The sequence symmetry analysis design (SSA) is a epidemiological design used for finding associations between exposures and outcomes using prescription registries. But what exactly does SSA estimate? How does the choice of exposure window affect the results? And what happens when researchers adjust the crude sequence ratio (SR) with a so-called null-effects SR?
In this talk, I will show theoretical results for the design, including how it estimates a hazard ratio while implicitly adjusting for time-invariant confounders – even in the presence of competing risks. I will also discuss the consequences of adjusting the SR, showing that this adjustment reintroduces the need for a no unmeasured confounding assumption. These results have the following practical implications: Researchers should perform sensitivity analyses to ensure robustness of the results. Furthermore, the adjusted SR should be abandoned.