CQC Care Service Change & Risk Signals avatar

CQC Care Service Change & Risk Signals

Pricing

from $20.00 / 1,000 cqc care service change signals

Go to Apify Store
CQC Care Service Change & Risk Signals

CQC Care Service Change & Risk Signals

Detect decision-useful changes in official CQC care service data.

Pricing

from $20.00 / 1,000 cqc care service change signals

Rating

0.0

(0)

Developer

Starshape Tools

Starshape Tools

Maintained by Community

Actor stats

0

Bookmarked

2

Total users

1

Monthly active users

2 days ago

Last modified

Categories

Share

This Actor compares complete official CQC public snapshots and emits one machine-readable row per meaningful change. It is a change-intelligence Actor, not a facility-directory scraper.

Supported signals

NEW_REGISTRATION, DEACTIVATED, RATING_DOWNGRADE, RATING_UPGRADE, and PROVIDER_CHANGE. Identity is the official CQC Location ID; provider changes require the official Provider ID. A missing active row alone is never treated as deactivation: explicit evidence from CQC's archived-locations sheet is required. Archived does not necessarily mean the business closed.

Severity (HIGH, MEDIUM, LOW) is a conservative heuristic classification of an observed change, not a prediction or risk probability.

Official source

The Actor resolves the current files from CQC's Using CQC data page: the weekly Care Directory CSV plus the monthly ratings and deactivated-location ODS files. CQC warns that its digital-system migration can delay directory updates, including registration cancellations. This Actor therefore does not claim real-time freshness.

First run and state

The first successful live run creates a baseline and emits no fake NEW_REGISTRATION signals. baselineOnly: true creates a missing baseline but never resets an existing one. State uses the named KVS cqc-care-service-change-risk-signals-state, a small scope-specific CURRENT pointer, complete generation metadata, and deterministic chunks of at most 2,000 compact entities. localAuthorities is source scope and changes the scope fingerprint. Signal, service, rating, postcode, and inactive filters are output-only and cannot change entity identity.

Detected signals enter a durable queue before CURRENT is committed. Dataset delivery advances its cursor only after a successful write. Large batches are split across runs by maxItems; while a backlog exists, it is drained before fetching another source version. Consumers can additionally deduplicate with deterministic signalId.

Try it with sample data

{"mode":"sample"}

Sample mode is a user-facing deterministic demonstration: it makes zero external requests, reads/writes no production state, uses the production diff engine, and emits rating downgrade, provider change, and explicit deactivation examples.

Example live input

{"mode":"live","localAuthorities":["Leeds"],"baselineOnly":true,"maxItems":100}

Schedule live mode weekly or monthly. A later run with the same complete source state and no pending backlog emits zero rows.

Example output

{"signalId":"…","signalType":"RATING_DOWNGRADE","severity":"HIGH","signalSummary":"Rating changed from Good to Requires improvement","detectedAt":"2026-01-02T00:00:00.000Z","sourceObservedAt":"2026-01-02T00:00:00.000Z","entityType":"location","entityId":"L1","name":"Oak Care","address":null,"postcode":null,"localAuthority":null,"serviceType":null,"previousValue":"Good","currentValue":"Requires improvement","previousRating":"Good","currentRating":"Requires improvement","reportDate":null,"providerId":"P1","providerName":"Alpha","sourceUrl":"https://www.cqc.org.uk/location/L1"}

Limitations and compute

The public sheets do not expose a sufficiently stable report identifier/publication semantic for NEW_INSPECTION_REPORT, so it is not emitted. Manager, bed-capacity, service-type, and specialism signals are also excluded from this MVP. Whole-England runs download and parse large public files and use more memory/compute; authority scope filters storage after the official country file is fully validated. Required-source or schema failures fail closed and preserve the last committed baseline.