https://www.nejm.org/doi/10.1056/NEJMc2302462
"We fit the Cox regression model with a time-varying hazard ratio for severe infection and fit the proportional-rates model with a time-varying rate ratio for recurrent infection for each additional booster dose that was received... We estimated the booster effectiveness on a particular day as 1 minus the hazard ratio or rate ratio on that day multiplied by 100%."