Multidimensional Detection Of Outliers In Clinical Registers

Abstract

Incorrect data in clinical registers can lead to inaccurate or wrong results. This project is aimed at monitoring and evaluation of data in clinical registers. Usual methods to identify incorrect data are one-dimensional statistical methods per each variable in the register. Proposed method finds outliers in data using machine learning combined with multidimensional statistical methods that transform all column variables of clinical register to one, representing one record of a patient in the register.