Qualidade de dados na indústria da saúde: perspectivas de Big Data e aprendizado de máquina - uma revisão de escopo

Abstract

Introduction: Technologies such as Big Data, Machine Learning, and Deep Learning have great potential in the health industry and can be used to improve the process of diagnosis, treatment, and health management. However, to effectively use these technologies, it is necessary to ensure the quality of the data to make them capable of achieving the proposed objective. Objective: In this context, a scoping review was carried out to identify and map methods and frameworks used to treat and evaluate data quality in medical and health applications. Methods and Analysis: The scoping review was conducted based on the six-step process described in the framework proposed by Levac et al. in "Scoping Studies: Advancing the methodology" and reported following the PRISMA-ScR checklist (Preferred Reporting Items for Systematic Reviews and Meta-Analyses Extension for Scoping Reviews). In the search for articles, terms associated with data quality, big data, machine learning, deep learning, and health were used to search for primary studies in the Scopus Document Search, IEEE Xplore Digital Library, and ACM Digital Library databases. Results: The research included 93 articles that allowed us to answer the research questions, providing information on the benefits of applying these technologies in the health industry, the most commonly used data preprocessing methods, their benefits, and mitigated problems. It was also identified that the Data Quality Dimensions most addressed in the research are reliability, usability, and availability. Conclusions: The study lets us understand the main metrics used to assess data quality. It also highlighted gaps in the relationship between methods and dimensions, providing an opportunity to develop a framework to guide the selection and application of appropriate methods for each quality dimension.