hayat help
Help › Hayat Health

Predictions and prescribing

health.hayaterp.com ↗

For doctors — the risk flags Hayat Health raises from guideline rules (diabetes, hypertension, CKD, anaemia, dyslipidaemia, thyroid, liver, BMI, vitamin D) and from trained models, with their inputs and guideline; confirm, reject or share with the patient; trend projections; generic medicine suggestions with allergy, kidney, liver, pregnancy and interaction checks; anticipated needs; and writing and approving prescriptions.

Everything on this page is decision support for a licensed doctor. The patient sees nothing here until you share or approve it. You need the patient's predictions consent for flags and the medicines consent for suggestions and prescriptions.

Risk flags

Hayat looks at every patient's confirmed values (after each new value, and once a day) and raises flags for you to review. Each flag shows its title, stage, severity (info, watch, high, urgent), the plain explanation, the inputs used (each value with its date), the rule and the guideline it follows, and a confidence.

Rule flags work from the first day:

Area Examples Guideline
Diabetes Confirmed (two diagnostic results), possible, prediabetes — HbA1c ≥ 6.5 % / 5.7–6.4 %, FBS ≥ 126 / 100–125 ADA 2025
Hypertension High-normal, grade 1, 2, severe — from the 30-day home average (≥ 3 readings on ≥ 2 days) ISH 2020
Kidney CKD G / A stage from eGFR (CKD-EPI 2021) and UACR; chronic vs possible; rapid decline KDIGO 2024
Anaemia Mild / moderate / severe by sex and pregnancy; iron-deficiency hint (MCV, ferritin) WHO 2024
Lipids LDL ≥ 190 / 160 / 130, TC ≥ 240, low HDL, TG ≥ 150 and ≥ 500 ACC/AHA 2018
Thyroid TSH > 10, 4.5–10, < 0.1, 0.1–0.4 (with FT4) ATA
Liver ALT / AST patterns (fatty liver, hepatitis, acute), cholestatic ACG 2017
Weight BMI with Asian cut-offs (23 / 27.5) WHO 2004
Vitamin D < 20, < 12 ng/mL —

Model estimates (e.g. risk of progressing from prediabetes to diabetes in 12 months, CKD stage change, BP staying uncontrolled) appear only when the Hayat team has trained and activated a model that passed its quality checks, and only above a minimum probability. They show the model version and the probability.

Refresh re-runs the rules for a patient now.

Review a flag

Action Effect
Confirm (optional note, condition name, stage) Adds the condition to the patient's record; counts as a "yes" label for training
Reject (optional note) Closes the flag; counts as a "no" label
Share with patient (note required) The patient sees the finding with your note and a disclaimer, and gets a notification
Unshare Hides it from the patient again

Write the share note for the patient: what it means for them and what to do next. You can also add or resolve conditions yourself.

Projections

A test's chart shows a dotted projection for the next months from the patient's own values (at least 4 over 30 days or more), with an uncertainty band — e.g. HbA1c, weight, BP. It is labelled a projection: a trend, not a prediction or a diagnosis.

Medicine suggestions

For each confirmed or flagged condition you see first-line and second-line generic options from the guideline — e.g. type 2 diabetes → metformin (unless eGFR < 30), hypertension → amlodipine, an ACE inhibitor or an ARB, dyslipidaemia → a statin — with strengths, the usual dose, frequency and days, informational brand examples, and a safety level:

Safety Means
OK No problem found in the record
Caution Something to weigh — e.g. reduce the dose for eGFR 30–45
Avoid A clear reason not to — allergy group, eGFR below the limit, pregnancy, liver disease, high potassium, a serious interaction, or a duplicate of a current medicine

Each check says what it found (eGFR 28 — avoid below 30). Checks use only what is in the record: an empty allergy list means none recorded, not none.

Anticipated needs lists trends that suggest therapy may soon be indicated — e.g. LDL rising towards the statin threshold, with the projected date — and current medicines due for review. Refills shows when each prescribed medicine runs out.

Patients never see suggestion lists.

Write and approve a prescription

  1. Start a prescription for the patient — from a suggestion or blank. Add lines: generic (from the catalogue or typed), strength, dose, frequency, route, days, quantity, instructions. Add the diagnosis, notes and start date.
  2. Save: it is a draft. Each line is checked again and shows its warnings.
  3. Edit as needed. Only you can edit your drafts.
  4. Approve. If any line has an avoid warning you must confirm that you have considered the warnings.
  5. The patient is notified and sees the prescription (numbered HRX-…) — without the warning text. A linked Hayat hospital pharmacy can fill it.

Cancel a prescription with a reason; the patient sees it as cancelled.