3 ms·

Some of the numbers in the tables seem strange to me. Looking at the participants, it is hard to understand how the mean and sd of the ages and age distribution

Some of the numbers in the tables seem strange to me. Looking at the participants, it is hard to understand how the mean and sd of the ages and age distributions can be exactly the same. Moreover, apparently the odds of being hospitalized are 160-times greater for vaccinated vs unvaccinated second infection individuals. Perhaps the dramatic differences are due to very low numbers - I did not see the raw numbers for symptomatic infection and hospitalization. But if I were reviewing this paper, I would wonder how they got their age distributions to match so closely.

jsnell 5y ago
It's a matched cohort study. They're starting with a large population in both of their target groups, and then picking pairs of demographically identical people (e.g. age, sex, location), one from each group, for the study. So you'd certainly expect the age distributions to match up exactly. It'd be more worrying if criteria they weren't doing the matching by also were exactly identical. That's not the case, all kinds of secondary characteristics such as comorbidities are roughly but not exactly the same.
- fastaguy88 5y ago
  Thanks. One wonders how closely they matched time of infection and local infection rate.
  - jsnell 5y ago
    They couldn't have matched the time of infection, since not everyone in the study was (re-)infected during the study period. The point of the study was to see if there was a difference in the infection rates, so just picking from the people who got infected would have completely distorted the results. But they would have arranged the study such that both people in the matched pair were considered for exactly the same time period. E.g. if the vaccinated person in pair was fully vaccinated on April 20th, they would have started counting re-infections for the previously infected person of the pair from April 20th too. Local infection rates should be the same, since the pairs would be matched from the same geographic area and the study contstruction ensured they were considered for exactly the same time period.

jsnell 5y ago

It's a matched cohort study. They're starting with a large population in both of their target groups, and then picking pairs of demographically identical people (e.g. age, sex, location), one from each group, for the study. So you'd certainly expect the age distributions to match up exactly. It'd be more worrying if criteria they weren't doing the matching by also were exactly identical. That's not the case, all kinds of secondary characteristics such as comorbidities are roughly but not exactly the same.

fastaguy88 5y ago
Thanks. One wonders how closely they matched time of infection and local infection rate.
- jsnell 5y ago
  They couldn't have matched the time of infection, since not everyone in the study was (re-)infected during the study period. The point of the study was to see if there was a difference in the infection rates, so just picking from the people who got infected would have completely distorted the results. But they would have arranged the study such that both people in the matched pair were considered for exactly the same time period. E.g. if the vaccinated person in pair was fully vaccinated on April 20th, they would have started counting re-infections for the previously infected person of the pair from April 20th too. Local infection rates should be the same, since the pairs would be matched from the same geographic area and the study contstruction ensured they were considered for exactly the same time period.