Desenvolvimento de software de gerenciamento de informações em controle de infecção hospitalar

Abstract

Background: The Hospital Infection Control Committee (HICC) is responsible for adequate diagnosis, treatment, and prevention of Healthcare Associated Infections (HAI) in hospitals, and epidemiologic surveillance is its main activity. Epidemiologic surveillance requires collecting and managing a big amount of data, usually through different tools, such as paper forms, electronic sheets, small databases etc. Objectives: Understand the work process and information flow of a HICC to develop an application for mobile devices capable of managing all data collection, storage, access, and Key Performance Indicators (KPI) production in a HICC. Predict patients with enterobacteria producing ESBL infections before culture results are available with the use of Artificial Intelligence (AI). Methods: Value Stream Mapping (VSM) was applied to study information flow in the HICC and to find opportunities for improvement in the work process. We also performed a systematic review to identify the necessary characteristics a healthcare software must have to ensure security and privacy of data collected. The authors used the results of these studies to design the functionalities of the software. Application programming used Java language and the blockchain technology to manage data collection and storage. Standard programing produced the KPI required by the HICC team and hospital directors, and we created an AI algorithm to predict patients with ESBL producing enterobacteria infections before the microbiology cultures results are available. Results: We applied VSM to represent the current work process of the HICC (actual VSM) and then be critically reviewed by an interdisciplinary team, producing an ideal VSM with 15,7 less hours/month, less activities done retrospectively and automated KPI production. The research continued with the development of the software with blockchain and AI. A private blockchain was used (Ethereum platform) for data storage. Each blockchain node has access to the complete database but only to its respective hospital data. The patient identification data used a second blockchain contract with cryptography to reinforce data security and privacy. Blockchain latency was countered with SQLite database used as buffer until the next blocks were registered. The AI algorithm was created using all data from epidemiological surveillance charts, known risk factors for these infections and all other information available in the HICC recorded data. AI was applied using Artificial Neural Networks (ANN) with 10 neurons (or hidden layers) and supervised learning, where the training is guided by receiving feedback. A backpropagation algorithm with a feedforward network topology was used. Conclusion: A mobile device application using VSM, blockchain and AI was developed. The methods/technologies used can be complementary and the resulting software can bring substantial benefits when applied to a HICC. [resumo fornecido pelo autor]