It is suggested that the choice of convenient, seemingly effective proxies for ground truth can be an important source of algorithmic bias in many contexts.
Abstract:
Health systems rely on commercial prediction algorithms to identify and help patients with complex health needs. We show that a widely used algorithm, typical of this industry-wide approach and affecting millions of patients, exhibits significant racial bias: At a given risk score, Black patients are considerably sicker than White patients, as evidenced by signs of uncontrolled illnesses. Remedying this disparity would increase the percentage of Black patients receiving additional help from 17.7 to 46.5%. The bias arises because the algorithm predicts health care costs rather than illness, but unequal access to care means that we spend less money caring for Black patients than for White patients. Thus, despite health care cost appearing to be an effective proxy for health by some measures of predictive accuracy, large racial biases arise. We suggest that the choice of convenient, seemingly effective proxies for ground truth can be an important source of algorithmic bias in many contexts.
TL;DR: WILDS is presented, a benchmark of in-the-wild distribution shifts spanning diverse data modalities and applications, and is hoped to encourage the development of general-purpose methods that are anchored to real-world distribution shifts and that work well across different applications and problem settings.
TL;DR: In this article, the authors provide a review of federated learning in the biomedical space, and summarize the general solutions to the statistical challenges, system challenges, and privacy issues in federated Learning, and point out the implications and potentials in healthcare.
TL;DR: There is a need to sensitize developers, healthcare professionals, and legislators to the challenges and limitations of opaque algorithms in medical AI and to foster multidisciplinary collaboration moving forward to ensure that medical AI lives up to its promises.
TL;DR: This work shows the need to explicitly account for underspecification in modeling pipelines that are intended for real-world deployment in any domain, and shows that this problem appears in a wide variety of practical ML pipelines.
TL;DR: Key findings from a 2-year weekly effort to track and share key developments in medical AI are discussed, including prospective studies and advances in medical image analysis, which have reduced the gap between research and deployment.
TL;DR: In this article, a systematic review and meta-analysis of large-scale blood pressure lowering trials, published between Jan 1, 1966, and July 7, 2015, was performed.
TL;DR: Reducing SES disparities in health will require policy initiatives addressing the components of socioeconomic status (income, education, and occupation) as well as the pathways by which these affect health.
TL;DR: This article showed that applying machine learning to ordinary human language results in human-like semantic biases and replicated a spectrum of known biases, as measured by the Implicit Association Test, using a widely used, purely statistical machine-learning model trained on a standard corpus of text from the World Wide Web.
TL;DR: The Charlson Index, the CIRS, the ICED and the Kaplan Index are valid and reliable methods to measure comorbidity that can be used in clinical research.
Q1. What have the authors contributed in "Dissecting racial bias in an algorithm used to manage the health of populations" ?
Gomez-Uribe et al. this paper proposed a uniform pricing in US retail chains, which was later validated by the National Bureau of Economic Research.
Q2. What are the future works mentioned in the paper "Dissecting racial bias in an algorithm used to manage the health of populations" ?
As a first step, the authors suggested using the existing model infrastructure— sample, predictors ( excluding race, as before ), training process, and so forth—but changing the label: Rather than future cost, they created an index variable that combined health prediction with cost prediction. Building on these results, the authors are establishing an ongoing ( unpaid ) collaboration to convert the results of Table 3 into a better, scaled predictor of multidimensional health measures, with the goal of rolling these improvements out in a future round of algorithm development. These results suggest that label biases are fixable.
Q3. What are the main reasons for the complexity of the aggregation process?
Health care costs, though well measured and readily available in insurance claims data, are also the result of a complex aggregation process with a number of distortions due to structural inequality, incentives, and inefficiency.
Q4. Why does the program operate to improve the management of these conditions?
Because the program ultimately operates to improve the management of these conditions, patients with the most encounters related to them could also be a promising group on which to deploy preventative interventions.
Q5. What are the four counterfactual simulations used to put these numbers in context?
The authors then perform four counterfactual simulations to put these numbers in context; naturally, these simulations use only observable factors, not the many unobserved administrative and human factors that also affect enrollment.
Q6. What are the main mechanisms by which poverty can lead to disparities in use of health care?
Although the population the authors study is entirely insured, there are many other mechanisms by which poverty can lead to disparities in use of health care: geography and differential access to transportation, competing demands from jobs or child care, or knowledge of reasons to seek care (29–31).
Q7. What is the common way to understand an algorithm?
Without an algorithm’s training data, objective function, and predictionmethodology, the authors can only guess as to the actual mechanisms for the important algorithmic disparities that arise.
Q8. What is the effect of race on health care?
whether it is communication, trust, or bias, something about the interactions of Black patients with the health care system itself leads to reduced use of health care.
Q9. What is the unusual aspect of the algorithm?
An unusual aspect of their dataset is that the authors observe the algorithm’s inputs and outputsas well as its objective function, providing us a unique window into the mechanisms by which bias arises.
Q10. How do the authors compare the results of the high-risk care management program?
For those enrolled in the high-risk care management program (1.3% of their sample), the authors first show the fraction of the population that is Black, as well asthe fraction of all costs and chronic conditions accounted for by these observations.