I’ve been feeling depleted for two years. Not sick. Not in crisis. Just… 65%. Like the brightness on my screen was permanently dimmed.
My doctors said my labs were “normal.” My psychiatrist said the medication was working. My new job was going well. By every external measure, I was improving.
Then I fed five years of medical records to Claude and asked it to look at everything at once. What it found stopped me cold: every objective health marker was worse than it had ever been. My body was in the worst shape of my adult life — and I was the only one who couldn’t see it, because I finally felt “okay.”
The road got smoother. But the car was running on empty.
Why I Did This
We optimize our investment portfolios. We track our net worth to the penny. We run Monte Carlo simulations on our retirement date. But how many of us have ever looked at our health data with the same rigor?
Your body is an asset. Unlike your portfolio, it can’t be rebalanced after a crash. And unlike your career, you can’t pivot to a new one.
I had five years of records sitting in my patient portal — lab panels, imaging reports, specialist notes — and nobody was looking at them as a whole. Each doctor saw their silo. Nobody saw the movie.
So I downloaded everything and asked an AI to watch the whole movie at once.
The Setup
I’m a 40-year-old woman. Two kids under 8. Tech career. I’ve spent the past few years dealing with pregnancy complications, multiple COVID infections, postpartum recovery, and the creeping exhaustion that comes with managing a household, a career, and your own unraveling health.
I see an internist, a gynecologist, a psychiatrist, a neuropsychologist, and a geneticist. They’re all good. But each one sees me through their own lens. My internist looks at my bloodwork. My GYN tracks my periods. My psychiatrist manages my anxiety medication. Nobody connects the dots.
I told Claude: “You are my concierge physician. Review everything. Ask me questions. Then tell me what you see.”
What I Gave It
- 56 medical encounters (office visits, telemedicine, ER visits, lab draws) spanning 2021-2026
- 15 complete lab panels including CBC, metabolic panel, lipids, thyroid, iron studies, vitamins, hormones
- A brain MRI (with images)
- A neuropsychological evaluation (8-page report with 30+ cognitive test scores)
- A 70-gene cancer panel
- Breast imaging (mammogram + ultrasound reports)
- My current symptoms, medications, diet, exercise, and sleep patterns
Total: approximately 1,000 pages of medical records, images, and reports.
Claude read it all in about 15 minutes. Then it asked me questions for an hour. Then it showed me what nobody had seen.
Pattern 1: The Vitamin D Collapse
My Vitamin D levels over 4 years:
| Year | Vitamin D (ng/mL) | Status |
|---|---|---|
| 2021 | 53.8 | Sufficient |
| 2023 | 38.3 | Sufficient (declining) |
| 2024 | 33.8 | Barely sufficient |
| 2025 | 21 | Insufficient |
A 61% decline over 4 years. Each individual result was discussed at its annual visit. But nobody flagged the trajectory. A Vitamin D of 21 in isolation gets you a “take some supplements” and a follow-up in 6 months. But a Vitamin D that was 53.8 and is now 21? That’s a body running out of something fundamental — and it maps directly to my 4-year pattern of getting sick every few months.
Pattern 2: The Cholesterol Crash Nobody Questioned
| Year | Total Cholesterol | LDL |
|---|---|---|
| 2021 | 137 | 65 |
| 2023 | 140 | 57 |
| 2024 | 157 | 88 |
| 2025 | 114 | 54 |
My total cholesterol dropped 27% in a single year. In standard medicine, low cholesterol is celebrated. My doctor didn’t blink.
Claude flagged it immediately: “A 27% decline in total cholesterol in one year is not ‘good.’ Cholesterol is the building block for every steroid hormone in your body — estrogen, progesterone, testosterone, cortisol. At 114, you may not have enough raw material for hormone production.”
When I tracked my protein intake for a week, it was 65 grams per day. For my body weight, I needed 100-120 grams. I was eating 55% of what my body required to function. The cholesterol crash wasn’t a mystery. It was malnutrition hiding behind “normal” labs.
Pattern 3: The Sinus Disease Nobody Followed Up On
In early 2025, I had a brain MRI for cognitive complaints. The result: “No acute infarction, intracranial hemorrhage, or mass. Severe paranasal sinus disease.”
The brain was fine. Great news. But that second finding — “severe paranasal sinus disease” — was an incidental note in the radiology report. Nobody followed up.
Meanwhile, I’d been getting sick every few months for four years. That’s 9-10 respiratory illness events in 4 years — including three COVID infections while pregnant.
Claude connected two findings from two different medical silos: chronic sinus disease was serving as a reservoir for recurrent infections, and my crashed Vitamin D (critical for immune function) meant my body couldn’t clear them. Two findings. Two specialists. Nobody put them together.
Pattern 4: The Iron Paradox
| Marker | 2021 | 2025 | Ref Range |
|---|---|---|---|
| Iron | 183 (above range) | 146 | 50-170 |
| Iron Saturation | 60% (above range) | 49% (high) | 15-50% |
| Ferritin (stores) | 58 | 38 | 12-204 |
High iron saturation with declining ferritin is an unusual combination. Each individual lab was “in range” or close to it. But the pattern across time was significant — it meant supplementing with iron pills could potentially push me toward overload, even though my stores were dropping.
No doctor had compared these two draws side by side. Claude did it in seconds and flagged the risk.
Pattern 5: Every Blood Count Declining in Unison
| Marker | 2021 | 2024 | 2025 | Trend |
|---|---|---|---|---|
| White blood cells | 5.3 | 7.4 | 4.5 | Dropped 40% |
| Red blood cells | 4.16 | 4.36 | 3.80 | Below range |
| Hemoglobin | 13.2 | 13.5 | 11.8 | Near floor |
| Platelets | 200 | 224 | 190 | Gradual decline |
Not one individual marker was flagged, because each was still technically “in range.” But every single count moved in the same direction at the same time. That pattern — everything declining together — paints a picture of systemic depletion that individual snapshot readings miss entirely.
Pattern 6: The Mind-Body Disconnect
This was the one that stopped me.
I felt my worst in 2022-2023 — postpartum, anxious, overwhelmed, unable to function some days. But my labs from that period were actually still decent. My body had reserves. It was running through a hurricane, but the gas tank was full.
By 2025, I felt better. I’d started an antidepressant. Changed jobs. Resolved some personal issues. I was no longer in crisis. By every subjective measure, I was improving.
But my labs hit all-time lows. Every marker — vitamin D, iron stores, B12, cholesterol, hemoglobin — worse than it had ever been. My body was objectively failing while my mind was finally at peace.
Claude explained it in a way I’ll never forget:
“You burned through your reserves during the crisis. The medication masked the signals. The new job gave you psychological relief. But nobody refilled the tank. You feel better because the road is smoother. But the car is in worse shape than it’s ever been.”
I had been running on fumes — just more comfortably.
What Claude Actually Does (That Your Doctor Can’t)
Claude isn’t a doctor. It can’t examine you, order tests, or prescribe medication. But it can do three things no individual specialist can:
It reads everything at once. My internist doesn’t see my neuropsych report. My GYN doesn’t look at my immunology trends. My psychiatrist doesn’t examine my iron studies. Claude sees all of it simultaneously and asks: what story does this data tell when you read it as one document?
It tracks trends, not snapshots. A Vitamin D of 21 at one visit is a footnote. A Vitamin D that went from 53 to 21 over 4 years is a collapse. Doctors see individual frames. Claude sees the time-lapse.
It connects findings across silos. Severe sinus disease + crashing Vitamin D + declining immune markers = an explanation for 4 years of recurrent infections. That connection requires seeing records from radiology, immunology, and primary care at the same time. No single doctor does that. It’s not their fault. The system isn’t built for it.
The AI doesn’t replace your doctor. It gives your doctor a cheat sheet they never had time to make.
What It Built For Me
After the analysis, Claude created a specific protocol based on my actual data — not generic wellness advice, but targeted interventions mapped to my specific deficiencies:
- A meal-by-meal protein protocol to get from 65g to 100-120g/day, adapted for my food preferences and dietary restrictions
- A supplement stack targeting only what my labs showed was depleted (not a kitchen-sink multivitamin)
- A sleep protocol addressing my specific pattern of delayed sleep onset
- A lab schedule for my next draw, including tests that had never been ordered
- Specialist referrals — the right type of specialist for my situation, with specific providers in my area
The Cost Comparison
| Claude | Concierge Medicine | |
|---|---|---|
| Records reviewed | All 56 encounters + labs + imaging | Same |
| Time to analysis | ~15 minutes | 2-4 weeks |
| Total consultation | ~2 hours | 3-5 hours |
| Cost | ~$20 | $2,000-5,000 |
| Follow-up | Instant, any time | By appointment |
How to Do This Yourself
Step 1: Download your records. Log into your hospital’s patient portal (MyChart, Epic, etc.). Look for “Download My Data” or “Export Health Records.” Download everything — you want the full CCD/CDA files, not just the summary view.
Step 2: Open Claude. You need Claude Pro ($20/month). The free version doesn’t have a large enough context window for this.
Step 3: Upload and prompt. Drag the files in. Use this exact prompt:
“You are my concierge physician in the style of Peter Attia. Review all of my medical records. Create a longitudinal table of every lab value across all years. Identify any declining trends, abnormal patterns, or findings that were noted in one specialty but never followed up by another. Ask me clarifying questions about my symptoms, diet, sleep, and lifestyle before giving your assessment.”
Step 4: Answer its questions honestly. Claude will ask about your diet, sleep patterns, exercise, medications, symptoms. The more specific and honest you are, the better the analysis. Don’t hold back the embarrassing stuff — it’s not judging you, and the details matter.
Step 5: Take the output to your real doctor. Print or email the summary. Bring it to your next appointment. Say: “I ran my records through an AI and it flagged these patterns. Can we discuss them?” Your doctor may disagree with some findings. That’s fine. The value is in the conversation it starts.
What’s Next
I’m implementing the protocol now. More protein — starting with a scoop of protein powder in my morning coffee, which took zero effort. Consistent sleep schedule. Targeted supplements for my specific deficiencies. The specialist appointments are booked.
In 90 days, I’ll recheck every lab. The hypothesis is clear: if the depletion was caused by pregnancy + repeated illness + insufficient nutrition, then targeted repletion should reverse the trends. My 2021 baseline proves my body knows how to be healthy. It just needs back what it lost.
I’ll share those results in Part 2.
Your health data is sitting in a patient portal right now. Nobody is looking at all of it at once. For $20 and two hours, you can change that. The patterns might surprise you — or they might explain something you’ve been feeling for years.