Add new value: hospital/ICU admissions
Opened this issue · 13 comments
Many provinces replaced active hospitalizations/ICU numbers with hospital/ICU admissions (usually weekly) (#82). These should be added, potentially as hospitalizations_admissions and icu_admissions (original idea: hospitalizations_new and icu_new, but this may be confusing). As well, hospitalizations and icu should probably be changed to hospitalizations_active and icu_active (or perhaps "current" or "occupancy"; "hospitalizations" should be shortened to "hosp"), but maybe it's too late to change these values.
Will also need to add these values to the README and values.json, although they should be labelled experimental.
- NS admissions time series (
ae74c46f-dec7-43d6-a42b-86c7242f0868) - NL admissions archive_ts daily/weekly values (
34f45670-34ed-415c-86a6-e14d77fcf6db) -
NB admissions archive_ts cumulative values (40d93d2b-5c85-4592-b844-fb78b5ae5cec) ~~ (Has "total hospitalized and "total discharged" values but they are much larger than the cumulative values given in the subsequent reports---probably a definition change, and so it doesn't make sense to report negative cumulative values, but could add as an extra dataset) - MB report
- Censor first value of MB report
- NB report
- NS report
- NT dashboard
- SK report
- PE (multiple sources)
- Canada-level time series and province-level yearly estimates from CIHI (excludes Quebec from April 2022 onward)
- Add value to README
Note, for the hospital admission time series for YT, there are two datasets: the CSV (d720af0b-070a-4f67-ba5e-b1393757ba79) and the JSON (2fb9f970-04e0-4502-ae4a-67ec866dcad9). The CSV ends slightly earlier and contradicts the JSON slightly. However, it also places the date of the first hospitalization one day earlier, so I have added this correction but otherwise preferred the JSON dataset.
A time series of admissions for NS can be extracted from ae74c46f-dec7-43d6-a42b-86c7242f0868.
An active_cumul strategy can be adopted for NL from 34f45670-34ed-415c-86a6-e14d77fcf6db.
A potential issue with admission data is that the full cumulative number may not be available, or only available beginning a certain date. Data notes will have to be included, at minimum. Will also have to edit datasets so that they have the correct daily values in the first row rather than the full cumulative value.
I added an old daily HR-level hosp_admissions dataset for ON (raw_data/static/on/hosp_admissions_hr_ts.csv) for posterity, but it should not be used over the current weekly dataset because the most recent weeks of data in the old time series are incomplete.
PEI starts tallying hosp/ICU admissions on their webpage on 2021-12-31 (68e5cbb9-0dcc-4a4f-ade0-58a0b06b1455) and ends with the final update of the original page on 2023-09-26. Data start to be split into years beginning with 2022-11-15.
It should be possible to censor the value_daily for when cumulative time series of admissions do not begin from 0 (such as MB's).
PE hosp/ICU admissions are reported on the old COVID-19 data page (68e5cbb9-0dcc-4a4f-ade0-58a0b06b1455) from 2021-12-31 to the page's retirement on 2023-09-26. On 2022-11-15, the page starts reporting hosp/ICU admissions by year, meaning the cumulative values for 2022 must be added to the cumulative values for 2021 (and later, 2023 + 2022 + 2021).
Hospital admissions are reported on the NT dashboard from 2021-08-25 to the end of the dashboard, 2022-06-13.
ICU admissions are reported on the NT dashboard from 2021-09-08 to the end of the dashboard, 2022-06-13. Prior to 2021-09-08, ICU admissions are simply reported as being "<5".
Decided against including SK hosp admission data from the period of the first weekly report (#64) (i.e., during the period when hospital/ICU occupancy were reported) for several reasons:
- ICU admissions are not reported, only hospital
- Data are presented as an average daily value of the past 7 days, leading to rounding error
- Data change significantly from when they are first reported ("current week") to when the are reported again in the subsequent week ("previous week"), with the following note: "Note: Because of the delay in date tested result, it affects the total number of COVID-19 admissions for a particular day. This lag in data impacts mostly the last couple of days from the day the report is updated." (By contrast, the hospital/ICU occupancy numbers do not change between reports)
Instead, I will begin using data starting with the monthly report.
Should suppress first date of value_daily for MB hospital admissions/ICU admissions.
See discussion of CCCS data for the Territories (#82 (comment)).
The graphic in this news release suggests that hospital admissions probably include ICU admissions:
The metadata could probably be updated with this new information (#37).
There is an issue with PE hosp/ICU admissions around 2023-02-23 or 2023-02-24, where the cumulative value declines for a single day.