From business to clinical trials: a systematic review of the literature on fraud detection methods to be used in central statistical monitoring

Abstract

Data-driven decisions can be suboptimal when the data are distorted by fraudulent behaviour. Fraud is a common occurrence in finance or other related industries, where large datasets are handled and motivation for financial gain may be high. In order to detect and prevent fraud, quantitative methods are used. Fraud, however, is also committed in other circumstances, e.g. during clinical trials. The article aims to verify which analytical fraud-detection methods used in finance may be adopted in the field of clinical trials. We systematically reviewed papers published over the last five years in two databases (Scopus and Web of Science) from the field of economics, finance, management and business in general. We considered the broad scope of data mining techniques including artificial intelligence algorithms. As a result, 37 quantitative methods were identified with the potential of being fit for application in clinical trials. The methods were grouped into three categories: pre-processing techniques, supervised learning and unsupervised learning. Our findings may enhance the future use of fraud-detection methods in clinical trials.