Development of a decision support system in haemovigilance

Laburpena

Nowadays, the use of computer systems in blood banks is universal. Since the identification of the donor, all the steps and staff involved in the process are recorded, allowing an effective traceability. In the 1990s, the area of Haemovigilance emerged with the aim of making transfusion even more secure: "... a set of surveillance procedures that cover the entire chain ... in order to collect and analyze information on the effects unexpected or unwanted results that result from the therapeutic use of labile blood products, and to prevent their occurrence and recurrence”. The Portuguese Hemovigilance System (PHvS) is managed the Portuguese Blood Institute since 2007. The Institute created a system of computerized records of the incidents reported by the Transfusion Medicine services, accessible through the website http://www.hemovigilancia.net. From 2007 to 2010, the number of rcords tripled, reinforcing its position as a key resource in routine Transfusion Medicine services. However, the objectives set initially focused on supporting data entry and storage, that is, obtaining records of adverse reactions. It was assumed that the description and analysis of these records would be "paid for" by the coordinating group of the PHvS. Specifically, the coordinating group of the PHvS was commissioned to prepare an annual report. To date, the production of this report has not been automated and its contents are limited to descriptive statistics and group comments. Since the registration system is consolidated, and the content volume continues to grow, there is the technical possibility and the interest of all the participants and decision-makers in the process to develop new automatic mechanisms with (pro) active inference capacity. In this work plan, the design and implementation of a support system for decision making in Hemovigilance, specialized in the analysis of adverse reaction reports is proposed. The initial objective is to automate the collection of statistics and other information for the production of reports, but it is also intended to evaluate the data mining models and their introduction as a new methodology to support the work of the coordinating group of the PHvS. In particular, this work plan includes the following stages: • discuss with the coordinating group of the PHvS the most urgent analytical requirements at this time and in the medium term; • analyze paper reports and define an analytical data model, with the capacity to cover information of analytical interest; • develop a system for the automatic production of reports, capable of summarizing or adding the data recorded when requested, that is, trying to give answers to the different questions instead of maintaining rigid queries; • define the architecture of a new computer system using Web technologies and database management systems, assuring compatibility with the existing system; • validate and consolidate the information system, performing on-site tests; • prepare different sets of data to evaluate the analysis capacity of different data mining algorithms; • develop an advanced analysis system with the most promising data mining models, in order to offer a different perspective of analysis of registered notifications. The proposed developments are new and are designed to complement a real system, in an application area with an undoubted social impact. The work is planned in order to meet current and pressing needs, but it already includes the introduction of advanced analytical methodologies. On the other hand, it is planned to work in close collaboration with the coordinating group of the PHvS to guarantee the suitability and usefulness of the developed systems.