Could be reverse causation. T2 diabetes is a major risk factor for COVID, so perhaps kids with undiagnosed T2 diabetes are more likely to have COVID bad enough to go to the doctor.
chiefalchemist 477 days ago [-]
Probably not a popular take:
- But they didn't account for the treatment for Covid as being a possible cause for the difference.
- Nor - conveniently? - is there mention of vaxinated vs non-vaxinated. Boosted vs non.
- The final check would be those who knew they had Covid vs those who were asymptomatic. Stress is a powerful force. Those aware of having Covid may have stressed more, which is known to effect the immune system, weight gain, etc.
defrost 477 days ago [-]
It's an ANOVA study of a cohort of 613,602 pediatric patients aged 10 to 19 years between 2020 to 2022 based on electronic health records that include factors such as vaccination status.
It reports the correlations found - those that had a COVID diagnosis (with no prior diabetes indicated) had a greater chance of developing diabetes.
It doesn't report that those vaccinated (with no prior diabetes indicated) had a greater chance of developing diabetes NOR does it report a great many other potential relationships.
The probable reason for that lack of reporting is the absence of pattern being present in the data ... it's reasonable to assume that multiple correlations across fields in the EHRs were run (including against vaccination status).
If you suspect a vaccination effect you can always look to Australian health studies - the country has decades of world class near total population records for the entire country and had a near total COVID vaccination program with multiple doses.
Any effects would clearly show up in the before and after M&M stats of 25 million people.
chiefalchemist 477 days ago [-]
> The probable reason for that lack of reporting is the absence of pattern being present in the data
So another "scientific study" with too little actual science? How does this build trust?
defrost 477 days ago [-]
The word you're looking for is epidemiology, it's applied medical statistics, and you build your trust by educating yourself ...
Perhaps start with some first year college texts on ANOVA applications in medicine and agriculture?
chiefalchemist 477 days ago [-]
What I'm suggesting is that "the science" seek some epidemiology.
There's a lot implied in this study. There's some great correlation. But it's not thorough. It missed key questions (already listed) to the point of being misleading.
It's the ethical equivalent of cherry-picking data to find something you wanted to find. That's science? I think not.
defrost 476 days ago [-]
> It missed key questions (already listed) to the point of being misleading.
That was your claim. What is your evidence for this? Did you read the actual paper cited in the article.
> It's the ethical equivalent of cherry-picking data to find something you wanted to find.
They reported what was there.
Are you suggesting they didn't run a multi variate regression against vaccination status?
> That's science? I think not.
Could you remind me when you first ran any regressions on medical data for a cohort of several thousand or more?
In the interestes of transparency I first did that in the mid 1980s .. although admittedly I continued in the direction of exploration geospatial statistical applications as part of geophysics.
jcoletti 477 days ago [-]
Headlines always get me. Correlation identified, but not causation.
mewse 477 days ago [-]
As a practical matter, studies about links between major diseases will always always be talking about correlation.
To reach 'causation' would require intentionally giving your experimental subjects Covid-19 (and in a way that didn't result in them knowing they'd had it!), and that's unlikely to pass muster with the ethics review board.
jcoletti 477 days ago [-]
Hm, I thought other methods could be used to establish causation, such as longitudinal cohorts, RCTs, and (another randomization I'm forgetting), that wouldn't require infecting humans, but maybe I'm confusing these
477 days ago [-]
Rendered at 12:43:05 GMT+0000 (Coordinated Universal Time) with Vercel.
- But they didn't account for the treatment for Covid as being a possible cause for the difference.
- Nor - conveniently? - is there mention of vaxinated vs non-vaxinated. Boosted vs non.
- The final check would be those who knew they had Covid vs those who were asymptomatic. Stress is a powerful force. Those aware of having Covid may have stressed more, which is known to effect the immune system, weight gain, etc.
It reports the correlations found - those that had a COVID diagnosis (with no prior diabetes indicated) had a greater chance of developing diabetes.
It doesn't report that those vaccinated (with no prior diabetes indicated) had a greater chance of developing diabetes NOR does it report a great many other potential relationships.
The probable reason for that lack of reporting is the absence of pattern being present in the data ... it's reasonable to assume that multiple correlations across fields in the EHRs were run (including against vaccination status).
If you suspect a vaccination effect you can always look to Australian health studies - the country has decades of world class near total population records for the entire country and had a near total COVID vaccination program with multiple doses.
Any effects would clearly show up in the before and after M&M stats of 25 million people.
So another "scientific study" with too little actual science? How does this build trust?
* https://www.ncbi.nlm.nih.gov/books/NBK7993
Here's a nice lady: https://en.wikipedia.org/wiki/Fiona_Stanley a noted epidemiologist who has demonstrated the need for folic acid in childhood development.
Perhaps start with some first year college texts on ANOVA applications in medicine and agriculture?
There's a lot implied in this study. There's some great correlation. But it's not thorough. It missed key questions (already listed) to the point of being misleading.
It's the ethical equivalent of cherry-picking data to find something you wanted to find. That's science? I think not.
That was your claim. What is your evidence for this? Did you read the actual paper cited in the article.
> It's the ethical equivalent of cherry-picking data to find something you wanted to find.
They reported what was there.
Are you suggesting they didn't run a multi variate regression against vaccination status?
> That's science? I think not.
Could you remind me when you first ran any regressions on medical data for a cohort of several thousand or more?
In the interestes of transparency I first did that in the mid 1980s .. although admittedly I continued in the direction of exploration geospatial statistical applications as part of geophysics.
To reach 'causation' would require intentionally giving your experimental subjects Covid-19 (and in a way that didn't result in them knowing they'd had it!), and that's unlikely to pass muster with the ethics review board.