-
Brain Communications 2021Alzheimer's disease is a neurodegenerative disorder and the most common form of dementia. Early diagnosis may assist interventions to delay onset and reduce the... (Review)
Review
Alzheimer's disease is a neurodegenerative disorder and the most common form of dementia. Early diagnosis may assist interventions to delay onset and reduce the progression rate of the disease. We systematically reviewed the use of machine learning algorithms for predicting Alzheimer's disease using single nucleotide polymorphisms and instances where these were combined with other types of data. We evaluated the ability of machine learning models to distinguish between controls and cases, while also assessing their implementation and potential biases. Articles published between December 2009 and June 2020 were collected using Scopus, PubMed and Google Scholar. These were systematically screened for inclusion leading to a final set of 12 publications. Eighty-five per cent of the included studies used the Alzheimer's Disease Neuroimaging Initiative dataset. In studies which reported area under the curve, discrimination varied (0.49-0.97). However, more than half of the included manuscripts used other forms of measurement, such as accuracy, sensitivity and specificity. Model calibration statistics were also found to be reported inconsistently across all studies. The most frequent limitation in the assessed studies was sample size, with the total number of participants often numbering less than a thousand, whilst the number of predictors usually ran into the many thousands. In addition, key steps in model implementation and validation were often not performed or unreported, making it difficult to assess the capability of machine learning models.
PubMed: 34805994
DOI: 10.1093/braincomms/fcab246 -
Revista Medica de Chile Jul 2021Students belonging to sexual and gender diversity experience chronic stress due to stigmatization and discrimination.
BACKGROUND
Students belonging to sexual and gender diversity experience chronic stress due to stigmatization and discrimination.
AIM
To identify the experiences of lesbian, gay, bisexual, transgender/transsexual, and queer (LGBTQ+) medical students.
MATERIAL AND METHODS
Systematic literature review using the PRISMA protocol in PubMed, ERIC, EMBASE, and LILACS databases. Articles published in Spanish or English were considered. Three authors independently reviewed and synthesized information from the selected articles, according to the PRISMA criteria.
RESULTS
Fifteen studies met the inclusion criteria. Forty-three experiences were reported, which were finally classified into four categories: i) Relationship between peers in the educational context (23%), ii) Relationship between students and teachers in the educational context (23%), iii) Relationship with the educational institution (34%), and iv) Curriculum and training experience (19%). The relationship with the educational institution was identified as the most relevant category. Students with a strong sense of belonging to their institution were more likely to be persistent and make an effort in learning. The second most relevant experiences, mainly negative, derive from interactions with peers and teachers.
CONCLUSIONS
LGBTQ+ medical students still experience more discrimination than inclusion during their training. Therefore, medical schools should render medical education a more inclusive space for the LGBTQ+ population.
Topics: Female; Gender Identity; Homosexuality, Female; Humans; Sexual Behavior; Sexual and Gender Minorities; Students, Medical
PubMed: 34751308
DOI: 10.4067/s0034-98872021000701058 -
Journal of Diabetes Science and... Mar 2023With the rising prevalence of diabetes, machine learning (ML) models have been increasingly used for prediction of diabetes and its complications, due to their ability...
BACKGROUND
With the rising prevalence of diabetes, machine learning (ML) models have been increasingly used for prediction of diabetes and its complications, due to their ability to handle large complex data sets. This study aims to evaluate the quality and performance of ML models developed to predict microvascular and macrovascular diabetes complications in an adult Type 2 diabetes population.
METHODS
A systematic review was conducted in MEDLINE®, Embase®, the Cochrane® Library, Web of Science®, and DBLP Computer Science Bibliography databases according to the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) checklist. Studies that developed or validated ML prediction models for microvascular or macrovascular complications in people with Type 2 diabetes were included. Prediction performance was evaluated using area under the receiver operating characteristic curve (AUC). An AUC >0.75 indicates clearly useful discrimination performance, while a positive mean relative AUC difference indicates better comparative model performance.
RESULTS
Of 13 606 articles screened, 32 studies comprising 87 ML models were included. Neural networks (n = 15) were the most frequently utilized. Age, duration of diabetes, and body mass index were common predictors in ML models. Across predicted outcomes, 36% of the models demonstrated clearly useful discrimination. Most ML models reported positive mean relative AUC compared with non-ML methods, with random forest showing the best overall performance for microvascular and macrovascular outcomes. Majority (n = 31) of studies had high risk of bias.
CONCLUSIONS
Random forest was found to have the overall best prediction performance. Current ML prediction models remain largely exploratory, and external validation studies are required before their clinical implementation.
PROTOCOL REGISTRATION
Open Science Framework (registration number: 10.17605/OSF.IO/UP49X).
Topics: Adult; Humans; Diabetes Mellitus, Type 2; Machine Learning; Neural Networks, Computer; ROC Curve
PubMed: 34727783
DOI: 10.1177/19322968211056917 -
Heart (British Cardiac Society) Jun 2022Atrial fibrillation (AF) is common and is associated with an increased risk of stroke. We aimed to systematically review and meta-analyse multivariable prediction models... (Meta-Analysis)
Meta-Analysis
OBJECTIVE
Atrial fibrillation (AF) is common and is associated with an increased risk of stroke. We aimed to systematically review and meta-analyse multivariable prediction models derived and/or validated in electronic health records (EHRs) and/or administrative claims databases for the prediction of incident AF in the community.
METHODS
Ovid Medline and Ovid Embase were searched for records from inception to 23 March 2021. Measures of discrimination were extracted and pooled by Bayesian meta-analysis, with heterogeneity assessed through a 95% prediction interval (PI). Risk of bias was assessed using Prediction model Risk Of Bias ASsessment Tool and certainty in effect estimates by Grading of Recommendations, Assessment, Development and Evaluation.
RESULTS
Eleven studies met inclusion criteria, describing nine prediction models, with four eligible for meta-analysis including 9 289 959 patients. The CHADS (Congestive heart failure, Hypertension, Age>75, Diabetes mellitus, prior Stroke or transient ischemic attack) (summary c-statistic 0.674; 95% CI 0.610 to 0.732; 95% PI 0.526-0.815), CHADS-VASc (Congestive heart failure, Hypertension, Age>75 (2 points), Stroke/transient ischemic attack/thromboembolism (2 points), Vascular disease, Age 65-74, Sex category) (summary c-statistic 0.679; 95% CI 0.620 to 0.736; 95% PI 0.531-0.811) and HATCH (Hypertension, Age, stroke or Transient ischemic attack, Chronic obstructive pulmonary disease, Heart failure) (summary c-statistic 0.669; 95% CI 0.600 to 0.732; 95% PI 0.513-0.803) models resulted in a c-statistic with a statistically significant 95% PI and moderate discriminative performance. No model met eligibility for inclusion in meta-analysis if studies at high risk of bias were excluded and certainty of effect estimates was 'low'. Models derived by machine learning demonstrated strong discriminative performance, but lacked rigorous external validation.
CONCLUSIONS
Models externally validated for prediction of incident AF in community-based EHR demonstrate moderate predictive ability and high risk of bias. Novel methods may provide stronger discriminative performance.
SYSTEMATIC REVIEW REGISTRATION
PROSPERO CRD42021245093.
Topics: Aged; Atrial Fibrillation; Bayes Theorem; Electronic Health Records; Heart Failure; Humans; Hypertension; Ischemic Attack, Transient; Risk Assessment; Risk Factors; Stroke
PubMed: 34607811
DOI: 10.1136/heartjnl-2021-320036 -
JCPP Advances Oct 2021There has been a rapid growth in the publication of new prediction models relevant to child and adolescent mental health. However, before their implementation into... (Review)
Review
BACKGROUND
There has been a rapid growth in the publication of new prediction models relevant to child and adolescent mental health. However, before their implementation into clinical services, it is necessary to appraise the quality of their methods and reporting. We conducted a systematic review of new prediction models in child and adolescent mental health, and examined their development and validation.
METHOD
We searched five databases for studies developing or validating multivariable prediction models for individuals aged 18 years old or younger from 1 January 2018 to 18 February 2021. Quality of reporting was assessed using the Transparent Reporting of a multivariable prediction models for Individual Prognosis Or Diagnosis checklist, and quality of methodology using items based on expert guidance and the PROBAST tool.
RESULTS
We identified 100 eligible studies: 41 developing a new prediction model, 48 validating an existing model and 11 that included both development and validation. Most publications ( = 75) reported a model discrimination measure, while 26 investigations reported calibration. Of 52 new prediction models, six (12%) were for suicidal outcomes, 18 (35%) for future diagnosis, five (10%) for child maltreatment. Other outcomes included violence, crime, and functional outcomes. Eleven new models (21%) were developed for use in high-risk populations. Of development studies, around a third were sufficiently statistically powered ( = 16%, 31%), while this was lower for validation investigations ( = 12, 25%). In terms of performance, the discrimination (as measured by the C-statistic) for new models ranged from 0.57 for a tool predicting ADHD diagnosis in an external validation sample to 0.99 for a machine learning model predicting foster care permanency.
CONCLUSIONS
Although some tools have recently been developed for child and adolescent mental health for prognosis and child maltreatment, none can be currently recommended for clinical practice due to a combination of methodological limitations and poor model performance. New work needs to use ensure sufficient sample sizes, representative samples, and testing of model calibration.
PubMed: 37431439
DOI: 10.1002/jcv2.12034 -
Emerging Microbes & Infections Dec 2021There is great interest in understanding the viral genomic predictors of phenotypic traits that allow influenza A viruses to adapt to or become more virulent in...
BACKGROUND
There is great interest in understanding the viral genomic predictors of phenotypic traits that allow influenza A viruses to adapt to or become more virulent in different hosts. Machine learning techniques have demonstrated promise in addressing this critical need for other pathogens because the underlying algorithms are especially well equipped to uncover complex patterns in large datasets and produce generalizable predictions for new data. As the body of research where these techniques are applied for influenza A virus phenotype prediction continues to grow, it is useful to consider the strengths and weaknesses of these approaches to understand what has prevented these models from seeing widespread use by surveillance laboratories and to identify gaps that are underexplored with this technology.
METHODS AND RESULTS
We present a systematic review of English literature published through 15 April 2021 of studies employing machine learning methods to generate predictions of influenza A virus phenotypes from genomic or proteomic input. Forty-nine studies were included in this review, spanning the topics of host discrimination, human adaptability, subtype and clade assignment, pandemic lineage assignment, characteristics of infection, and antiviral drug resistance.
CONCLUSIONS
Our findings suggest that biases in model design and a dearth of wet laboratory follow-up may explain why these models often go underused. We, therefore, offer guidance to overcome these limitations, aid in improving predictive models of previously studied influenza A virus phenotypes, and extend those models to unexplored phenotypes in the ultimate pursuit of tools to enable the characterization of virus isolates across surveillance laboratories.
Topics: Algorithms; Drug Resistance, Viral; Genome, Viral; Genotype; Humans; Influenza A virus; Machine Learning; Phenotype
PubMed: 34498543
DOI: 10.1080/22221751.2021.1978824 -
Applied Clinical Informatics Aug 2021The change in performance of machine learning models over time as a result of temporal dataset shift is a barrier to machine learning-derived models facilitating...
OBJECTIVE
The change in performance of machine learning models over time as a result of temporal dataset shift is a barrier to machine learning-derived models facilitating decision-making in clinical practice. Our aim was to describe technical procedures used to preserve the performance of machine learning models in the presence of temporal dataset shifts.
METHODS
Studies were included if they were fully published articles that used machine learning and implemented a procedure to mitigate the effects of temporal dataset shift in a clinical setting. We described how dataset shift was measured, the procedures used to preserve model performance, and their effects.
RESULTS
Of 4,457 potentially relevant publications identified, 15 were included. The impact of temporal dataset shift was primarily quantified using changes, usually deterioration, in calibration or discrimination. Calibration deterioration was more common ( = 11) than discrimination deterioration ( = 3). Mitigation strategies were categorized as model level or feature level. Model-level approaches ( = 15) were more common than feature-level approaches ( = 2), with the most common approaches being model refitting ( = 12), probability calibration ( = 7), model updating ( = 6), and model selection ( = 6). In general, all mitigation strategies were successful at preserving calibration but not uniformly successful in preserving discrimination.
CONCLUSION
There was limited research in preserving the performance of machine learning models in the presence of temporal dataset shift in clinical medicine. Future research could focus on the impact of dataset shift on clinical decision making, benchmark the mitigation strategies on a wider range of datasets and tasks, and identify optimal strategies for specific settings.
Topics: Clinical Decision-Making; Clinical Medicine; Cognition; Machine Learning
PubMed: 34470057
DOI: 10.1055/s-0041-1735184 -
Neuroscience and Biobehavioral Reviews Oct 2021Laboratory experiments using fear conditioning and extinction protocols help lay the groundwork for designing, testing, and optimizing innovative treatments for... (Review)
Review
Laboratory experiments using fear conditioning and extinction protocols help lay the groundwork for designing, testing, and optimizing innovative treatments for anxiety-related disorders. Yet, there is limited basic research on fear conditioning and extinction in obsessive-compulsive disorder (OCD). This is surprising because exposure-based treatments based on associative learning principles are among the most popular and effective treatment options for OCD. Here, we systematically review and critically assess existing aversive conditioning and extinction studies of OCD. Across 12 studies, there was moderate evidence that OCD is associated with abnormal acquisition of conditioned responses that differ from comparison groups. There was relatively stronger evidence of OCD's association with impaired extinction processes. This included multiple studies finding elevated conditioned responses during extinction learning and poorer threat/safety discrimination during recall, although a minority of studies yielded results inconsistent with this conclusion. Overall, the conditioning model holds value for OCD research, but more work is necessary to clarify emerging patterns of results and increase clinical translational utility to the level seen in other anxiety-related disorders. We detail limitations in the literature and suggest next steps, including modeling OCD with more complex conditioning methodology (e.g., semantic/conceptual generalization, avoidance) and improving individual-differences assessment with dimensional techniques.
Topics: Conditioning, Classical; Conditioning, Psychological; Extinction, Psychological; Fear; Humans; Obsessive-Compulsive Disorder
PubMed: 34314751
DOI: 10.1016/j.neubiorev.2021.07.026 -
Revista de Neurologia Jul 2021Different variables, such as repetition and cognitive load, may explain the neurophysiological differences observed from one task to another in motor learning. This...
INTRODUCTION
Different variables, such as repetition and cognitive load, may explain the neurophysiological differences observed from one task to another in motor learning. This learning can be measured with functional magnetic resonance imaging.
AIM
The aim of this systematic review was to document motor learning by functional magnetic resonance imaging during the performance of different simple or complex motor tasks in healthy subjects.
MATERIAL AND METHODS
The search for articles was carried out in the MEDLINE, PEDro, CINHAL and EBSCO databases in May 2020. The systematic review followed the PRISMA criteria.
RESULTS
Nine studies were selected for a qualitative analysis. The quality of the studies ranged from 5 to 7 points on the PEDro scale. The qualitative analysis shows strong evidence that after repeating a motor task a motor learning process is generated. There is both strong and moderate evidence to show that action observation and sleep restriction are involved in motor learning. The results on sensory discrimination training were controversial.
CONCLUSIONS
The results show, with high quality evidence, that repetition of a motor task is associated with the learning process, which seems to be related to a thickening of the motor cortex after the intervention measured with functional magnetic resonance imaging. These results are not conclusive, owing to the limiting factors of this systematic review.
Topics: Brain Mapping; Humans; Learning; Magnetic Resonance Imaging; Motor Cortex; Motor Skills; Psychomotor Performance
PubMed: 34170004
DOI: 10.33588/rn.7301.2020657 -
Acta Orthopaedica Aug 2021Background and purpose - External validation of machine learning (ML) prediction models is an essential step before clinical application. We assessed the proportion,...
Background and purpose - External validation of machine learning (ML) prediction models is an essential step before clinical application. We assessed the proportion, performance, and transparent reporting of externally validated ML prediction models in orthopedic surgery, using the Transparent Reporting for Individual Prognosis or Diagnosis (TRIPOD) guidelines.Material and methods - We performed a systematic search using synonyms for every orthopedic specialty, ML, and external validation. The proportion was determined by using 59 ML prediction models with only internal validation in orthopedic surgical outcome published up until June 18, 2020, previously identified by our group. Model performance was evaluated using discrimination, calibration, and decision-curve analysis. The TRIPOD guidelines assessed transparent reporting.Results - We included 18 studies externally validating 10 different ML prediction models of the 59 available ML models after screening 4,682 studies. All external validations identified in this review retained good discrimination. Other key performance measures were provided in only 3 studies, rendering overall performance evaluation difficult. The overall median TRIPOD completeness was 61% (IQR 43-89), with 6 items being reported in less than 4/18 of the studies.Interpretation - Most current predictive ML models are not externally validated. The 18 available external validation studies were characterized by incomplete reporting of performance measures, limiting a transparent examination of model performance. Further prospective studies are needed to validate or refute the myriad of predictive ML models in orthopedics while adhering to existing guidelines. This ensures clinicians can take full advantage of validated and clinically implementable ML decision tools.
Topics: Decision Support Techniques; Humans; Machine Learning; Models, Statistical; Orthopedic Procedures; Treatment Outcome; Validation Studies as Topic
PubMed: 33870837
DOI: 10.1080/17453674.2021.1910448