Looking for hidden deaths in Japan's population and flu stats

Over the weekend, I took 2018-12, 2019-01 and 2019-02 death stats and compared them to 2019-12, 2020-01, 2020-02. The data comes from the JP govt [1] The results (and a copy of the source data) are are in a Google Sheet [2].

Basically the number of deaths is just too high and changes too much from year to year to see corona in there. Japan has 250k deaths in Jan and 120k in Dec and Feb. Even if they are under-reporting by a factor of 10, it would still be lost in the noise. There's no smoking gun in there and I think the excess deaths concept really only becomes useful when you have emergency medicine collapse and people start dying from all kinds of usually-treatable things. Maybe when you have good categorization and you can remove all the car crashes and cancers etc. Then you might start to see corona effects clearly but we don't have that.

The flu stats [3] were also updated today and they added 4 weeks of data instead of just one and that does look like a smoking gun to me.

To me that looks like about 100-150 deaths counted as flu above what would be expected. That's for Tokyo. Data is missing for many of the other cities. Most that have data show a drop, in line with expectations as due to everyone wearing masks and washing their hands. Why would Tokyo be different? Well it's tempting to go with a Tokyo-Olympic related conspiracy :) but maybe Tokyo has had a bigger corona problem for longer than anyone really knew. It's the biggest destination for foreigners, it's super easy to spread and contact tracing in Tokyo is a nightmare, so spread could easily happen under the radar, given the low testing we had. So the less touristy cities would be less infected and would see flu drop as it should while Tokyo saw flu replaces by corona. Also Nagoya seems to drop and then jump up, consistent with it briefly being the worst infected area. I should look at the per-prefecture graphs of corona to see if Nagoya's corona increase lines up.

So nationally it's still 100-150 tops. Maybe corona deaths were under-counted and should have been 50% higher but it's not the 5-10x smoking Howitzer that it might have been.

1 https://www.mhlw.go.jp/toukei/list/81-1a.html#link05

2 https://docs.google.com/spreadsheets/d/1Lz71ov-DrSiLiwyUOvROzJtTmCBaHdpTXzpHj8juKAE/edit

3 https://www.niid.go.jp/niid/ja/flu-m/2112-idsc/jinsoku/1852-flu-jinsoku-7.html

Prof. Hitoshi Oshitani's lecture

This [1] (dammit, you need a username and password unless you come from a Google search, so click here [3] and then click the first result) is a long but detailed and interesting transcription of a lecture from Prof Hitoshi Oshitani, one of the expert group, a few days ago. Google translates it into fairly natural, easy to read English. In general it all seems fairly good.
There are some very surprising bits in there like the fact that they think that 80% of cases do not transmit and that 10% transmit to 1 person. This is not something I've heard anywhere (inside or outside Japan) and it implies that the remaining 10% must be transmitting a hell of lot to get us up to epidemic levels. It also seems like it would be very hard to reliably establish this when you're not testing very much.
It's unclear to me whether this 80/10/10 split is actually something that's driving testing policy. Its not that he denies that asymptomatic transmission is a problem but there's a lot of talk about what influences how infectious someone else. Given how little we know about this, I'd rather we just over-tested instead of trying to be smart.
Finally, he ends with an oblique reference to The Hammer and the Dance with a chart pulled from Thomas Pueyo's post [2]. This describes what I would love to see, a short, real lockdown to get to manageable levels (the hammer), followed by aggressive tracing, hygiene and other measures (the dance) were we try to find the best way to keep infections low and controlled. Still no concerts etc but we have a mostly normal life until a vaccine arrives. That puts us where Taiwan, South Korea etc are.
It gives me some hope that he refers to this but from everything else I've read, actual the policies have resulted in an inflatable hammer. We have already failed once at the the dance. He is calling for increased testing but others are still saying no that. Without that the dance will fall apart again.

1 https://www.m3.com/open/iryoIshin/article/759517/

2 https://medium.com/@tomaspueyo/coronavirus-the-hammer-and-the-dance-be9337092b56
3 https://www.google.com/search?q=%E3%80%8C3%E3%81%A4%E3%81%AE%E5%AF%86%E3%80%8D%E3%80%81%E8%AA%95%E7%94%9F%E3%81%AE%E8%83%8C%E6%99%AF%E3%81%A8%E3%81%AF%EF%BC%9F%20-%20%E6%8A%BC%E8%B0%B7%E3%83%BB%E6%9D%B1%E5%8C%97%E5%A4%A7%E6%95%99%E6%8E%88&oq=%E3%80%8C3%E3%81%A4%E3%81%AE%E5%AF%86%E3%80%8D%E3%80%81%E8%AA%95%E7%94%9F%E3%81%AE%E8%83%8C%E6%99%AF%E3%81%A8%E3%81%AF%EF%BC%9F%20-%20%E6%8A%BC%E8%B0%B7%E3%83%BB%E6%9D%B1%E5%8C%97%E5%A4%A7%E6%95%99%E6%8E%88&fbclid=IwAR0iDgCweB8WRA9o8GQIWg3E-W5ZaW5bJgHD5twPch3zQPmdgKON0j7TVEU

Fixing a truncated movie file with untrunc

Just putting this here because it worked

https://github.com/ponchio/untrunc

It was a bit of a mess. Ended up building a docker image which is such a sledge hammer to crack a but whatever. The files was from a phone that had run out of space during recording. The error when trying to examine the file with ffprobe was "moov atom not found".