Update NS deaths time series

[jeanpaulrsoucy]

Nova Scotia is now reporting "New Deaths" and "New Deaths from Previous Reporting Periods". For now, I am reporting deaths as the sum of these two numbers.

There is also a "Total Deaths" value now---will have to check if this was there before, and use it to calibrate if any deaths are missing from the totals. Can also compare to PHAC data.

[jeanpaulrsoucy]

Looks like NS was reporting 587 deaths as of last week's report, with 14 more today (all from previous reporting periods), for a total of 601. Yet their reported total deaths is 602. Could start using this new cumulative number for reporting.

[jeanpaulrsoucy]

The D_TOTAL entry in the weekly report JSON (COVID_WEEKREPT_V6_PROD) seems to have been there since the beginning of the file. I could use this to rewrite the report time series potentially. Of course, I would have to use historical data because not all historical deaths have unclassified HRs for deaths...

I have a vague recollection about a missing/disappearing NS death around the time they started producing report data...

[jeanpaulrsoucy]

The original weekly report JSON did not report total deaths, but the one adopted with the second weekly report (2022-03-18) did. Our time series gives total deaths on 2022-03-15 (the as-of date for the 2022-03-18 report) as 231. The total_deaths value is 232. So it really is as simple as figuring out where that missing death was.

[jeanpaulrsoucy]

Digging into the email archive, from January 20, 2022:

I wanted to inquire regarding a discrepancy in the number of COVID-19 deaths reported by Nova Scotia.

On January 17, the Nova Scotia dashboard reported 121 total deaths (9 Western / 9 Northern / 10 Eastern / 93 Central).
On January 18, the Nova Scotia press release reported an additional death in Eastern Zone
On January 19, the Nova Scotia press release reported an additional 3 deaths in Central Zone

However, the Nova Scotia dashboard reported 124 deaths on January 19, despite an increase in 4 (1 + 3) deaths since January 17, which would imply a total of 125 deaths in the province. Can you help me understand this discrepancy. If a death was removed, from which Zone and which date?

The response was:

This data is pulled from the Panorama system and one of the deaths was not updated in the system by the time the dashboard was updated. The dashboard should be updated today to reflect the additional death.

However, this is actually a case of there being one extra death, rather than one missing death.

The best solution here is probably to start using cumulative deaths from the second NS report date (the first NS report did not have total deaths). Will have to write a script to extract all of this from the archives. Additionally, in the calculations, the "Not Reported" health region data will have to be subtracted so that the time series are calculated correctly.

[jeanpaulrsoucy]

In this week's data update, the NS cumulative deaths incremented by 15 but the the new deaths (this week) was 1 and new deaths (from previous reporting periods) was 15. Based on this and how the variables are coded in the JSON, I am assuming that deaths (from previous reporting periods) actually includes the death from the current reporting period as well. This would mean deaths_previous_reporting_period and deaths don't need to be tracked separately in the report.

Will wait for another instance of this occurring before finalizing this observation. For now, have coded both of the aforementioned fields in the report as 15.

[jeanpaulrsoucy]

This week, we have 14 deaths reported in the previous period and 1 in the current period, as well as a cumulative difference of 15. This contradicts the previous observation of 1 + 15 = 15. Evidently, this field should be ignored in favour of cumulative deaths.

[jeanpaulrsoucy]

Another mismatch: 1 + 19 = 20 weekly deaths (covering a two-week period due to the holidays), but a cumulative increase of 694 - 675 = 19. Again, ignoring the weekly data in favour of the cumulative.