{Reference Type}: Journal Article
{Title}: The Conclusion Generator.
{Author}: Schmidt M;Parner E;
{Journal}: Ann Epidemiol
{Volume}: 0
{Issue}: 0
{Year}: 2024 Jun 28
{Factor}: 6.996
{DOI}: 10.1016/j.annepidem.2024.06.008
{Abstract}: <B>OBJECTIVE: </B>Reliance on null hypothesis significance testing often leads to misinterpretation of research results. Common misinterpretations include that a statistically nonsignificant difference (p≥0.05) implies no difference between groups, and that a statistically significant finding (p<0.05) is unbiased and clinically important. We aimed to develop a tool - the Conclusion Generator - to mitigate these misconceptions.<BR><B>METHODS: </B>We reviewed the content of the Conclusion Generator and validated its output using published and simulated data.<BR><B>RESULTS: </B>The Conclusion Generator is a free online application designed to generate conclusions for scientific papers based on the values and clinical interpretation of the point estimate and confidence interval. Both relative and absolute measures of effect are supported. It offers two modes for interpretation: (1) Statistical mode provides an accurate statistical interpretation of results, with an optional specification of superiority and noninferiority bounds; (2) Clinical mode evaluates the clinical importance of the point estimate and confidence limits as specified by the user. Both modes assume no uncontrolled biases. Users must specify the number of decimals, the direction of a beneficial effect (e.g., relative risk <1 vs. >1), and the level of detail (concise vs. elaborated) for the output. The validation confirmed the Conclusion Generator's capability to interpret research results, considering random error and clinical relevance, while avoiding common misinterpretations associated with null hypothesis significance testing.<BR><B>CONCLUSIONS: </B>The Conclusion Generator facilitates an appropriate interpretation of research results by emphasizing estimation and clinical relevance over hypothesis testing.