Comments on my new KCOR algorithm?
I'm working on a novel analysis method called KCOR for analyzing retrospective observational data in heterogeneous cohort. I'd love to get your thoughts.
Executive summary
I invented a new method, Kirsch Cumulative Outcomes Ratio (KCOR) for analyzing vaccine data. I needed to do that because existing methods simply cannot provide an accurate assessment of risk/benefit when you have retrospective observational data involving vaccinated and unvaccinated cohorts.
KCOR paper and resources
KCOR methods paper (see Box 1 for the one-pager explanation of the method)
KCOR substack article: less technical than the paper
Code at Github:skirsch/KCOR that reproduces the method
How KCOR works
KCOR uses record level data. Only dates of birth, vaccination, and death are need.
The four key steps are:
Define fixed cohorts based on each person’s vaccine status on a specific date (the “enrollment date”)
Characterize and remove the gamma frailty of each cohort which creates and adjusted cumulative hazard function. This is the key step that people have missed.
Take the ratio of the adjusted cumulative hazards
Normalize that ratio relative to the ratio during the baseline period (“quiet period” with no COVID shortly after enrollment)
See “Box 1” in the paper for a more detailed description.
Let me know what you think
Let me know what you think of the KCOR methods paper in the comments below on the approach.
Thanks!



Hi Steve, have you tried performing a validation through ChatGPT?
So grateful for all of your work, Steve. Have you ever connected with RFK?