Incorrect inference in prevalence trend analysis due to misuse of the odds ratio

Because public health agencies usually monitor health outcomes over time for surveillance, program evaluation, and policy decisions, a correct health outcome trend analysis is vital. If the analysis is done incorrectly and/or results are misinterpreted, a faulty trend analysis could jeopardize key aspects of public health initiatives such as program planning, implementations, policy development, and clinical decision making. It is essential then that accurate health outcome trend analysis be implemented in any data-driven decision-making process. Unfortunately, there continues to be common statistical mistakes in prevalence trend analysis. In this article, using recently published results from the Pediatric Nutrition Surveillance System, we will show the effect that an incorrect trend analysis and subsequent interpretation of results can have. We will also propose more appropriate statistical processes, such as the log-binomial model, for these situations.