ABSTRACT
This chapter describes some techniques for checking whether the assumptions of the self-controlled case series method are reasonable, in the sense that departure from them is unlikely to produce grossly inaccurate inferences. Event-dependence of observation periods need only be investigated when the event carries substantial short-term mortality. Long-term event-dependence of exposures requires attention only when there is reason a priori to suspect that it might be an issue. Like all statistical methods, the self-controlled case series (SCCS) method relies on assumptions. For the standard SCCS model, modelling assumptions also include the number and placing of age and exposure categories. A rare disease assumption is required to apply the SCCS method to non-recurrent events. The SCCS likelihood for recurrent events is derived under the assumption that the events arise in Poisson processes modulated within individuals by age and time-varying exposures.
