Why is Taiwan's peck lethality changing so much?

On October 6, 2021, exactly as I had predicted in August, the number of peck-attributed deaths in Taiwan overtook the number of birdemic-attributed deaths. A month later, I predicted that by Chinese New Year 2022, the pecks would have killed twice as many Taiwanese people as the birdemic.

That second prediction didn't pan out. Instead, we're seeing a "second wave" of birdemic-attributed deaths, while peck-attributed deaths have slowed to a trickle, so now the two a pretty much neck-and-neck. As of yesterday, the birdemic-attributed death toll was 1,436. Peck-attributed deaths are no longer reported anywhere near as regularly as they used to be, but on May 17 the total peck-attributed death toll was 1,477.

Curious as to what had changed, I looked up the relevant statistics and plotted the number of reported deaths per million doses for each of the pecks for two different periods: up to January 17, and from January 18 to May 17. It turns out peck lethality has changed a lot. Overall, peck lethality is about half what it used to be -- 14 deaths per million doses, down from 30.

The first thing to notice is that AZ pecks -- the deadliest of the lot, 28% of all doses but responsible for 57% of peck-attributed deaths to date -- have been quietly discontinued. (There were 28 new AZ deaths from January 18 to May 17, but no way to calculate deaths per million doses without dividing by zero.) I didn't know this until I looked up the data to make this chart. It was never announced. I guess they didn't want to admit, after administering 16 million doses to the trusting citizenry and killing more than 800 of them, that they'd made a mistake. They did discontinue it, though, suggesting they've developed some capacity, however limited, for learning from experience.

Eliminating AZ from the mix is only part of the story, though. Every one of the other peck brands has seen a dramatic change in lethality. Mod is killing half as many people as it used to, and Med just a third. What's going on? I can think of a few possibilities.

1. The first dose is the riskiest. If you survived the first dose, you're likely to survive additional doses of the same thing. By January 17, 81% of the population had already received a first dose, so most of the doses since that time have been second or additional doses.

2. The number of "cases" (positive test results) is much higher than it used to be. Before January 17, about 18,000 people had tested positive. From January 18 to May 17, there were 896,000 positive tests. This is partly because more people are being tested and partly because the definition of "case" has changed. Before, a positive antigen test had to be confirmed by a positive PCR test before it could be considered a "case"; starting this month, though, all positive DIY antigen tests are counted as "cases." People who die of peck side effects after testing positive for the birdemic are presumably counted as birdemic deaths, not peck deaths.

3. The pecks themselves could have changed. I mean, why wouldn't they? Quietly replacing most of the doses with normal saline would seem to be the best way of minimizing deaths without admitting anything.

Then there's the question of the Pfi pecks, which show the opposite trend. While the other brands have dramatically decreased in lethality, Pfi is now more than twice as deadly as it used to be. No idea what to make of that.