Multidimensional reference regions is a new tool to optimize the personalized care of patients.

Medical diagnostic processes often rely on comprehensive biological assessments, but current methods have drawbacks, such as insufficient consideration of variable dependence. Modern databases enable precise estimation of multidimensional variable distributions, prompting this study to enhance methodologies for biological variables. The focus is on establishing better reference regions and defining more accurate decision boundaries. Using an American database (1999-2017) with 19,231 healthy and 24,257 diseased individuals, the study examined plasma biochemical markers. The methodology involved constructing reference regions based on level sets of the healthy and diseased distributions and establishing decision boundaries. An example involved selecting diseased liver patients and considering 9 biological variables characterizing liver dysfunction. The proposed method is consistently more sensitive and specific than traditional approaches. Our results show that other biological functions, such as renal and cardiac functions, in a patient with liver dysfunction are also likely to be altered, even if the biochemical variables measuring these functions remain within the reference range. Therefore, it is preferable to consider a large panel of biological variables rather than just 9 to accurately characterize liver function. Despite potential limitations in identifying diseases, the proposed methodology provides valuable insights into physiological dysfunctions, enhancing early detection.