Picking the hour on the day itself: the quiet window held on 88.6 % of days
One objection comes up every time hourly trends are shown: nobody plans an ER visit days ahead, you decide on the day. The objection holds. It leaves a question open anyway: once the day is decided, does the hour change anything. I ran the calculation over the archive one evening in September, and this is what came out.
The test
Window: 163 consecutive hours ending 10 September 2026, across the 120 ERs in the file. 108 of them have at least 14 readings a day between 07:00 and 23:00 on at least 5 days; the other 12 are dropped for lack of data, not for their result. The column measured is the ministry’s own: people waiting to see a physician.
Each day is tested using the other days only. The hourly shape is built from the other days, the quietest 3-hour window between 07:00 and 23:00 is read off it, and that window is then scored on the held-out day against that same day’s 07:00–23:00 average. The window is fixed before anyone looks at the day it gets scored on.
The result
- 647 ER-days tested. The window chosen without seeing the day came in under that day’s average 574 times, or 88.6 %.
- Average margin: 4.8 fewer people waiting than the day’s average.
- 93 of 108 ERs are at 70 % or better. Three are under 50 %: Radisson (17 %), Centre de santé Isle-Dieu (33 %), Centre hospitalier de St. Mary (33 %).
- Median gap between an ER’s calmest and busiest window: 8.9 people.
Where the window lands
Across 108 ERs the calm window is 07:00–10:00 in 75 cases, 08:00–11:00 in 4, 09:00–12:00 in 11, 10:00–13:00 in 3. Ten ERs run the other way and are calmest from 20:00 to 23:00. The busiest window is 20:00–23:00 in 20 cases, 15:00–18:00 in 20, and 14:00–17:00 in 19.
The eight largest gaps
| Emergency room | Calm window | Busy window | Gap (people waiting) |
|---|---|---|---|
| Hôpital général juif | 7 h – 10 h (17.0) | 15 h – 18 h (46.4) | 29.4 |
| Hôpital de Lanaudière (Joliette) | 7 h – 10 h (20.6) | 20 h – 23 h (48.1) | 27.5 |
| Hôpital Charles-Le Moyne | 7 h – 10 h (34.1) | 20 h – 23 h (60.9) | 26.8 |
| CMSSS Chauveau | 7 h – 10 h (3.0) | 16 h – 19 h (27.4) | 24.4 |
| Hôpital de la Cité-de-la-Santé (Laval) | 9 h – 12 h (29.0) | 20 h – 23 h (52.7) | 23.7 |
| Hôpital Pierre-Le Gardeur | 9 h – 12 h (24.1) | 20 h – 23 h (46.1) | 22.0 |
| Hôpital Honoré-Mercier (Saint-Hyacinthe) | 7 h – 10 h (19.2) | 20 h – 23 h (39.4) | 20.2 |
| Hôpital général du Lakeshore | 7 h – 10 h (7.0) | 17 h – 20 h (26.4) | 19.4 |
What the measurement does not say
The result is largely “go early”. What took 163 hours of archive is the rest: which ERs run the other way, and by how many people.
For anyone who cannot go in the morning, the evening fallback is weak: the best window after 15:00 sits a median of only 2.3 people below the day’s worst. The window is also under seven days, so day of week is under-sampled and is not tested here. Triage comes before arrival time, whatever the window.
Source and reuse
Readings come from the hourly index published by Santé Québec, kept hour by hour because the public page is overwritten on every update. Each facility’s history is open: /api/facility/<permis>/history?hours=250 (the archive returns 163 for now; the parameter caps higher). Journalists who want to redo the calculation can ask for the raw hourly file; it is sent as CSV.
Measured 10 September 2026 over 163 hours and 108 emergency rooms.