More about Steve Kirsch's Medicare vaccination data from Connecticut

When gold rules, the world falls silent.

Recently, Steve Kirsch published an article on Medicare data from Connecticut that he fed to ChatGPT - Jesus.

He kindly published the data, a 66MB file, for which I am very grateful. ChatGPT’s results revealed significant differences in the risk of death depending on the brands Pfizer and Moderna. Steve's conclusion is that the vaccines are unsafe if they lead to different increased risks of death. We assume here that the substances have no benefit in the sense of their intended purpose.

I took a closer look at this data, which Steve calls "Gold standard". During the usual plausibility checks, I noticed:

1. In 22 cases, people died but were later vaccinated. One case was a 116-year-old who died in 1983 and received a Pfizer shot in February 2021. For further amusement, there is also a 117-year-old who died in 1991 and even received two Moderna shots in 2021. I don't appreciate desecration of corpses in general, but disturbing the peace of the dead in databases is outrageous. Because... it's just not Gold standard.

2. In two cases, the brand "MODERNA" appears, although Steve writes that the brand names have been obscured, not to stir up suspicion at ChatGPT. I can't say whether ChatGPT jumped on this.

Technical errors are sympathetic and forgivable. With this attitude, I considered the data worthy of further analysis, after elimination of the few logical error cases. It includes around 481,000 insured people who had taken a C19 vaccination. The data includes gender, date of doses 1-3, brand, age when receiving the first dose and date of death, if died between 2020-12-23 and 2024-07-29. Perfect.

Survival time

As a first probe, I calculated the survival time of those who died and presented it in histograms (Fig. 1).

Fig. 1: Histogram of the survival time in days (if died) since 1st shot by brand. PURPLE=Pfizer, MIXED=Moderna, CHERRY=Janssen.

We first would assume all brands not to have been pushed into people who were close to death or seriously ill. This well-known effect is called Healthy User Bias (HUB). It means that the observed risk of death is reduced during the first days or weeks after a vaccination. We should expect to observe the HUB equally across all brands, but surprisingly they show different histograms of the survival time. A bias suggestive for HUB is actually visible with Moderna=MIXED. The other two diagrams deviate from this pattern. A likely reason is that the brands superpose different additional mortality risks. They can partially compensate for the HUB (Pfizer=PURPLE) or even reverse it (Janssen=CHERRY). We should also be aware of contrary confounders, i. e. possible inaccurate reporting as well as vaccination prioritization in care homes and generally to the elderly and those at risk.

Age dependent annual risk

Another approach is to calculate the annual risk of death separately for each age. I therefore chose the period of 365 days after the first shot (Fig. 2). It’s perhaps useful to compare them with the official mortality risks, published in life tables.

Fig. 2: Risk of death for all vaccinated (left) and the various brands within 365 days after 1st shot. Green: Gompertz–Makeham-Curve of Mortality obtained from the life table 2019 of Connecticut. Red: Age 65. Vertical axis log-scaled.

These charts tell everything there is to know. First, the nadir from age 64 to age 65 is due to the composition of Medicare's population. Under the age of 65, people with disabilities are enrolled there. So their risk of death is higher than in the general population of Connecticut. This has nothing to do with the vaccinations. However, it cannot be said that the vaccinations did not cause harm to these people. Besides, the spread increases among very young people because the number of cases is lower among them than among older people.

Second, the people 65+. In terms of brands, Moderna looks better than Pfizer, just as Steve Kirsch notes, and Janssen seems to be produce pretty bad mortality risks. A slight benefit is only visible for the Moderna cohort, probably due to HUB.

In order to expand the previous observations, I conducted a second risk analysis for those who died one to two years after the first dose (Fig. 3).

Fig. 3: Risk from day 366 to 730 after 1st dose. Explanation see Fig. 2.

The main pattern appears to be similar but not equal. Regarding the individuals age 65+, MIXED=Moderna is now straight on the green line, just as to be expected, if we consider a temporary effect such as HUB. PURPLE=Pfizer remains slightly above and CHERRY=Janssen again clearly above the green line. The latter is suspicious. Is it possible that Janssen was given to cohorts, which are not comparable to the others? Finally, it should be noted that the risks for the very old vaccinated people exceeded the green curve in both left diagrams, although usually some years later, the curves decrease, due to the continuous improvement of life expectancy.

Summary

Steve Kirsch’s main hypothesis of different risk outcomes of the various brands can be confirmed on various other ways. If a benefit can be seen, it is most likely with Moderna, but only in the first year and maybe as a result of the HUB. In addition to the contradictions listed at the beginning, the data raise another question. What about mixed vaccinations? They are missing.