Tervis ja toetused

Tervisestatistika

Tervise Arengu Instituudi ametlikud tervisenäitajad masinloetavate tabelitena.

Tervise Arengu Instituudi tervisestatistika andmebaas (statistika.tai.ee) on Eesti ametlik tervisestatistika: sünnid, surmad, haigestumine, ravi, tervisekäitumine ja ravimid, tuhandete tabelitena. Iga tabel kannab oma metoodikamärkusi ja osa neist on TAI enda märke järgi kohustuslik välja tuua, sest just seal on kirjas, millal mõiste, nimetaja või katvus muutus ja mis aastad seetõttu enam ei võrdle. Andmed on koondnäitajad, ühegi patsiendi kirjeid siin ei ole, ja tühi lahter tähendab avaldamata või kogumata arvu, mitte nulli.

Mida su agent peab oskama

POSTVajab vormipäringut. Vähemalt üks vajalik päring on vormi saatmine (POST). Küpsiseid ega turvamärgiseid vahele ei jää, aga ainult lehti tõmbav agent seda päringut teha ei saa.

Milline agent selle läbi teeb

Juhend ise

Inglise keeles, sest seda loeb mudel.

Estonia Health Statistics

Tervise Arengu Instituut publishes Estonia's official health statistics — births, deaths, disease, treatment, health behaviour, medicines — as 2,406 PxWeb tables at https://statistika.tai.ee/api/v1/et/Andmebaas. Public, no authentication.

Retrieve

Use the module below. Two calls answer almost everything: find the table, then read it.

const f = await findHealthTable("surmad");   // ONE request, the database's own search
show(f.rows.slice(0, 5));                    // { tableId, title, path, updated }

const r = await healthTable("SR001", { query: { Elulisus: ["1"] } });
show(r.totals);           // { rows, series, first, last, firstValue, lastValue, changePct }
show(r.mandatoryNotes);   // the caveats TAI marks as must-show
show(r.rows.slice(0, 5));

A dimension you do not name comes back in full. Leaving Aasta out of the query returns all 34 years of SR001 in one call. So a ten-year question is one call, not ten — and totals.first/last/changePct are computed here over the whole series rather than by picking two printed rows and subtracting.

Get the exact codes from healthTableMeta(tableId) when a query is rejected: they are Estonian words (Aasta, Elulisus, Maakond, Vanuserühm) and must be sent exactly as returned.

Methodology notes are part of the answer

Each table ships its own note array, and extension.noteMandatory flags the ones TAI requires to be shown. The module surfaces those as mandatoryNotes, keeping the methodology page's URL rather than stripping it. That is where a changed definition, denominator or coverage break is written down — two years either side of such a break are not comparable, and saying so is part of answering.

Limits

Reading a table is a form POST: the selection goes in a JSON request body, and a plain GET of the table URL returns its dimension list only — the codes, never the numbers. Table search is a GET. So we cannot offer this source to an agent that can issue GET requests and nothing else; it can find a table here but not read one.
Public aggregates only; no patient-level records.
A blank cell is suppressed or not collected, never a zero. rows carries only cells that have a value, so a missing year is missing on purpose.
Five table ids exist in two folders each (RV31, PH5, KP15, TTO60, RV301). The module throws naming both paths instead of picking one.
The API rate-limits: 429 - Too many requests in too short timeframe on concurrent requests. Do not fan out.
?query= is the whole search contract — adding PxWeb's documented &filter=* returns [] with HTTP 200 here, which reads exactly like "no such subject".

Verify

class: "dataset", a value count matching the product of size, and the table's source and updated. A search that matches nothing returns an empty list with HTTP 200 — totals.tables: 0 is the module saying that out loud.

Module API

healthTable(tableId, {query, lang}) -> {source, page, tableId, title, producer, updated, totals, rows, mandatoryNotes, notes, note} findHealthTable(term) -> {source, page, term, totals, rows, note}