DEV Community: Sergey Kasev

Doctor, patient, insurer: why nobody has time to explain your lab results

Sergey Kasev — Sat, 24 Oct 2026 12:16:15 +0000

In every visit there are three parties in the room: the patient, the medical record, and whoever pays. Everyone acts reasonably, and the patient still leaves with a lab report that nobody explained.

I'm building ZaZdorovie, an early-stage prototype that turns lab reports into clear summaries for patients and clinicians. It's not a medical device and has no revenue yet. Recently I watched a pitch from a Russian startup that turns doctor-patient conversations into medical records. Investors got most excited about one phrase: "insurance terror". Clinics fear a mistake in the paperwork more than a confused patient. That made me look at how the three links actually connect.

Link 1. The doctor: paperwork eats the visit

For every hour of face time with patients, doctors spend nearly two more hours on the record and desk work (Sinsky et al., Ann Intern Med 2016).
The record isn't bureaucracy for its own sake. It shows what was done, and it is what gets paid.
That's why AI scribes that turn the conversation into notes are growing fast. The time they free up could go to explaining.

Link 2. The payer: pays and checks

United States. In ACA Marketplace plans, insurers denied nearly 1 in 5 in-network claims in 2023 (KFF). In the American Medical Association survey, physicians and their staff reported about 13 hours a week spent on prior authorizations (AMA, 2024).
Russia. In compulsory insurance (OMS), insurers and funds run medical-economic and quality reviews and can cut payment to clinics for defects in records (Ministry of Health order No. 231n of 19.03.2021). Voluntary insurance (DMS) has its own checks in each contract.

The result is the same everywhere: documentation wins over explanation, because documentation is what gets checked.

Link 3. The patient: has the result, not the meaning

Patients forget 40–80 % of what they hear in a visit (Kessels, 2003).
Results now reach the phone before a clinician looks at them. Among patients with abnormal results, 16.5 % felt more worried after reading them, compared with 5 % with normal results (Steitz et al., JAMA Netw Open 2023).
When nobody explains, forums do. In 154 public complaints we collected, "nobody explained it" and "everyone says something different" appear in almost every topic, from ferritin to cholesterol.

Where the chain breaks

Who	What they lose	What helps
Patient	Doesn't understand the result or the next step; comes back with the same questions	Preparing for the visit, saying the plan back, getting it in writing
Doctor	Time spent explaining the same things; follow-up messages	Clear handouts and one consistent format for results
Payer and clinic	"Just to ask" visits; patients who never come back for follow-up	Transparent, consistent results and a plain-language explanation for the person

What lab explanations can and can't do for payers

I want to be precise here, because it's easy to overpromise.

What a transparent lab explainer can give:

Reproducible interpretation. Every calculated index uses a published formula with a source. The same report always gives the same output, so anyone can check it.
One format. Reports from different labs, languages and units become one table.
A clear summary for the member after a check-up. Whether this reduces "just to ask" visits and calls is a hypothesis. It can only be proven in a pilot.

What it can't do:

It doesn't review or fix the clinic's visit documentation.
It doesn't protect a clinic from claim denials or payer audits.
It doesn't judge whether treatment follows guidelines.
It doesn't make medical decisions. The clinician does.

So it's a different job from AI scribes, and the two fit together: the scribe covers the conversation and the record, the lab explainer covers the results and the patient's understanding.

What you can do as a patient

Ask who will explain your results. Many insurance plans have a nurse line or a care coordinator.
Get copies of your results and keep them in one place, sorted by date. Trends matter more than one number.
Bring three questions: what is important, what to do before the next visit and when to retest, which symptoms mean "come back sooner".
Say the plan back in your own words before you leave.
Ask the insurer, not the clinic, what your plan covers.

Chest pain, shortness of breath at rest, sudden weakness or trouble speaking: call your local emergency number (911 in the US, 112 in the EU).

Where we are (honestly)

ZaZdorovie is an early-stage prototype, built by a solo founder, with no revenue and no paying clients yet. The working version is in Russian. It is not available to US patients. Outside Russia we offer it only as software licensed to clinics, labs, insurers' health programs and health-tech teams, running on their own servers, with patient-facing interpretation switched off where regulation requires it. It's an information tool, not a medical device.

If you work at a clinic, a lab, an insurer or an AI-scribe company and want to test whether clear lab explanations change anything measurable, write to me: info@projectinvest.eu · project page.

This article is not medical, legal or insurance advice. Sources are linked in the text.

Disclosure: drafted with help from AI tools; the facts, numbers and sources were checked by the author.

You left the doctor's office and understood nothing. That's normal, and here's what helps

Sergey Kasev — Sat, 24 Oct 2026 12:16:04 +0000

"We talked for an hour, and at home I couldn't remember a single word except 'everything is fine'."

If that sounds familiar, nothing is wrong with you. This is how memory works under stress, and it is well studied. We read 154 public complaints about lab results and doctor visits, and "nobody explained it to me" was one of the most common messages.

I'm building ZaZdorovie, an early-stage prototype that turns lab reports into clear summaries and questions for the doctor. It is not a medical device and does not serve patients at scale. This article is not about the product, though. It is about the problem and what research says helps.

Why we forget what the doctor said

Most of it is gone right away. A review in the Journal of the Royal Society of Medicine found that patients forget 40–80 % of medical information immediately after the visit, and almost half of what they remember is remembered wrongly. The more information at once, the less stays (Kessels, 2003).
Worry blocks listening. When the doctor says something scary, attention locks onto one word and the rest passes by.
Doctors are short on time. A time-and-motion study found that for every hour of face time with patients, doctors spend nearly two more hours on the medical record and desk work (Sinsky et al., Ann Intern Med 2016). Explanations are the first thing to get cut.
Results get lost on the way. A systematic review of outpatient care found that the share of lab results without timely follow-up ranged from 7 % to 62 % across studies (Callen et al., J Gen Intern Med 2012). Another study found that about 1 in 14 clinically significant abnormal results was not communicated to the patient or not documented as communicated (Casalino et al., Arch Intern Med 2009).
Results now arrive before the explanation. In a survey of 8,139 US patients, almost everyone wanted immediate access to results. But among people with abnormal results, 16.5 % felt more worried after reading them, compared with 5 % with normal results (Steitz et al., JAMA Netw Open 2023).

An example

This example is based on typical complaints; any similarity is a coincidence.

Olga, 52, remembers only one sentence from her visit: "Your sugar is a bit high, keep an eye on it." At home she can't recall whether she needs a new test, when, or what "keep an eye on it" means. Six months later she comes back with the same questions and one more lab report.

Six things that help

Write down your top 3 questions before the visit. Put the most important one first. Time may run out at the end.
Bring your previous results. A trend over 2–3 years tells the doctor more than one new number.
Take notes, or bring someone with you. You can also ask if the doctor is OK with an audio recording.
Say it back in your own words at the end: "So I need a new test in 3 months, and until then I should…?" This is called teach-back. A systematic review found it improves understanding and adherence (Talevski et al., PLoS One 2020).
Ask for the plan in writing: what to do, when to retest, which symptoms mean "come back sooner".
Ask how you'll get your results. "We'll call you if something is wrong" is a weak plan: no call doesn't mean everything is fine.

Three questions for your doctor

"What is important in my results, and what can I ignore?"
"What should I do before the next visit, and when should I retest?"
"Which symptoms mean I should come back sooner or call emergency services?"

Chest pain, shortness of breath at rest, sudden weakness in an arm or leg, trouble speaking, fainting: call your local emergency number (911 in the US, 112 in the EU). Don't search for an explanation online.

Where we fit (honestly)

What these studies describe is a gap between having results and understanding them. That's the one problem ZaZdorovie works on: put all lab values in one table with each lab's own reference ranges, show trends, explain in plain words what each marker is, and suggest questions for the clinician. No diagnoses, no treatment advice.

Where we are: an early-stage prototype built by a solo founder. No revenue yet. The working version is in Russian and still in testing. It is not available to US patients; outside Russia we only offer it as software licensed to clinics, labs and health-tech teams, running on their own servers. It is an information tool, not a medical device, and the clinician always decides.

If you work in a clinic, a lab or a patient portal and see this gap every day, I'd like to hear how you handle it: info@projectinvest.eu · project page.

This article is not medical advice. Sources are linked in the text.

Disclosure: drafted with help from AI tools; the facts, numbers and sources were checked by the author.

A white-label lab-report explainer for clinics: what's inside

Sergey Kasev — Sat, 24 Oct 2026 09:22:19 +0000

Your patients see their lab results in the portal before you do. Then they google. Then they call.

A lab-report explainer under your clinic's brand can take some of that load. Here's an honest look inside one: ZaZdorovie. It's an early-stage prototype we license to clinics and health-tech teams. Below: the components, the US configuration, and the parts we deliberately keep out.

Status: prototype, built by a solo founder, not providing services at scale, no revenue yet. Not FDA-cleared, not a medical device. Not offered to US patients directly.

The components

Component	What it does
Report reader	Takes a PDF or photo. Extracts values, units, the lab's reference ranges, patient name, date of birth and test date. Text-layer PDFs are read exactly. Scans go through OCR.
Marker dictionary	607 markers in 37 sections. Maps the many names labs use for one marker to a single entry.
Results table	All reports, all dates, one table: the lab's own range, the change in %, a chart per marker.
Indices engine	88 calculated indices (HOMA-IR, TyG, FIB-4, APRI, eGFR, NLR, TSAT…) with formula, thresholds and source, computed only when all inputs are present.
Clinician report	The table plus indices, each with its basis visible, so the clinician can check every number.
Marker guide	Plain-language "what is this marker and why is it measured" for patients. Educational, no conclusions about the individual.
Practice workspace	Clinicians, patient files, sharing by link with permissions, page-based billing, admin panel.
Branding	Your logo, colours, domain and wording.

The US configuration: what's off and why

The full product also has patient-facing interpretation ("this pattern may relate to…"), risk scores and a symptom concierge. In the US configuration, all of that is switched off. Here's why:

FDA. Software that only stores, converts or displays lab results is excluded from the device definition (FD&C Act 520(o)(1)(D)). Software that interprets results for a patient isn't excluded, and FDA's January 2026 General Wellness guidance excludes anything with clinical thresholds, diagnoses or treatment suggestions. Clinician-facing decision support can be non-device when the clinician can independently review the basis (FDA Clinical Decision Support guidance, January 2026). So in the US we keep interpretation on the clinician side, with every basis visible. Your counsel should confirm this for your use case. This isn't legal advice.
HIPAA. Your clinic is a covered entity. Any vendor that hosts or sees your patients' data is a business associate and needs a BAA. We don't host US patient data and don't sign BAAs. The software runs on your infrastructure, under your HIPAA program and your cloud provider's BAA, and we work only with test or de-identified data.
FTC and state laws. No ad pixels in the patient area. No data sharing without explicit consent, which also matters under Washington's My Health My Data Act and California's CMIA. If you use AI-generated patient messages in California, AB 3030 requires a disclosure, which we add to the white-label wording.
Data location. Our own hosted service runs in Russia for Russian users. US data never goes there. That also matters under the US Department of Justice data security rule (28 CFR Part 202), which restricts access to bulk US health data from countries of concern.

How a deployment looks

Demo call on fictional data: 30 minutes, free.
Installed pilot, from $1,500: we help install it in your cloud (AWS, Azure or GCP) for 60 days and test it on your sample report formats. No patient data reaches us. The fee is credited toward the licence.
Licence: source-code or white-label, non-exclusive, from $3,500. Updates from $120/month. An exclusive US licence is possible (from $3,500/year + 4 % royalty) for a partner who takes the regulatory lead.

What we'd tell you before you ask

US lab formats are untested. Quest and Labcorp layouts need adaptation; that's what the pilot is for. mg/dL is partly supported.
The knowledge base is most complete in Russian. English UI exists, and the English texts are partial.
It's one founder. You get code, documentation and direct support, not a 24/7 vendor.
For scans, OCR currently uses an external vision model. For your deployment it runs locally, or through a provider covered by your BAA. This is fixed before install.

If you run a clinic, a lab or a health-tech product and want to see it on fictional data: info@projectinvest.eu · project page.

ZaZdorovie is an information tool, not a medical device. It doesn't diagnose, treat or replace a clinician. Prices are starting prices; the final price is agreed in writing before payment.

Disclosure: drafted with help from AI tools; the facts, numbers and sources were checked by the author.

Building a lab-report assistant as a solo founder (build in public #1)

Sergey Kasev — Sat, 24 Oct 2026 09:17:45 +0000

Every lab prints the same blood test differently: different names, units, column orders. I wanted all my results from different labs in one table, so I started building one. This is the first build-in-public post about ZaZdorovie: an early-stage prototype, built by one person with a lot of help from AI coding tools.

Upfront, so nobody is misled:

It's a prototype. It isn't providing services at scale and has no revenue yet.
It isn't a medical device and doesn't diagnose.
It isn't available to US patients. The hosted version runs in Russia for Russian users.

What it does today

You upload a PDF or a photo of a lab report.
It reads the values, units, the lab's own reference ranges, the patient's name and date of birth.
It matches each line to a marker in a dictionary of 607 markers in 37 sections: CBC, chemistry, lipids, glucose and insulin, iron, vitamins, hormones, thyroid, coagulation, organic acids and more.
It puts every result from every date in one table, with the change in % and a chart.
It calculates 88 indices where the inputs exist: HOMA-IR, TyG, FIB-4, APRI, eGFR, NLR, TSAT and others. Each comes with its formula, thresholds and the published source.
It produces two reports: a detailed one for a clinician and a plain-language one.
It lets you share a link with a doctor, with permissions.

There's an 86-second demo on fictional data on the project page.

The stack (boring on purpose)

Backend: Python + FastAPI, PostgreSQL via SQLAlchemy.
PDF text: PyMuPDF. Most lab PDFs have a real text layer, which is fast and exact.
Scans and photos: a vision model reads the page into text; local Tesseract OCR is the fallback. The model only transcribes. All interpretation happens in our own rules.
The "brain": a rules-and-indices engine in plain code and JSON knowledge files, not a chatbot. Every conclusion points back to the markers and thresholds behind it.
Frontend: a web app with a practitioner workspace, patient files and page-based billing.

Why not just send the PDF to a big model and ask "what does this mean"? Three reasons:

Reproducibility. The same report must give the same result every time.
Traceability. A clinician needs to see why a number is flagged.
Liability. A free-text answer about someone's health is the hardest thing to make safe.

What was hard

1. Marker names. "Ferritin", "Ферритин", "FER", "Feritină", "Ferritin (CMIA)" are all the same marker. Most of the work is a normalisation dictionary and tests for every lab format we've seen. Each new lab still breaks something.

2. Units and ranges. The lab's own reference range is the source of truth, not a textbook range. Conventional US units (mg/dL) are partly supported. Reports from US labs haven't been tested yet.

3. Photos. Phone photos of crumpled paper are where OCR fails. We ask the user to confirm doubtful values instead of guessing.

4. Dates and people. One upload can contain three reports from different dates, sometimes for different family members. The app reads name and date of birth and suggests the profile, but asks before attaching.

What I got wrong (so far)

Privacy shortcut in OCR. For scans, the current recognition step sends the whole page image, including the patient's name, to an external AI service. That's how the current Russian version works, and it's not acceptable for the EU or US. It's the first thing to fix before any deployment outside Russia: crop the header or keep OCR fully local.
Too much in one product. It has a symptom concierge, a disease encyclopaedia, metabolic maps… For the US I'm cutting it down to the core: organize results, show trends, a clinician-facing report.
Talking about features before regulation. In the US, software that interprets lab results for a patient is likely a medical device. Software that stores and displays lab results is not (FD&C Act 520(o)(1)(D)). Clinician-facing support whose basis the clinician can review may fall outside the device definition (FDA CDS guidance, January 2026). So the US configuration keeps interpretation for clinicians only. I wrote this up properly before writing any US marketing.

What's next

Fix the OCR privacy issue.
Test reports from the most common US lab formats on fictional or de-identified samples.
Find 1–2 US clinics or health-tech teams for an installed pilot: the software runs on their servers, and their patient data never leaves their environment.

We're licensing the software (source-code or white-label, on the buyer's own infrastructure) from $3,500, with an installed pilot from $1,500. Those are honest prototype prices, not enterprise ones.

If you've built anything that parses lab PDFs, I'd love to compare notes: what broke for you? info@projectinvest.eu

Not medical advice. ZaZdorovie is an information tool and doesn't replace a clinician.

Disclosure: drafted with help from AI tools; the facts, numbers and sources were checked by the author.

Why people don't understand their own lab results: what 154 complaints told us

Sergey Kasev — Sat, 24 Oct 2026 09:17:18 +0000

"I got my results, nobody explained anything. What's the next step?"

That's one real forum post. We read 153 more like it.

I'm building ZaZdorovie, an early-stage prototype of a lab-report organizer. It is not a medical device and does not serve patients at scale. Before writing more code, I wanted to know what people actually struggle with when they hold a lab report in their hands. So we collected public complaints and checked each recurring pain against the research.

How we collected them (and the limits)

154 public quotes with links, 2012–2026, most from 2024–2026.
Sources: Russian-language forums and app reviews. A women's forum makes up 56 % of the sample, then Pikabu, a doctor Q&A site, regional forums and lab-app reviews in RuStore.
Skewed sample. We searched by topic ("ferritin", "results normal but…", "check-up"). The counts show what hurts for people who post online, not how common it is in the population. Posts that looked like paid promotion were excluded.

So this is qualitative research with rough numbers, not an epidemiological study. Still, the patterns match what US studies find. Links are below.

What people complain about

We grouped 22 raw topics into 9 clusters:

Cluster	Complaints	Share
Iron and ferritin: "haemoglobin is normal, but I'm exhausted"	27	17.5 %
"I have my results and they mean nothing to me" (nobody explained)	26	16.9 %
Myths, contradictory advice, distrust	24	15.6 %
Metabolism: blood sugar, weight, cholesterol	23	14.9 %
The system doesn't lead anywhere (check-ups "for the record")	18	11.7 %
Money: expensive check-up panels	17	11.0 %
Results scattered across labs, no history	7	4.5 %
Vitamin D: "I tested it, now what?"	7	4.5 %
Thyroid: TSH vs how I feel	5	3.2 %

There's no neat 80/20 here: 5 of 9 clusters cover 77 % of complaints. But one thread runs through almost every cluster: "nobody explained it to me" and "everyone tells me something different."

People don't lack tests. They lack a translation from numbers into understanding and a good question for their doctor.

Five reasons lab reports are hard to read

1. "Normal" doesn't mean what people think

A reference range is usually set so that about 95 % of healthy people fall inside it. A value slightly outside can be perfectly fine for you, and a value inside the range doesn't guarantee everything is fine. MedlinePlus explains this well (How to Understand Your Lab Results). Almost nobody gets this explanation at the counter.

Example from our sample: ferritin of 10 with "normal" haemoglobin, and a doctor saying "it's normal." A recent review puts ferritin below 30 µg/L as a reliable sign of iron deficiency even with normal haemoglobin (PMC11247274). The two values answer different questions, and the patient sees only the green tick on one of them.

2. Results arrive before the explanation

In the US, the 21st Century Cures Act made many results visible in patient portals immediately, often before a clinician looks at them. A survey of 8,139 patients at four US academic centres found that 96 % still prefer immediate access. But among people with abnormal results, 16.5 % felt more worried after reading them, compared with 5 % of those with normal results (Steitz et al., JAMA Netw Open 2023). Access solved one problem and exposed another: understanding.

3. Different labs, different numbers

Units, methods and reference ranges differ between labs, so values from two labs can't always be compared directly (MedlinePlus, same page). People who switch labs, which is most people, end up with a pile of PDFs that don't line up. Seven complaints in our sample were exactly that: "my history is lost across lab apps."

4. Health literacy is low everywhere

In the last US national assessment, only 12 % of adults had proficient health literacy (NCES, The Health Literacy of America's Adults, 2006). Lab reports are written for clinicians. Most of the people reading them are not clinicians.

5. Short visits and check-ups without a next step

A Cochrane review found that general health checks for healthy adults don't reduce mortality by themselves (Krogsbøll et al., 2019). The benefit comes from what happens after the result. Our "check-up for the record" cluster (18 complaints) describes exactly the missing "after."

The myths fill the gap

When nobody explains, forums do. In our sample:

"Diabetes is when sugar is 30" (it isn't; diagnostic thresholds are much lower);
one value is read as a diagnosis ("you have hypothyroidism");
hidden supplement advertising sits in the threads.

The research says something calmer:

Prediabetes is often reversible. About 150 minutes of activity a week and losing 5–7 % of body weight cut the risk of progressing to diabetes (Echouffo-Tcheugui et al., JAMA 2023;329(14):1206–1216).
A weight plateau is normal physiology, not a "broken metabolism" (Endotext, NBK278963).
Fatigue in adults is most often about sleep, stress or mood. Anaemia and other organic causes are a minority (Deutsches Ärzteblatt review, PMC8579431).

What would actually help

From the complaints, the useful thing is not another diagnosis engine. It is:

One table: all results from all labs and dates together, with each lab's own range and the trend.
Plain language: what each marker is and why it's measured.
A good question for the doctor instead of a conclusion.
Linked values side by side: ferritin next to haemoglobin, HbA1c next to glucose.

That's what we're building. To be clear about where we are: ZaZdorovie is an early-stage prototype. The working version is in Russian and still in testing, and it is not available to US patients. Its US version is offered only as software licensed to clinics and health-tech teams, running on their own servers, with patient-facing interpretation switched off. It's an information tool, not a medical device, and nothing in it replaces a clinician.

If you build patient portals or work in a clinic and recognise these problems, I'd like to hear how you handle them: info@projectinvest.eu · project page.

Method note: 154 public quotes, Russian-language sources, collected October 2026. The dataset (quotes with links) is available on request. This article is not medical advice.

Disclosure: drafted with help from AI tools; the facts, numbers and sources were checked by the author.