The Alliance for Artificial Intelligence in Healthcare (AAIH)的动态

A Year of Journal Articles (341/365): Davoudi, Anahita, Sena Chae, Lauren Evans, Sridevi Sridharan, Jiyoun Song, Kathryn H. Bowles, Margaret V. McDonald, and Maxim Topaz. "Fairness gaps in Machine learning models for hospitalization and emergency department visit risk prediction in home healthcare patients with heart failure."?International Journal of Medical Informatics?(2024): 105534. Summary: Problem: Traditional risk prediction models might miss patients due to bias against certain subgroups. Goal: Assess fairness of ML models in predicting hospitalizations and ED visits for heart failure patients. Data: Used electronic health records (EHR) data from a large HHC agency (12,189 patients). Methods: Developed ML models (LightGBM and AutoGluon) using demographics, vital signs, comorbidities, etc. Evaluated fairness using metrics like Equal Opportunity, Predictive Parity, and Statistical Parity across subgroups (gender, race/ethnicity, socioeconomic status). Results: Significant disparities in model performance across subgroups. AutoGluon showed slightly better fairness metrics than LightGBM. Model favored some subgroups (e.g., Hispanic males with high socioeconomic status) in metrics like Equal Opportunity. Other subgroups (e.g., females from other races with low socioeconomic status) were disadvantaged. Discussion: Importance of considering fairness when using ML models in healthcare. Potential reasons for bias include data imbalances and algorithmic biases. Strategies to improve fairness: using diverse datasets, bias-aware algorithms, and involving stakeholders. Limitations: Data from a single HHC agency, potentially limiting generalizability. Not all aspects of fairness were captured (e.g., cultural differences). Conclusion: ML models for healthcare require ongoing monitoring to ensure fairness and avoid discrimination. Fairness is crucial for equitable healthcare delivery. I love when authors do this so I am direct quoting here: "What is this study added to our knowledge Fairness Assessment in HHC Settings:?This study is the first to comprehensively assess fairness metrics in ML-based risk prediction models for HF patients receiving HHC, focusing on detecting biases in predicting hospitalizations or ED visits. Nuanced Understanding of Model Performance:?The study reveals AutoGluon's slight fairness advantage over LightGBM across most patient subgroups, emphasizing the nuanced nature of model performance based on specific subgroup characteristics. Insights for Future Model Development:?Recommendations include employing calibration techniques to address performance disparities, integrating fairness metrics in early model development stages, and involving diverse stakeholders to ensure fairness and equity."

Yesterday, I gave a talk on "Mathematical modeling of obstetric unit closures: From the descriptive to the prescriptive" at the Harvard Center for Research on Computation and Society's workshop on Computing for Population Maternal and Child Health. The workshop aimed to bring together experts in different fields utilizing computational methods to improve public health related to maternal and child healthcare. There were talks and panel discussions from a great group of computer scientists, operations researchers, biomedical engineers, pediatricians, neonatologists, OB/GYNs, and epidemiologists who gathered to highlight innovation in this area and identify important problems that are opportunities for future research. The work I presented is motivated by the lack of access to risk-appropriate maternity services in the United States, which is thought to be a leading contributor to rural disparities in?maternal morbidity and mortality. Moreover, rural access to care is declining as rural hospitals close their obstetric units. I gave an overview of my group's work forecasting the potential impact of obstetric facility closures on geographic access and facility delivery volume using mathematical modeling. We're hoping these models can help inform actions that are proactive rather than reactive. This research is led by PhD candidate Meghan Meredith, and MS Analytics students Ayushi Goyal and Manikant Thatipalli are building interactive dashboards to communicate these findings. Here's a related pre-print if you're interested: https://lnkd.in/exvQBzbG

Research led by the University of Michigan shows that Black patients have significantly fewer medical tests in the emergency department than White patients, which could lead to biases in artificial intelligence (AI) programs based on this data. The study, published in PLOS Global Public Health, shows emergency department tests are 4.5% higher for White than for Black patients of the same age, sex, and with the same medical complaints and emergency department triage score. The researchers expressed concern that, if built using this data, AI-based tools used to guide clinician decision making, which are becoming increasingly more common, could help increase pre-existing testing biases and result in substandard care for Black patients. For example, they could assume that Black patients are less likely to get ill, when actually they are less likely to be tested and admitted to hospital, which is not necessarily the same thing.

?? Exciting News! Our latest study, published in the International Journal of Medical Informatics, addresses crucial fairness gaps in machine learning models predicting hospitalization and emergency department visits for home healthcare patients with heart failure. ?? Key Highlights: * First study assessing fairness and biases in ML risk prediction models in home healthcare settings for heart failure patients. * We identified significant disparities in model performance across demographic subgroups, highlighting an urgent need to address these biases to ensure equitable healthcare delivery. * This study advances ethical considerations in ML, promoting fairness and inclusivity in patient care. Proud of our team's hard work on this important research! #MachineLearning #HealthcareEquity #HomeHealthcare #AIinHealthcare VNS Health Columbia University Irving Medical Center The Data Science Institute at Columbia University Columbia University School of Nursing Jiyoun Song Sena Chae Margaret McDonald

A recent UK study -- the first part of a three-phase project aiming to reduce bias in #AI health prediction models trained on real-world patient data -- assessed the available detail of ethnicity data in NHS England, including across different types of ethnicity codes. The study analyzed how more than 61 million people in England identified their ethnicity in more than 250 different groups. Noting that NHS hospitals record patient data via 19 ethnicity codes, while GPs use the 489 globally recognized #SNOMEDCT ethnicity codes, the researchers say the lack of representative data can lead to biased models that ultimately produce incorrect health assessments and that better data from real-world settings can lead to better health outcomes. More here: https://lnkd.in/gHtv9s55

??Scientific paper: Classification Methods Based on Machine Learning for the Analysis of Fetal Health Data Abstract: The persistent battle to decrease childhood mortality serves as a commonly employed benchmark for gauging advancements in the field of medicine. Globally, the under-5 mortality rate stands at approximately 5 million, with a significant portion of these deaths being avoidable. Given the significance of this problem, Machine learning-based techniques have emerged as a prominent tool for assessing fetal health. In this work, we have analyzed the classification performance of various machine learning models for fetal health analysis. Classification performance of various machine learning models, such as support vector machine (SVM), random forest(RF), and attentive interpretable tabular learning (TabNet) have been assessed on fetal health. Moreover, dimensionality reduction techniques, such as Principal component analysis (PCA) and Linear discriminant analysis (LDA) have been implemented to obtain better classification performance with less number of features. A TabNet model on a fetal health dataset provides a classification accuracy of 94.36%. In general, this technology empowers doctors and healthcare experts to achieve precise fetal health classification and identify the most influential features in the process. Continued on ES/IODE ?? https://etcse.fr/hdQ ------- If you find this interesting, feel free to follow, comment and share. We need your help to enhance our visibility, so that our platform continues to serve you.

The University of Illinois Chicago, the University of Iowa, the University of Missouri, and Loyola University have initiated an innovative project to integrate AI with health data from various professionals such as nurses and therapists. With $10 million in funding from ARPA-H, this initiative aims to create unified, holistic datasets that will transform patient care. By utilizing data from diverse healthcare professionals, this AI-powered initiative aims to enhance outcomes for challenging scenarios, such as falls and neonatal care. Multidisciplinary insights will drive innovation and deliver more personalized care solutions. Read more here: https://lnkd.in/edWgV7ne #MedTech #AIinHealthcare #HealthInnovation #DataScience #PatientCare #HealthcareTransformation

Insightful read by the Orlando Sentinel on how Orlando Health and AdventHealth Central Florida are widely using Artificial Intelligence for administrative work and early detection of illnesses like pancreatic cancer and sepsis. It seems Central Florida's hospitals are embracing AI. While this offers exciting possibilities in reducing administrative burnout and delivering earlier diagnoses, it is important to navigate this new frontier with equal parts optimism and caution. Legal and ethical concerns around data privacy, bias, and potential harm to patients, must be carefully considered. What are your thoughts on AI in healthcare? Share in the comments!