Search

A significant level of stigma and inequality exists in mental healthcare, especially in under-served populations. Inequalities are reflected in the data collected for scientific purposes. When not properly accounted for, machine learning (ML) models learned from data can reinforce these structural inequalities or biases. Here, we present a systematic study of bias in ML models designed to predict depression in four different case studies covering different countries and populations. We find that standard ML approaches regularly present biased behaviors. We also show that mitigation techniques, both standard and our own post-hoc method, can be effective in reducing the level of unfair bias. There is no one best ML model for depression prediction that provides equality of outcomes. This emphasizes the importance of analyzing fairness during model selection and transparent reporting about the impact of debiasing interventions. Finally, we also identify positive habits and open challenges that practitioners could follow to enhance fairness in their models., (© 2024. The Author(s).)

Aim: Evidence-based practice is a way to minimize the theory-to-practice gap. However, translating evidence into practice is not merely appraising the evidence, and for applying it to patients, the context factors need to be included. This study aims to analyze the meaning of the concept of "evidence translating ability" in nursing., Method: Walker and Avant's method was utilized in this study. Keywords such as evidence, evident, transferring, translating, and ability were mixed for literature searching in PubMed, CINAHL, and Google Scholar from 2017 to April 2022., Results: Nine studies that met inclusion criteria were analyzed. Attributes of evidence translating ability included four themes: identification of contextual barriers and facilitators to implementation, evaluation of evidence, collaboration with colleagues to modify recommendations, and pilot and dissemination. Antecedents included two themes: facilitating environment and experience of nurses. Consequences included three themes: motivating and sustaining evidence-based practice, providing high-quality care, and reducing inequity in health care., Conclusion: Evidence translating ability requires a set of skills that nurses have to do such as identifying the challenges and analyzing the setting for searching the evidence; the best evidence is not the highest evidence hierarchy, the best evidence is the evidence that suits patient, organization, and practitioner abilities.

Deep learning techniques for 3D brain vessel image segmentation have not been as successful as in the segmentation of other organs and tissues. This can be explained by two factors. First, deep learning techniques tend to show poor performances at the segmentation of relatively small objects compared to the size of the full image. Second, due to the complexity of vascular trees and the small size of vessels, it is challenging to obtain the amount of annotated training data typically needed by deep learning methods. To address these problems, we propose a novel annotation-efficient deep learning vessel segmentation framework. The framework avoids pixel-wise annotations, only requiring weak patch-level labels to discriminate between vessel and non-vessel 2D patches in the training set, in a setup similar to the CAPTCHAs used to differentiate humans from bots in web applications. The user-provided weak annotations are used for two tasks: (1) to synthesize pixel-wise pseudo-labels for vessels and background in each patch, which are used to train a segmentation network, and (2) to train a classifier network. The classifier network allows to generate additional weak patch labels, further reducing the annotation burden, and it acts as a second opinion for poor quality images. We use this framework for the segmentation of the cerebrovascular tree in Time-of-Flight angiography (TOF) and Susceptibility-Weighted Images (SWI). The results show that the framework achieves state-of-the-art accuracy, while reducing the annotation time by ∼77% w.r.t. learning-based segmentation methods using pixel-wise labels for training., (Copyright © 2021 The Authors. Published by Elsevier B.V. All rights reserved.)

The objective of this study was to describe the overall pattern of morbidity and mortality of children seen at the Thai Binh Paediatric Hospital in Vietnam, with a focus on infectious diseases. A retrospective review of hospitalisation records was conducted from 1 January 2015 to 31 December 2019. Data were obtained from a total of 113,999 records. The median age of patients was 18 months, with 84.0% of patients aged <5 years. Infectious diseases accounted for 61.0% of all cases. The most prevalent diseases were lower respiratory tract infections (32.8%), followed by gastrointestinal infections (13.3%) and confirmed influenza (5.4%). Most infections were not microbiologically documented. A total of 81.4% patients received at least one antibiotic. Most patients (97.0%) were hospitalised for less than 15 days. Regarding outcomes, 87.8% patients were discharged home with a favourable outcome. Twelve percent were transferred to the Vietnam National Children's Hospital because their condition had worsened and 0.1% died. In total, infectious diseases accounted for 40.4% of deaths, followed by neonatal disorders (34.6%). Our data serves a basis for the identification of needs for diagnostic tools and for future evaluation of the effect of the targeted implementation of such facilities. Point-of-care tests, including real-time polymerase chain reaction assays to identify common pathogens should be implemented for more accurate diagnosis and more appropriate antibiotic use., Competing Interests: The authors declare they have no conflicts of interest., (© 2020 The Authors. Published by Atlantis Press International B.V.)

Vietnam is one of the 15 countries where the prevalence of child pneumonia is highest. It is a major cause of admission in pediatric hospitals. However, little is known on the burden of severe pneumonia and their risk factors in children <5 years of age in Vietnam. A case-control study was conducted among children aged 2-59 months presenting with pneumonia at the Pediatric Provincial Hospital of Thai Binh. Cases were children with severe pneumonia while controls included those with non-severe pneumonia as defined by the World Health Organization (WHO) classification of 2005. Eighty-three cases and 83 controls were included. Sex ratio was 2.19. Children with severe pneumonia were significantly less likely to receive antibiotics preadmission compared to children with non-severe pneumonia [odds ratio (OR) = 0.16, 95% confidence interval (CI) = 0.06-0.42]. The main risk factors of severe pneumonia were a lack of immunization (OR = 4.77, 95% CI = 1.80-12.65), an exposure to cigarette smoke (OR = 3.87, 95% CI = 1.62-9.23), and having a mother with a low level of education. Children with severe pneumonia were 25 times more likely to present with associated measles with p < 0.0001 and five times more likely to present with diarrhea than children with non-severe pneumonia ( p < 0.0001). Improving immunization coverage, educating parents about the risks of passive smoking and the recognition of respiratory distress signs, and facilitating early antibiotic access for infants with acute pulmonary disease should reduce the burden of such illnesses. To implement a national, multicenter study about pneumonia in children, more precise inclusion criteria should be chosen, including radiological and/or biological assessment., Competing Interests: The authors declare they have no conflicts of interest., (© 2019 Atlantis Press International B.V.)

This paper illustrates a method to identify and classify scenarios generated in a dynamic event tree (DET) analysis. Identification and classification are carried out by means of an evolutionary possibilistic fuzzy C-means clustering algorithm which takes into account not only the final system states but also the timing of the events and the process evolution. An application is considered with regards to the scenarios generated following a steam generator tube rupture in a nuclear power plant. The scenarios are generated by the accident dynamic simulator (ADS), coupled to a RELAP code that simulates the thermo-hydraulic behavior of the plant and to an operators' crew model, which simulates their cognitive and procedures-guided responses. A set of 60 scenarios has been generated by the ADS DET tool. The classification approach has grouped the 60 scenarios into 4 classes of dominant scenarios, one of which was not anticipated a priori but was "discovered" by the classifier. The proposed approach may be considered as a first effort towards the application of identification and classification approaches to scenarios post-processing for real-scale dynamic safety assessments.