De-identification of clinical free text using natural language processing: A systematic review of current approaches.

BACKGROUND: Electronic health records (EHRs) are a valuable resource for data-driven medical research. However, the presence of protected health information (PHI) makes EHRs unsuitable to be shared for research purposes. De-identification, i.e. the process of removing PHI is a critical step in making EHR data accessible. Natural language processing has repeatedly demonstrated its feasibility in automating the de-identification process. OBJECTIVES: Our study aims to provide systematic evidence on how the de-identification of clinical free text written in English has evolved in the last thirteen years, and to report on the performances and limitations of the current state-of-the-art systems for the English language. In addition, we aim to identify challenges and potential research opportunities in this field. METHODS: A systematic search in PubMed, Web of Science, and the DBLP was conducted for studies published between January 2010 and February 2023. Titles and abstracts were examined to identify the relevant studies. Selected studies were then analysed in-depth, and information was collected on de-identification methodologies, data sources, and measured performance. RESULTS: A total of 2125 publications were identified for the title and abstract screening. 69 studies were found to be relevant. Machine learning (37 studies) and hybrid (26 studies) approaches are predominant, while six studies relied only on rules. The majority of the approaches were trained and evaluated on public corpora. The 2014 i2b2/UTHealth corpus is the most frequently used (36 studies), followed by the 2006 i2b2 (18 studies) and 2016 CEGS N-GRID (10 studies) corpora. CONCLUSION: Earlier de-identification approaches aimed at English were mainly rule and machine learning hybrids with extensive feature engineering and post-processing, while more recent performance improvements are due to feature-inferring recurrent neural networks. Current leading performance is achieved using attention-based neural models. Recent studies report state-of-the-art F1-scores (over 98 %) when evaluated in the manner usually adopted by the clinical natural language processing community. However, their performance needs to be more thoroughly assessed with different measures to judge their reliability to safely de-identify data in a real-world setting. Without additional manually labeled training data, state-of-the-art systems fail to generalise well across a wide range of clinical sub-domains.

Building the European Social Innovation Database with Natural Language Processing and Machine Learning.

Social innovation is widely defined as technological and non-technological new products, services or models that simultaneously meet social needs and create new social relationships or collaborations. Despite a significant interest in the concept, the lack of reliable and comprehensive data is a barrier for social science research. We created the European Social Innovation Database (ESID) to address this gap. ESID is based on the idea of large-scale collection of unstructured web site text to classify and characterise social innovation projects from around the world. We use advanced machine learning techniques to extract features such as social innovation dimensions, project locations, summaries, and topics, among others. Our models perform as high as 0.90 F1. ESID currently includes 11,468 projects from 159 countries. ESID data is available freely and also presented in a web-based app. Our future workplan includes expansion (i.e., increasing the number of projects), extension (i.e., adding new variables) and dynamic retrieval (i.e., retrieving and extracting information in regular intervals).

Effects of Sodium Bicarbonate Ingestion on Recovery in High-Level Judokas.

BACKGROUND: The aim of this study was to examine the sodium bicarbonate (NaHCO3) effect on recovery in high-level judokas. METHODS: The sample of participants consisted of 10 male judokas (Age = 20 ± 2.1 years) who are judo masters (black belt holders) with a minimum of 10 years of training and competition experience. The study was designed as a double-blinded crossover design with the order of treatments being randomly assigned. The washout period was 72 h. All subjects received a dose of sodium bicarbonate (0.3 g/kg body weight) or a placebo 120 min before the fatigue caused by the special judo fitness test (SJFT). Lactate concentration (LC), countermovement jump (CMJ), hand grip strength and degree of perceived fatigue on Borg's scale (RPE) were tested two times before SJFT and four times after SJFT. RESULTS: There was no interaction between groups and type of recovery at any time during the two types of recovery for RPE, grip strength, VJ and lactate concentration (p > 0.05). However, there was a main effect of time for dominant grip strength (F(1,8)= 3.3; p = 0.01; η2 = 0.25, (small)), non-dominant grip strength (F(1,8) = 3.2; p = 0.01; η2 = 0.24, (small)), CMJ (F(1,8) = 8.8; p = 0.01; η2 = 0.47, (small)), and LC (F(1,8) = 124.2; p = 0.001; η2 = 0.92, (moderate)). CONCLUSIONS: The results of the present study show no significant difference between the NaHCO3 and placebo groups in RPE, handgrip strength, CMJ, and lactate concentration.

Aggressiveness in Judokas and Team Athletes: Predictive Value of Personality Traits, Emotional Intelligence and Self-Efficacy.

Combat sports and martial arts are often associated with aggressiveness among the general public, although data on judo and/or martial arts and aggressiveness seem to be unclear. This research aims to compare athletes who have trained judo for a prolonged time (minimum 5 years) and athletes from various team sports, primarily regarding the manifestation of aggression, but also regarding personality traits, emotional intelligence, and self-efficacy. Also, the potential predictive value of personality traits, emotional intelligence, and self-efficacy for aggression within subsamples of judokas and team athletes was tested. The research findings showed that professional judo athletes are characterized by a low degree of aggression, especially low indirect and physical manifestations of aggression. In addition, the personality traits Honesty-Humility and Openness to experience are well expressed, contrary to Emotionality and Extraversion, which are less pronounced. They are also characterized by moderate general self-efficacy. On the other hand, members of team sports produced the opposite results, as they are characterized by increased aggression, pronounced traits of Emotionality and Extraversion, somewhat less pronounced traits of Honesty-Humility, Openness to new experience, and less pronounced general self-efficacy. The percentage of explained variability of aggression is slightly higher in the subsample of team sports and constitutes 49.9% of the variability, while in the subsample of judokas it constitutes 47.8% of the variability of the criteria. Practical implications, limitations, and future research directions were discussed.