Poisson based model for adjusting for non-compliance and contamination in cohort studies
(English)Manuscript (preprint) (Other academic)
In cohort studies evaluating the effectiveness of an exposure such as an intervention or treatment on an outcome, self-selection in the form of non-compliance and contamination may lead to biased estimates of effectiveness. A previously published adjustment method by Cuzick et al for randomised controlled trials is further developed to a log-linear Poisson model for adjustment in cohort studies, allowing for estimation and testing also in subgroups with varying effectiveness. An example with both non-linear and log-linear interaction models for a Swedish mammography screening material with subgroups based on parity is presented.
cohort, poisson, non-compliance, contamination, subgroups
Cancer and Oncology
Research subject Oncology
IdentifiersURN: urn:nbn:se:umu:diva-86213OAI: oai:DiVA.org:umu-86213DiVA: diva2:698140