Sobre a qualidade das bases de dados públicas brasileiras: estudo de caso na saúde

Abstract

With the ever-growing digitalization of governments worldwide, great effort has been put towards data transparency. One important milestone of this trend is the Open Government Data workshop, which took place in 2007 in Sebastopol, California, where the eight Open Government Data principles were defined. Since then, large scale initiatives from government institutions worldwide were put in place to make their datasets more transparent, and among them, Brazilian institutions. With initiatives such as the Information Access Law and the Brazilian Portal of Open Data (Lei de Acesso à Informação and Portal Brasileiro de Dados Abertos, respectively, unofficially translated), data from municipal, state and federal Brazilian government institutions are accessible to any interested citizen. However, effort towards data quality is as important as the openness of the data itself, so it can be used clearly, optimally, and as granularly as wished. Picking up from selected datasets made available by the Brazilian unified healthcare system (Sistema Único de Saúde - SUS), through the Datasus Tabnet platform, and the national agency for health insurance providers (Agência Nacional de Saúde Suplementar - ANS ), this paper aims to analyze the quality of this data, proposing meaningful and reusable improvements to any government entity that wishes to open its data, with great care for quality. Leveraging acquired knowledge from a project executed by the Chicago-based multinational consulting firm Kearney for a hospital based in São Paulo, where numerous market analyses were based on open data provided by SUS and the ANS, the dimensions for data quality proposed by Parker et al. were used to analyze the available sets. In order to exemplify where such quality effectively interferes in the easiness of the work being done, sanitized analyses were brought up, showcasing the value of opening government data, and how they can be continuously improved.