Dear Department of Internal Affairs,

Could you please provide a data set in Excel or csv file with the following columns - for the last 5 years by week to the end of March 2020 for the whole of New Zealand.

Column 1 - Deaths During Period (Sunday to Saturday - or Monday to Sunday)
Column 2 - Age Group at Death (in 5 year bands)

I would expect that these numbers are available from the Deaths Register. If there are any concerns I may be happy to have numbers by 10 year bands. Please consult early so that these number can be provided on time by the end of the OIA period.

If there is a public portal where these numbers can be queried directly please direct me to this as soon as practical - you do not need to wait until the end of the OIA period.

Yours faithfully,

Chris Johnston

Hi Emma,

Thanks for your prompt reply.

The Stats NZ website is useful but not as up to date as it could be. The monthly data set was last updated on 19 Feb 2020.

Monthly data may be OK, but it is needed up to 31 March 2020.

Where your team can point out any caveats on the data that would be useful - e.g. death certificates may not be typically or legally finalised until a certain number of days after death for example.

I would not want data caveats to delay the delivery of the data set though.

Yours sincerely,

Chris Johnston

Dear Alysha Stevenson,

Thank you for the information forwarded in response to my request.

OBSERVATIONS SO FAR
The data set has thrown up a few things which I have outlined below. i hope that you can help.

1) There is significant evidence of additional deaths in February 2020 and probably March 2020 (although all March deaths do not appear to have been registered when the data set was extracted on 23 April 2020).
- I am getting my analysis of these potential additional deaths reviewed at the moment.
2) The additional deaths appear to be concentrated int he over 60's and over 80's.

SOME PROBLEMS
Some elements of the way the OIA was responded to cause some problems and constraints that I believe we can work around - but it requires a refined data extract.
3) Data was extracted on 23 April 2020 - which has meant that March 2020 was incomplete - this is just life but I note it for completeness (and you did state potential completeness issues in your response).
4) Data was extracted in 10 year age groups - which turned out to be OK but you did not consult/confirm with me what was going to be produced.
5) Monthly data was supplied - which has proven to be at too high a level to see what has occurred with enough granularity.
6) Individual cells were redacted if below 10 deaths. While I understand a policy like this (Stats NZ randomly round cells of 5 of less), I would have refined my request if I had be consulted on this.
7) The data was provided in separate tables in a Word Document and not in the CSV file format requested (which is a standard format for data extracts). This causes additional work for me - and other people that may access this OIA on the FYI.ORG.NZ website in future.
8) The 90-99 group was labelled 90-100 and then another age group was labelled 100+ - which appears to indicate some form of overlap. it is unclear whether this is intended or not - e.g. Does 100+ not include 100?)

REFINED REQUEST
Having looked at the data it seems that the following would meet the need better and overcome the limitations above.

a) Use the following Age Groups: 0-19, 20-59, 60-79, 80+
b) Provide weekly data as originally requested
c) Provide in a CSV dataset
Column 1 - Age Group as defined in (a) above
Column 2 - Start date of the week - let's agree on a Monday to Sunday definition
Column 3 - The number of deaths
d) Keep the suppression of deaths to under 10 - but the bigger aggregations/groupings in (a) should minimise this occurrence.
e) Extract up to the end of March 2020 again - which should look better as we are now at the 10 May 2020 which is 2 weeks further on from 23 April 2020.

OLD + NEW OIAs
Now you may be concerned about how the KPIs look on the OIAs if we re-do this one. My suggestion is that I also lodge a new OIA for the monthly and weekly datasets up to the end of April 2020. You can answer these both at once and while this one will be late you can answer the other one early.

As there are a number of OIAs around COVID19 death rates I am keeping the Ombudsman abreast of these as they are interdependent and impact on the delivery of information in a whole of government approach is useful for making the momentous decisions that we are presented with as a country.

I would like to impress upon you that a timely turnaround is important given the importance of this topic. If there were significant deaths in NZ over summer this is very important to measure and know about so it can be further diagnosed and understood by both the health sector, the rest of government and the private sector.

MEDIUM TERM - KEY INDICATOR DATASETS
My medium term suggestion is that these health statistic related KPIs need to be uploaded daily (overnight) and available on demand on an Open Data portal. This means that they can be downloaded at will and citizens / industry analysts can contribute to the management of government more promptly. An example of Open Data with DIA data on it is below, but it is very static data and not the type of data where timeliness is important:

https://catalogue.data.govt.nz/group/1a6...

I recognise that Statistics NZ provide this service but it is months later and the mode seems to be as one of historical record - rather than as a data set from which monitoring can be undertaken within a shorter time frame (and compared to other operational data sets).

Being able to provide information so that a forecast of likely deaths based on the "arrival rate of death registrations" also seems to be necessary so that timely information can be estimated. This would be helpful for quicker monitoring of future pandemics. Given how granular this raw dataset would be it is unlikely that it could be published raw. Therefore I suggest that DIA provide a data science engine that provides both actual and estimated (with statistical error) in any data portal as core functionality. Statistics NZ might provide the algorithm, or upgrade its own portal (on behalf of the DIA).

Could I ask you to please raise these issues with Ann-Marie Cavanagh (Deputy Govt CDO), and Peter Murray (Deputy CE Info & Knowledge Services) who seem to have relevant responsibilities within DIA (including for all-of-government).

THE GRAPH
Finally - You may be interested in the graph of some of the data you provided.
https://1drv.ms/b/s!ArUmXIClDUEfgaxYwAaH...

Yours sincerely,

Chris Johnston

Dear Alysha Stevenson,

Thanks for your reply.

APOLOGY FOR DELAY IN RESPONSE
The fyi.org.nz portal has been out of action for the last week. The response below did not send as a result of that outage.

ACKNOWLEDGEMENTS
I cannot see a CSV attachment to your email - are you able to check if it was attached?

Thanks for clarifying the age range for 90-99.

Thanks also to the DIA for looking into publishing the deaths more quickly in some form. Before any decision to terminate this OIA I would appreciate the opportunity to review any "solution" to confirm that it would meet my need.

SUMMARY OF COMMUNICATIONS
In response to your release of some of the data I needed on 8 May 2020 I replied on 10 May 2020 to say that the way that this OIA had been executed was useful but not all the information that was required. In particular:
1) I had asked for weekly and indicated monthly might be OK - but this proved to not be the case.
2) The DIA in its response suppressed counts below 10 in any given cell.
- This is a reasonable action
- I was not aware of this proposed action
- So with this information and the need in (1) above, it became clear to me that the age groups would need to be merged if the data set was going to be useful at a weekly level.
- Hence I suggested some consolidated groupings.

So we learn as we travel along the road of this OIA together.

ADDITIONAL AND SEPARATE OIA
Given the passage of time (and that March appeared to be incomplete) I also raised an additional OIA to request a dataset for an extended period of time. The link to that is here:
https://fyi.org.nz/request/12826-number-...

I assume that the response to this OIA and the new OIA are what you refer to when mentioning the "2 emails".

To clarify the status - the new OIA stands independently and on its own merits. It is my definitive version that will meet the need for ongoing monitoring of the NZ death rates.

NOT COMBINING THE OIAs
I outlined that I was open to combining the OIAs if you were concerned about the DIA's OIA KPIs and both OIAs could be completed expeditiously. Given the reply to my 10 May 2020 was 15 May 2020, and the DIA is indicating that this would be a "new request", we are not on on an "expeditious" path. Therefore this offer is revoked.

My case is that this OIA is not complete and it needs to be rectified. The DIA should rectify this on a more accelerated time frame than a normal OIA that has just been lodged.

It is not legally sustainable to say that a similar request has been lodged after the original OIA was supposed to have been delivered, and therefore we will combine those and deliver the material at the date the second later OIA is due. Where OIAs for a similar topic were submitted for monthly updates this would mean the relevant government department would never have to respond as the deadline would simply move ahead each month.

I did indicate to the Ombudsman that I would wait until the DIA had a chance to respond to my offer before officially lodging a complaint - but this appears to be the point at which some guidance would be useful.

THIS OIA
For this OIA the request was:
a) Use the following Age Groups: 0-19*** (see below), 20-59, 60-79, 80+
b) Provide weekly data as originally requested
c) Provide in a CSV dataset
Column 1 - Age Group as defined in (a) above
Column 2 - Start date of the week - let's agree on a Monday to Sunday definition
Column 3 - The number of deaths
d) Keep the suppression of deaths to under 10 - but the bigger aggregations/groupings in (a) should minimise this occurrence.
e) Extract up to the end of March 2020 again - which should look better as we are now at the 10 May 2020 which is 2 weeks further on from 23 April 2020.

Note: *** I suspect that the 0-19 age group will not prove to be useful on a weekly basis. So if it makes it easier you can ignore that group for this OIA. This makes completing this OIA even easier.

This adjustment request should not make it harder as in the data request there are only two "aggregations" to change:
- Date of Death - instead of using the "start of month"/"month" function use the "start of week" function
- The age group classification (consolidation) can be undertaken in the data extract or after the extract. No splitting of existing categories is required.

These simple aggregation changes will enable us to close this OIA down.

Yours sincerely,

Chris Johnston

Dear Alysha Stevenson,

Thanks for the prompt response.

I did refer this OIA to the Ombudsman so there is an opportunity for an I dependent review of reasonableness.

In essence, I asked for weekly and said monthly data “may” be reasonable ... which implies there are conditions. So when the DIA is making decisions it is worth consulting to get it right first time. Not consulting leads to this type of problem.

Not providing a reasonable comms approach to get it right first time (or remediate) leads to the DIA “training” requesters to be less flexible. I think ultimately this decreases the levels of trust. It may be a naive view.

I can confirm that the cab file is attached to the DIA’s previous response. I am at a loss to explain my incompetence in not spotting it.

If DIA provides a marginal/incremental dataset the DIA needs to state that:
- All previous numbers have not changed (eg by additional deaths being registered for a period previously supplied).
- New and incremental datasets fit together.

The Stats NZ dataset is a good first step. I am reminded of my Stats101 course where the lecturer said that a number is all well and good, but you need to know what to compare it to. Therefore previous years of comparable data is important ... and I suspect of marginal cost to the DIA data team to produce. Having worked with and been lectured by senior StatsNZ people I am surprised that their governance did not encourage this when published on their website under their brand.

I look forward to the DIA’s responses.

Kind regards
Chris Johnston

Yours sincerely,

Chris Johnston