Extensions of SPM to non-independent events, terminal events, and parallel-trends violations

Investigate extensions of the symmetric pair matching (SPM) design for settings involving non-independent events, terminal events, or violations of the parallel trends assumption.

Background

The paper introduces the symmetric pair matching (SPM) design as a self-controlled method for estimating transient exposure effects while controlling by design for time-invariant confounding and temporal effects. Its stated assumptions include recurrent events, independence of events from past events, and a common outcome time trend across individuals, corresponding to a parallel trends assumption.

The authors note that existing extensions of the self-controlled case series design may be preferable when some assumptions are violated, but explicitly identify the development of analogous SPM extensions as future work. The unresolved scope includes non-independent recurrent events, terminal events that preclude subsequent events, and departures from the parallel trends assumption.