What the numbers mean.
Hospital waiting times are not a single standard measure. Our job is to make each published figure understandable and traceable.
1. Original sources
We read public pages or endpoints published by hospitals and health authorities. A connector reads one source for all the departments it covers. We link every figure to its original publisher. We do not collect patient records or estimate missing values.
2. Different measures stay separate
- Average total stay: time spent in the department, including assessment, investigations and treatment. This is not the wait to see a doctor.
- Longest current wait: the longest wait currently reported for a specified assessment or treatment stage.
- Average wait to triage: the published average until the initial triage assessment.
- Median time already waited: the middle elapsed waiting time among people currently waiting. Northern Ireland publishes this measure; it is not a forecast for a new arrival.
- Bands and lower bounds: ranges such as “over four hours” are kept as ranges, without inventing a midpoint.
3. Publication time and retrieval time
“Published by source” is the source’s own timestamp, where supplied. “Retrieved by us” records when our system read the source. We do not substitute our clock for a missing publication time.
A recent source update means the publisher timestamp is within the last hour and our retrieval is within the last 30 minutes. An older observation remains explicitly labelled. If the source has no reliable timestamp, we say so, even after a successful retrieval.
4. Historical statistics when live data is missing
Where no recent waiting-time observation is available, we show an official historical statistic where one can be matched reliably. The card states “Historical · Not live”, the reporting month and whether it describes one hospital or the whole NHS trust. Historical statistics do not predict the queue today.
A mean is the arithmetic average. A median is the middle recorded time: half the recorded visits were shorter and half were longer. We name them separately. The English historical initial-assessment indicator covers ambulance arrivals only; it is not an all-patient triage wait. Monthly English ECDS figures are provisional and may be revised. Their release date is separate from the month of attendance. Total stay includes assessment, investigations and treatment; it is not just a wait to see a doctor. A trust-wide figure covers the provider’s reporting departments and patient groups, and must not be read as a separate measurement for each hospital or age group.
When only a percentage of visits completed within four hours is published, we label it as a percentage. We never turn those percentages into an average time. Official statistics can be several months old: the reporting period remains visible, separate from when we retrieved the file. Machine-readable statistical sources are checked daily; sources that fail keep their previous figures and original periods. Some official tables are supplied as verified extracts from a dated publication, with their original verification date and a link to that release.
5. Missing, closed and zero are different
A missing figure is shown as unavailable. A department is marked closed only when the source reports it. A genuine published zero remains zero. None of these states is converted into another.
6. Coverage and service types
The directory includes A&E departments, urgent treatment centres, minor injury units and some specialist services. A UTC at a hospital may publish a figure even when that hospital’s A&E does not. We keep those department identities separate.
Directory coverage does not mean every service has a live feed, that its opening hours have been confirmed today, or that it can treat a particular condition. Check restrictions directly with the service.
7. History and corrections
Changed observations can be stored for up to 30 days to support quality checks. On selected hospital pages we plot recorded observations from the original source over a 24-hour or seven-day window. The actual first and latest publication times are shown: selecting seven days does not mean seven days of data have already been collected. Each dot uses the original publication time and an exact published duration. Repeated checks with the same publication time appear once; if a value is revised for that time, the latest retrieved version is used. Missing publication times, ranges and readings that were already stale when collected are excluded from these plots. Blank periods are not zeros. We do not interpolate, join dots across gaps or generate forecasts. Tables provide the latest source publications and their separate retrieval times. If a source changes format, collection may fail; the previous observation keeps its original timestamp and becomes older.
Report a data issue with the hospital name and original source, without including personal medical information.