The Limitations of Big Data in Healthcare

Abstract

It has become commonplace in modern society to collect and process large bodies of data in order to establish trends in the data and thereby identify business opportunities, areas of cost-saving, etc. This article questions whether, in the healthcare context, the data sets from biomarker-type tests are reliable indicators of dysfunction and whether the processing of such data sets can be expected to identify significant areas for improvement in the clinical context which would lead to significant improvements and cost-savings regarding the diagnosis and treatment of morbidity. In particular, there is not yet an accepted understanding of the mechanisms which are responsible for maintaining the body’s highly regulated function therefore the idea that the current plethora of diagnostic tests can be accurate measures of the rate of pathological onset of a particular medical condition(s) is only an assumption and may have significant inherent limitations. The development of the first mathematical model of the autonomic nervous system by Grakov leads us to question basic assumptions upon which modern medicine is based and whether large-scale data mining of available data sets will lead to a better understanding e.g. (i) Is the biomarker the cause or the consequence of the condition? (ii) Does the marker relate to the genotype, phenotype or neither? (iii) How does the brain regulate the autonomic nervous system and the coherent function of the organ networks/physiological systems?