AMIA Joint Summits on Translational Science proceedings. AMIA Joint Summits on Translational Science最新文献

Clinical predictive models that include race as a predictor have the potential to exacerbate disparities in healthcare. Such models can be respecified to exclude race or optimized to reduce racial bias. We investigated the impact of such respecifications in a predictive model - UTICalc - which was designed to reduce catheterizations in young children with suspected urinary tract infections. To reduce racial bias, race was removed from the UTICalc logistic regression model and replaced with two new features. We compared the two versions of UTICalc using fairness and predictive performance metrics to understand the effects on racial bias. In addition, we derived three new models for UTICalc to specifically improve racial fairness. Our results show that, as predicted by previously described impossibility results, fairness cannot be simultaneously improved on all fairness metrics, and model respecification may improve racial fairness but decrease overall predictive performance.

Varying case definitions of COPD have heterogenous genetic risk profiles, potentially reflective of disease subtypes or classification bias (e.g., smokers more likely to be diagnosed with COPD). To better understand differences in genetic loci associated with ICD-defined versus spirometry-defined COPD we contrasted their GWAS results with those for heavy smoking among 337,138 UK Biobank participants. Overlapping risk loci were found in/near the genes ZEB2, FAM136B, CHRNA3, and CHRNA4, with the CHRNA3 locus shared across all three traits. Mediation analysis to estimate the effects of lead genotyped variants mediated by smoking found significant indirect effects for the FAM136B, CHRNA3, and CHRNA4 loci for both COPD definitions. Adjustment for mediator-outcome confounders modestly attenuated indirect effects, though in the CHRNA4 locus for spirometry-defined COPD the proportion mediated increased an additional 8.47%. Our results suggest that differences between ICD-defined and spirometry-defined COPD associated genetic loci are not a result of smoking biasing classification.

Machine learning classification problems are widespread in bioinformatics, but the technical knowledge required to perform model training, optimization, and inference can prevent researchers from utilizing this technology. This article presents an automated tool for machine learning classification problems to simplify the process of training models and producing results while providing informative visualizations and insights into the data. This tool supports both binary and multiclass classification problems, and it provides access to a variety of models and methods. Synthetic data can be generated within the interface to fill missing values, balance class labels, or generate entirely new datasets. It also provides support for feature evaluation and generates explainability scores to indicate which features influence the output the most. We present CLASSify, an open-source tool for simplifying the user experience of solving classification problems without the need for knowledge of machine learning.

While modelling and simulation are powerful techniques for exploring complex phenomena, if they are not coupled with suitable real-world data any results obtained are likely to require extensive validation. We consider this problem in the context of search game modelling, and suggest that both demographic and behaviour data are used to configure certain model parameters. We show this integration in practice by using a combined dataset of over 150,000 individuals to configure a specific search game model that captures the environment, population, interventions and individual behaviours relating to winter health service pressures. The presence of this data enables us to more accurately explore the potential impact of service pressure interventions, which we do across 33,000 simulations using a computational version of the model. We find government advice to be the best-performing intervention in simulation, in respect of improved health, reduced health inequalities, and thus reduced pressure on health service utilisation.

Obstructive sleep apnea is a sleep disorder that is linked with many health complications and severe form of apnea can even be lethal. Overnight polysomnography is the gold standard for diagnosing apnea, which is expensive, time-consuming, and requires manual analysis by a sleep expert. Recently, there have been numerous studies demonstrating the application of artificial intelligence to detect apnea in real time. But the majority of these studies apply data pre-processing and feature extraction techniques resulting in a longer inference time that makes the real-time detection system inefficient. This study proposes a single convolutional neural network architecture that can automatically extract spatial features and detect apnea from both electrocardiogram (ECG) and blood-oxygen saturation (SpO₂) signals. Using segments of 10s, the network classified apnea with an accuracy of 94.2% and 96% for ECG and SpO₂ respectively. Moreover, the overall performance of both models was consistent with an AUC score of 0.99.

Fairness is crucial in machine learning to prevent bias based on sensitive attributes in classifier predictions. However, the pursuit of strict fairness often sacrifices accuracy, particularly when significant prevalence disparities exist among groups, making classifiers less practical. For example, Alzheimer's disease (AD) is more prevalent in women than men, making equal treatment inequitable for females. Accounting for prevalence ratios among groups is essential for fair decision-making. In this paper, we introduce prior knowledge for fairness, which incorporates prevalence ratio information into the fairness constraint within the Empirical Risk Minimization (ERM) framework. We develop the Prior-knowledge-guided Fair ERM (PFERM) framework, aiming to minimize expected risk within a specified function class while adhering to a prior-knowledge-guided fairness constraint. This approach strikes a flexible balance between accuracy and fairness. Empirical results confirm its effectiveness in preserving fairness without compromising accuracy.

Automatic HIV phenotyping is needed for HIV research based on electronic health records (EHRs). MIMIC-IV, an extension of MIMIC-III, contains more than 520,000 hospital admissions and has become a valuable EHR database for secondary medical research. However, there was no prior phenotyping algorithm to extract HIV cases from MIMIC-IV, which requires a comprehensive knowledge of the database. Moreover, previous HIV phenotyping algorithms did not consider the new HIV-1/HIV-2 antibody differentiation immunoassay tests that MIMIC-IV contains. Our work provided insight into the structure and data elements in MIMIC-IV and proposed a new HIV phenotyping algorithm to fill in these gaps. The results included MIMIC-IV's data tables and elements used, 1,781 and 1,843 HIV cases from MIMIC-IV's versions 0.4 and 2.1, respectively, and summary statistics of these two HIV case cohorts. They could be used for the development of statistical and machine learning models in future studies about the disease.

In the rapidly evolving field of healthcare, the integration of artificial intelligence (AI) has become a pivotal component in the automation of clinical workflows, ushering in a new era of efficiency and accuracy. This study focuses on the transformative capabilities of the fine-tuned KoELECTRA model in comparison to the GPT-4 model, aiming to facilitate automated information extraction from thyroid operation narratives. The current research landscape is dominated by traditional methods heavily reliant on regular expressions, which often face challenges in processing free-style text formats containing critical details of operation records, including frozen biopsy reports. Addressing this, the study leverages advanced natural language processing (NLP) techniques to foster a paradigm shift towards more sophisticated data processing systems. Through this comparative study, we aspire to unveil a more streamlined, precise, and efficient approach to document processing in the healthcare domain, potentially revolutionizing the way medical data is handled and analyzed.

ChatGPT is a popular information system (over 1 billion visits in August 2023) that can generate natural language responses to user queries. It is important to study the quality and equity of its responses on health-related topics, such as vaccination, as they may influence public health decision-making. We use the Vaccine Hesitancy Scale (VHS) proposed by Shapiro et al.¹ to measure the hesitancy of ChatGPT responses in English, Spanish, and French. We find that: (a) ChatGPT responses indicate less hesitancy than those reported for human respondents in past literature; (b) ChatGPT responses vary significantly across languages, with English responses being the most hesitant on average and Spanish being the least; (c) ChatGPT responses are largely consistent across different model parameters but show some variations across the scale factors (vaccine competency, risk). Results have implications for researchers interested in evaluating and improving the quality and equity of health-related web information.