Survivor average causal effects and long-term outcomes after severe infections

causal inference; survivor average causal effect; truncation by death; principal stratification; potential outcomes.Severe infections in children can have consequences that last for years and can persist into adulthood. 1,2For instance, in a meta-analysis on bacterial meningitis, Edmond and colleagues 3 estimated that the median risk of a major or minor sequela was 19.9%.Given the potential clinical impact of these conditions, it is unsurprising that many epidemiologic studies 4-9 compare outcomes in children who develop severe infection versus outcomes for those with no history of severe infection.A complicating issue, however, is that not all children diagnosed with severe infection survive to outcome assessment.For example, fatality rates of infections caused by Group B Streptococcus have been estimated 10 to range from 4.7% to 18.9%, depending on the geographical region.Evidence of postacute mortality has also been reported: Bedford and colleagues 11 observed that some participants who survived bacterial meningitis died within 5 years (32 of 1717 children); in another study 12 that recruited children aged 2-59 months with pneumonia, sepsis, or meningitis, 2.8% of study participants died after discharge.In this research letter, we discuss survivor average causal effects in the context of studies on the long-term impact of infections.In the following sections, we formally define the survivor average causal effect and discuss assumptions under which it coincides with, or is bounded by, the crude comparison (that is, the comparison of outcomes in actual survivors by exposure status).