When time-dependence in disease outcome risk is not captured by impact evaluation modeling studies: a measles vaccination case study

Abstract

AbstractBackgroundIn modeling studies that evaluate the effects of health programs, the risk of secondary outcomes attributable to infection can vary with underlying disease incidence. Consequently, the impact of interventions on secondary outcomes would not be proportional to incidence reduction. Here we use a case study on measles vaccine program to demonstrate how failure to capture this non-linear relationship can lead to over- or under-estimation.MethodsWe used a published model of measles CFR that depends on incidence and vaccine coverage to illustrate the effects of: (1) assuming higher CFR in “no-vaccination” scenarios; (2) time-varying CFRs over the past; and (3) time-varying CFRs in future projections on measles impact estimation. We evaluated how different assumptions on vaccine coverage, measles incidence, and CFR levels in “no-vaccination” scenarios affect estimation of future deaths averted by measles vaccination.ResultsCompared to constant CFRs, aligning both “vaccination” and “no-vaccination” scenarios with time variant measles CFR estimates led to larger differences in mortality in historical years and lower in future years.ConclusionsTo assess consequences of interventions, impact estimates should consider the effect of “no-intervention” scenario assumptions on model parameters to project estimated impact for alternative scenarios according to intervention strategies and investment decisions.