Clinical AI tools must convey predictive uncertainty for each individual patient

Article Statistics & Probability

Conformal prediction beyond exchangeability

Rina Foygel Barber et al.

ANNALS OF STATISTICS (2023)

Article Computer Science, Artificial Intelligence

Learning from data with structured missingness

Robin Mitra et al.

Summary: Gathering big datasets is crucial for machine learning in scientific areas, but missing data values pose challenges, especially when dealing with structured missingness. While tools exist to handle 'missing at random' data, structured missingness, where missing values show association or structure, presents new obstacles. This issue becomes increasingly important as machine learning studies aim to learn from larger volumes of heterogeneous data.

NATURE MACHINE INTELLIGENCE (2023)

添加到收藏夹

Review Infectious Diseases

Antibiotic prescribing in patients with COVID-19: rapid review and meta-analysis

Bradley J. Langford et al.

Summary: The study estimated that approximately three-quarters of patients with COVID-19 receive antibiotic treatment, which is significantly higher than the estimated prevalence of bacterial co-infection. This suggests a potential for high rates of unnecessary antibiotic use in patients with COVID-19.

CLINICAL MICROBIOLOGY AND INFECTION (2021)

添加到收藏夹

Article Ethics

Privacy and artificial intelligence: challenges for protecting health information in a new era

Blake Murdoch

Summary: The development of healthcare artificial intelligence is raising concerns about the protection of patient data in the hands of private entities. Additionally, the use of AI-driven methods also poses risks of privacy breaches. There is a need for appropriate regulations and safeguards to ensure data privacy and patient agency are maintained.

BMC MEDICAL ETHICS (2021)

添加到收藏夹

Letter Medicine, General & Internal

The Clinician and Dataset Shift in Artificial Intelligence

Samuel G. Finlayson et al.

NEW ENGLAND JOURNAL OF MEDICINE (2021)

添加到收藏夹

Review Medicine, General & Internal

Risk of bias in studies on prediction models developed using supervised machine learning techniques: systematic review

Constanza L. Andaur Navarro et al.

Summary: A systematic review was conducted to assess the methodological quality of prediction models developed using machine learning techniques in various medical specialties between January 1, 2018 and December 31, 2019. The review found that most studies exhibited poor methodological quality and high risk of bias, with factors such as small study size, inadequate handling of missing data, and improper overfitting assessment contributing to this bias. Efforts to improve the design, conduct, reporting, and validation of these studies are necessary to enhance the application of machine learning based prediction models in clinical practice.

BMJ-BRITISH MEDICAL JOURNAL (2021)

添加到收藏夹

Article Health Care Sciences & Services

Do as AI say: susceptibility in deployment of clinical decision-aids

Susanne Gaube et al.

Summary: This study found that radiologists rated diagnostic advice as lower quality when it appeared to come from an AI system, while less experienced physicians did not have this bias. Diagnostic accuracy significantly decreased when participants received inaccurate advice, regardless of the purported source being AI or human experts. Therefore, important considerations need to be made when deploying advice in clinical environments, whether it is coming from AI or non-AI sources.

NPJ DIGITAL MEDICINE (2021)

添加到收藏夹

Editorial Material Infectious Diseases