We check ourselves. Our wait forecasts land within 20 minutes of what the hospital actually posted next 46% of the time — about 18% sharper than assuming the wait won't change — measured against 296,903 real hospital updates over the last 8 days. We only publish this figure once there's enough history to stand behind it.
Every few minutes we record what each ER posts. Later we compare what we forecast for a given moment against what the hospital actually posted then — and against a naive “no change” guess. That backtest is exactly what the figure above reports; nothing is hand-tuned, and it covers every tracked ER at once.
Canada has no common definition of an ER wait time. Of the 729 emergency departments we track, 215 publish something live — but Quebec posts stretcher occupancy, New Brunswick posts length of stay, and others post an estimate of the wait to be seen. We keep each in its own units and never blend them into a fake single number. See the national picture →
We read first-party hospital feeds directly and stamp every reading with the time the hospital said it was current. If a feed goes stale or freezes on one value, we flag it instead of showing an old number as if it were live — and each hospital page carries a “How reliable is this number?” note for its own feed.