RESUMO
OBJECTIVE: To describe and appraise the use of artificial intelligence (AI) techniques that can cope with longitudinal data from electronic health records (EHRs) to predict health-related outcomes. METHODS: This review included studies in any language that: EHR was at least one of the data sources, collected longitudinal data, used an AI technique capable of handling longitudinal data, and predicted any health-related outcomes. We searched MEDLINE, Scopus, Web of Science, and IEEE Xplorer from inception to January 3, 2022. Information on the dataset, prediction task, data preprocessing, feature selection, method, validation, performance, and implementation was extracted and summarized using descriptive statistics. Risk of bias and completeness of reporting were assessed using a short form of PROBAST and TRIPOD, respectively. RESULTS: Eighty-one studies were included. Follow-up time and number of registers per patient varied greatly, and most predicted disease development or next event based on diagnoses and drug treatments. Architectures generally were based on Recurrent Neural Networks-like layers, though in recent years combining different layers or transformers has become more popular. About half of the included studies performed hyperparameter tuning and used attention mechanisms. Most performed a single train-test partition and could not correctly assess the variability of the model's performance. Reporting quality was poor, and a third of the studies were at high risk of bias. CONCLUSIONS: AI models are increasingly using longitudinal data. However, the heterogeneity in reporting methodology and results, and the lack of public EHR datasets and code sharing, complicate the possibility of replication. REGISTRATION: PROSPERO database (CRD42022331388).
Assuntos
Inteligência Artificial , Registros Eletrônicos de Saúde , HumanosRESUMO
Risk taking (RT) is a component of the decision-making process in situations that involve uncertainty and in which the probability of each outcome - rewards and/or negative consequences - is already known. The influence of cognitive and emotional processes in decision making may affect how risky situations are addressed. First, inaccurate assessments of situations may constitute a perceptual bias in decision making, which might influence RT. Second, there seems to be consensus that a proneness bias exists, known as risk proneness, which can be defined as the propensity to be attracted to potentially risky activities. In the present study, we take the approach that risk perception and risk proneness affect RT behaviours. The study hypothesises that locus of control, emotion regulation, and executive control act as perceptual biases in RT, and that personality, sensation seeking, and impulsivity traits act as proneness biases in RT. The results suggest that locus of control, emotion regulation and executive control influence certain domains of RT, while personality influences in all domains except the recreational, and sensation seeking and impulsivity are involved in all domains of RT. The results of the study constitute a foundation upon which to build in this research area and can contribute to the increased understanding of human behaviour in risky situations.
RESUMO
Fixation identification is an essential task in the extraction of relevant information from gaze patterns; various algorithms are used in the identification process. However, the thresholds used in the algorithms greatly affect their sensitivity. Moreover, the application of these algorithm to eye-tracking technologies integrated into head-mounted displays, where the subject's head position is unrestricted, is still an open issue. Therefore, the adaptation of eye-tracking algorithms and their thresholds to immersive virtual reality frameworks needs to be validated. This study presents the development of a dispersion-threshold identification algorithm applied to data obtained from an eye-tracking system integrated into a head-mounted display. Rules-based criteria are proposed to calibrate the thresholds of the algorithm through different features, such as number of fixations and the percentage of points which belong to a fixation. The results show that distance-dispersion thresholds between 1-1.6° and time windows between 0.25-0.4 s are the acceptable range parameters, with 1° and 0.25 s being the optimum. The work presents a calibrated algorithm to be applied in future experiments with eye-tracking integrated into head-mounted displays and guidelines for calibrating fixation identification algorithms.
