Fig. 1.

Sampling Schematic for Case-Crossover Design. Panel A depicts the random L:M CCWR (Case:Control Window Ratio) sampling design used in our primary analysis, wherein we randomly sampled L = 1 case window and M = 4 control windows (by randomly choosing a window start date), each with S = 90 days in length. A patient’s index test date is denoted by the red line. The “Acute and Short COVID-19 period” is from −14 days to + 28 days, the “post-COVID-19 period” is from + 28 days to + 1 year, and the “pre-COVID-19 period” is from −14 days to −2 years from the index test date. In this instance, one 90-day case window is randomly selected from the “post-COVID-19 period,” and four 90-day control windows are selected from the “pre-COVID-19 period.” Panel B depicts the fixed scheme where two windows of S = 90 days length are selected from each of the periods with fixed start dates. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)