Interpreting statistics from published research to answer clinical and management questions

Abstract

Appropriate statistical analysis is critical in interpreting results from published literature to answer clinical and management questions. Internal validity is an assessment of whether the study design and statistical analysis are appropriate for the hypotheses and study variables while controlling for bias and confounding. External validity is an assessment of the appropriateness of extrapolation of the study results to other populations. Knowledge about whether treatment or observation groups are truly different is unknown, but studies can be broadly categorized as exploratory or discovery, based on knowledge about previous research, biology, and study design, and this categorization affects interpretation. Confidence intervals, P-values, prediction intervals, credible intervals, and other decision aids are used singly or in combination to provide evidence for the likelihood of a given model but can be interpreted only if the study is internally valid. These decision aids do not test for bias, study design, or the appropriateness of applying study results to other populations dissimilar to the population tested. The biologic and economic importance of the magnitude of difference between treatment groups or observation groups as estimated by the study data and statistical interpretation is important to consider in clinical and management decisions. Statistical results should be interpreted in light of the specific question and production system addressed, the study design, and knowledge about pertinent aspects of biology to appropriately aid decisions.