tell us. We cannot account for your full medical history, medications, or circumstances you haven't disclosed.
• AI imperfection: While we use state-of-the-art AI with clinical grounding, AI outputs can occasionally contain errors or oversimplifications.
• General guidelines: Protocols are general recommendations and may not be appropriate for individuals with specific health conditions.
• Not real-time: Blood tests are a snapshot in time. Results and recommendations should be considered alongside trends and clinical context.
• US-centric: Reference ranges and clinical guidelines are primarily sourced from US institutions and may not apply globally.

Questions?

If you have questions about our methodology or want to learn more about how Hema works, contact us at science@hema.health.