Search citation statements
Paper Sections
Citation Types
Year Published
Publication Types
Relationship
Authors
Journals
Importance. Discriminatory language in clinical documentation impacts patient care and reinforces systemic biases. Scalable tools to detect and mitigate this are needed. Objective. Determine utility of a frontier large language model (GPT-4) in identifying and categorizing biased language and evaluate its suggestions for debiasing. Design. Cross-sectional study analyzing emergency department (ED) notes from the Mount Sinai Health System (MSHS) and discharge notes from MIMIC-IV. Setting. MSHS, a large urban healthcare system, and MIMIC-IV, a public dataset. Participants. We randomly selected 50,000 ED medical and nursing notes from 230,967 MSHS 2023 adult ED visiting patients, and 500 randomly selected discharge notes from 145,915 patients in MIMIC-IV database. One note was selected for each unique patient. Main Outcomes and Measures. Primary measure was accuracy of detection and categorization (discrediting, stigmatizing/labeling, judgmental, and stereotyping) of bias compared to human review. Secondary measures were proportion of patients with any bias, differences in the prevalence of bias across demographic and socioeconomic subgroups, and provider ratings of effectiveness of GPT-4's debiasing language. Results. Bias was detected in 6.5% of MSHS and 7.4% of MIMIC-IV notes. Compared to manual review, GPT-4 had sensitivity of 95%, specificity of 86%, positive predictive value of 84% and negative predictive value of 96% for bias detection. Stigmatizing/labeling (3.4%), judgmental (3.2%), and discrediting (4.0%) biases were most prevalent. There was higher bias in Black patients (8.3%), transgender individuals (15.7% for trans-female, 16.7% for trans-male), and undomiciled individuals (27%). Patients with non-commercial insurance, particularly Medicaid, also had higher bias (8.9%). Higher bias was also seen in health-related characteristics like frequent healthcare utilization (21% for >100 visits) and substance use disorders (32.2%). Physician-authored notes showed higher bias than nursing notes (9.4% vs. 4.2%, p < 0.001). GPT-4's suggested revisions were rated highly effective by physicians, with an average improvement score of 9.6/10 in reducing bias. Conclusions and Relevance. A frontier LLM effectively identified biased language, without further training, showing utility as a scalable fairness tool. High bias prevalence linked to certain patient characteristics underscores the need for targeted interventions. Integrating AI to facilitate unbiased documentation could significantly impact clinical practice and health outcomes.
Importance. Discriminatory language in clinical documentation impacts patient care and reinforces systemic biases. Scalable tools to detect and mitigate this are needed. Objective. Determine utility of a frontier large language model (GPT-4) in identifying and categorizing biased language and evaluate its suggestions for debiasing. Design. Cross-sectional study analyzing emergency department (ED) notes from the Mount Sinai Health System (MSHS) and discharge notes from MIMIC-IV. Setting. MSHS, a large urban healthcare system, and MIMIC-IV, a public dataset. Participants. We randomly selected 50,000 ED medical and nursing notes from 230,967 MSHS 2023 adult ED visiting patients, and 500 randomly selected discharge notes from 145,915 patients in MIMIC-IV database. One note was selected for each unique patient. Main Outcomes and Measures. Primary measure was accuracy of detection and categorization (discrediting, stigmatizing/labeling, judgmental, and stereotyping) of bias compared to human review. Secondary measures were proportion of patients with any bias, differences in the prevalence of bias across demographic and socioeconomic subgroups, and provider ratings of effectiveness of GPT-4's debiasing language. Results. Bias was detected in 6.5% of MSHS and 7.4% of MIMIC-IV notes. Compared to manual review, GPT-4 had sensitivity of 95%, specificity of 86%, positive predictive value of 84% and negative predictive value of 96% for bias detection. Stigmatizing/labeling (3.4%), judgmental (3.2%), and discrediting (4.0%) biases were most prevalent. There was higher bias in Black patients (8.3%), transgender individuals (15.7% for trans-female, 16.7% for trans-male), and undomiciled individuals (27%). Patients with non-commercial insurance, particularly Medicaid, also had higher bias (8.9%). Higher bias was also seen in health-related characteristics like frequent healthcare utilization (21% for >100 visits) and substance use disorders (32.2%). Physician-authored notes showed higher bias than nursing notes (9.4% vs. 4.2%, p < 0.001). GPT-4's suggested revisions were rated highly effective by physicians, with an average improvement score of 9.6/10 in reducing bias. Conclusions and Relevance. A frontier LLM effectively identified biased language, without further training, showing utility as a scalable fairness tool. High bias prevalence linked to certain patient characteristics underscores the need for targeted interventions. Integrating AI to facilitate unbiased documentation could significantly impact clinical practice and health outcomes.
Bioethics increasingly recognizes the impact of discriminatory practices based on social categories such as race, gender, sexual orientation or ability on clinical practice. Accordingly, major bioethics associations have stressed that identifying and countering structural discrimination in clinical ethics consultations is a professional obligation of clinical ethics consultants. Yet, it is still unclear how clinical ethics consultants can fulfill this obligation. More specifically, clinical ethics needs both theoretical tools to analyze and practical strategies to address structural discrimination within clinical ethics consultations. Intersectionality, a concept developed in Black feminist scholarship, is increasingly considered in bioethical theory. It stresses how social structures and practices determine social positions of privilege and disadvantage in multiple, mutually co-constitutive systems of oppression. This article aims to investigate how intersectionality can contribute to addressing structural discrimination in clinical ethics consultations with a particular focus on mental healthcare. To this end, we critically review existing approaches for clinical ethics consultants to address structural racism in clinical ethics consultations and extend them by intersectional considerations. We argue that intersectionality is a suitable tool to address structural discrimination within clinical ethics consultations and show that it can be practically implemented in two complementary ways: 1) as an analytic approach and 2) as a critical practice.
Disabled people face many social problems in their lives, as outlined by the UN Convention on the Rights of Persons with Disabilities. These problems often increase when disabled people also belong to another marginalized identity. The first aim of this study was to report on the extent and what intersectionalities are mentioned in academic abstracts in conjunction with disabled people. Various intersectional concepts are used to discuss intersectionality-related issues. The second aim was to ascertain the use of intersectionality-based concepts to discuss the intersectionality of disabled people. The field of intersectional pedagogy emerged to discuss the teaching of intersectionality linked to various marginalized identities. The third aim was to ascertain the coverage of how to teach about the intersectionality of disabled people in the intersectional pedagogy-focused academic literature we covered. Ability judgments are a general cultural reality. Many ability judgment-based concepts have been developed within the disability rights movement, disability studies, and ability-based studies that could be used to discuss the impact of ability judgments on the intersectionality of disabled people and enrich the area of intersectional pedagogy. The fourth aim was to ascertain the use of ability judgment-based concepts to analyze the intersectionality of disabled people. To obtain data for the four aims, we performed a manifest coding and qualitative content analysis of abstracts obtained from SCOPUS, the 70 databases of EBSCO-HOST and Web of Science, and an online survey in which we ascertained the views of undergraduate students on social groups experiencing negative ability-based judgments. As to the 34,830 abstracts that contained the term “intersectionality”; the 259,501 abstracts that contained the phrase “intersection of”; and the 11,653 abstracts that contained the 35 intersectionality-based concepts, the numbers for these abstracts that also contained the disability terms we used for our analysis were 753, 2058, and 274 abstracts, respectively, so 2.16%, 0.79%, and 2.35%, indicating a low academic engagement with the intersectionality of disabled people. We found many different intersectionalities mentioned in conjunction with disabled people, but most were mentioned only once or twice, with the main ones mentioned being race and gender. The literature covered made little use of most of the 52 intersectionality-based concepts we looked at (35 identified before the study and 17 more identified during the analysis). The literature covered also did not link to the area of intersectional pedagogy. Of the 25 ability judgment-based concepts, only the term ableism was used. As to the surveys, most students saw many of the social groups experiencing negative ability judgments, suggesting that the ability judgment-based concepts might be a useful tool to discuss intersectional consequences of ability judgments, such as intersectional conflict. Our data might be useful for intersectionality studies, intersectional pedagogy, disability studies, ability-based studies, and other academic fields that engage with intersectionality or with disability issues. Our study might also be useful for academics covering various topics to engage with the intersectionality of disabled people as part of their inquiries.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
customersupport@researchsolutions.com
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
This site is protected by reCAPTCHA and the Google Privacy Policy and Terms of Service apply.
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.