Bias In Health Care Research Articles

Terez Shea-Donohue: Optimism helps, and confidence in your work is critical.

Terez Shea-Donohue is the program director of the Division of Digestive Diseases and Nutrition at the National Institute of Diabetes and Digestive and Kidney Diseases of the National Institutes of Health. As a program director, Terez supports basic and translational research related to neurogastroenterology, gastrointestinal (GI), and GI epithelial barrier function. We spoke to Terez about the transition from active research to a predominantly administrative job, the need for life-long mentorship, and the continued sex/gender bias in health care.

Exploring Bias in Health Care: Using Art to Facilitate a Narrative Medicine Approach among Third-Year Medical Students.

Teaching medical students about biases in health care poses challenges due to the complexity of modifying preexisting beliefs and the possibility of triggering strong emotional reactions. Narrative medicine fosters reflection and discussion among healthcare providers, and visual thinking strategies (VTS) enable providers to improve diagnostic ability through close and methodical examination of art. This study investigated the effectiveness of a session integrating narrative medicine and VTS to enhance third-year medical students' understanding of bias in health care. A 90-minute "Narrative Medicine Rounds" session was implemented in the Family Medicine clerkship for all third-year medical students in the Florida International University Herbert Wertheim College of Medicine class of 2024. The session included a faculty-facilitated discussion of Robert Colescott's painting Emergency Room, an overview of narrative medicine principles, and a sharing of student narratives related to bias experienced in the healthcare setting. After the session, students wrote and submitted their narrative essays. Students also completed an anonymous, optional, computer-based evaluation with Likert-type and free-response questions. Likert questions were analyzed using descriptive statistics. For the short-answer questions, teams of two researchers (D.G. and S.E.S., E.E. and M.S.) analyzed responses for themes and subthemes, met to reach consensus on a codebook through discussion, and recoded data as needed until frequencies of themes were mutually determined. Of 101 students, 87 (86%) provided responses to the Likert questions. Seventy percent of respondents agreed that Robert Colescott's Emergency Room was a useful tool for discussing bias in medicine, whereas 25% were neutral. Fifty-one percent of respondents agreed that the session enhanced their understanding of bias in patient care, with 33% being neutral. Forty-seven percent of respondents agreed that the session made them more familiar with narrative medicine as a tool that they can use in patient care, with 28% being neutral. Students identified five themes for recognizing bias: empathy (16%), self-reflection (60%), active listening (17%), communication (13%), and education (7%). When addressing bias, students outlined five strategies: drawing from personal experiences (1%), self-reflection (64%), communication (29%), education (15%), and advocacy (26%). Combining narrative medicine and VTS in a single session offers a promising approach for discussing bias in healthcare education. Through reflective experiences and art interpretation, students demonstrated an awareness of often interconnected strategies to identify and mitigate bias while caring for patients. Future next steps for this study involve exploring longitudinal impacts and integrating narrative medicine strategies throughout the medical school curriculum.

Applying a community-engaged participatory machine learning model.

Although predictive algorithms have been described as the definitive solution to bias in health care, machine learning techniques may also propagate existing health inequities within the community context. However, there may be ways in which machine learning techniques can help community psychologists, public health researchers and practitioners identify patterns in data in a way that empowers improved outcomes. Incorporating community insight in all stages of machine learning research mitigates bias by positioning members of underrepresented communities as the experts of their lived experiences. As community psychologists already prioritize community-based participatory practices, we propose three core guiding principles for a community-engaged participatory model for research using machine learning techniques: shared decision-making, reflexivity and structural humility, and flexibility and adaptability. Guided by these three principles, we emphasize grounding priority setting, problem formation, model assumptions, and interpretation of the resulting algorithmic patterns in the truths born from the lived experiences of people closest to the problem. We also suggest opportunities for bidirectional and mutually empowering partnerships between algorithmic scientists and the communities to which their algorithms will be applied. Inclusion of community stakeholders in all stages of machine learning for health research provides an opportunity to develop algorithms that are both highly effective and ethically grounded in the lived experiences of target populations.

Addressing AI Algorithmic Bias in Health Care

This Viewpoint discusses the bias that exists in artificial intelligence (AI) algorithms used in health care despite recent federal rules to prohibit discriminatory outcomes from AI and recommends ways in which health care facilities, AI developers, and regulators could share responsibilities and actions to address bias.

Reporting of health equity considerations in vaccine trials for COVID-19: a methodological review

Background and ObjectivesAn emerging body of randomized controlled trials (RCTs) on COVID-19 vaccines has served as the evidence base for public health decision-making. While it is recommended that RCTs report results by health equity stratifiers to reduce bias in health care and gaps in research, it is unknown whether this was done in COVID-19 vaccine trials. To critically examine the use of health equity stratifiers in COVID-19 vaccine trials. Study Design and SettingWe conducted a methodological review of published COVID-19 vaccine trials available in the COVID-19 living Network Meta-Analysis systematic review database through February 8, 2023. Based on the PROGRESS-Plus framework, we examined the following health equity stratifiers: place of residence, race/ethnicity, occupation, gender/sex, religion, education, socio-economic status, social capital, age, disability, features of relationships, and temporary situations. We assessed each study in duplicate according to three criteria for comprehensive health-equity reporting: 1) describing participants, 2) reporting equity-relevant results, and 3) discussing equity-relevant implications of trial findings. ResultsWe reviewed 144 trial manuscripts. The most frequently used PROGRESS-Plus stratifiers to describe participants were age (100%), place of residence (100%), gender/sex (99%), and race/ethnicity (64%). Age was most often used to disaggregate or adjust results (67%), followed by gender or sex (35%). Discussions of equity-relevant implications often indicated limited generalizability of results concerning age (40% of studies). Half (47%) of the studies considered at least one health equity stratifier for all three criteria. No trials included stratifiers related to religion, socioeconomic status, sexual orientation, or features of relationships. ConclusionCOVID-19 vaccine trials provided a limited description of health equity stratifiers as defined by PROGRESS-Plus and infrequently disaggregated results or discussed the study implications as they related to health equity. Considering the health disparities exacerbated during the pandemic, increased uptake of PROGRESS-Plus in RCTs would support a more nuanced understanding of health disparities and better inform actions to improve health equity.

Review on clinician bias and its impact on racial and socioeconomic disparities in pediatric heart transplantation.

This expert review seeks to highlight implicit bias in health care, transplant medicine, and pediatric heart transplantation to focus attention on the role these biases may play in the racial/ethnic and socioeconomic disparities noted in pediatric heart transplantation. This review breaks down the transplant decision making process to highlight points at which implicit bias may affect outcomes and discuss how the science of human decision making may help understand these complex processes.

Assessing the potential of GPT-4 to perpetuate racial and gender biases in health care: a model evaluation study

Large language models (LLMs) such as GPT-4 hold great promise as transformative tools in health care, ranging from automating administrative tasks to augmenting clinical decision making. However, these models also pose a danger of perpetuating biases and delivering incorrect medical diagnoses, which can have a direct, harmful impact on medical care. We aimed to assess whether GPT-4 encodes racial and gender biases that impact its use in health care. Using the Azure OpenAI application interface, this model evaluation study tested whether GPT-4 encodes racial and gender biases and examined the impact of such biases on four potential applications of LLMs in the clinical domain-namely, medical education, diagnostic reasoning, clinical plan generation, and subjective patient assessment. We conducted experiments with prompts designed to resemble typical use of GPT-4 within clinical and medical education applications. We used clinical vignettes from NEJM Healer and from published research on implicit bias in health care. GPT-4 estimates of the demographic distribution of medical conditions were compared with true US prevalence estimates. Differential diagnosis and treatment planning were evaluated across demographic groups using standard statistical tests for significance between groups. We found that GPT-4 did not appropriately model the demographic diversity of medical conditions, consistently producing clinical vignettes that stereotype demographic presentations. The differential diagnoses created by GPT-4 for standardised clinical vignettes were more likely to include diagnoses that stereotype certain races, ethnicities, and genders. Assessment and plans created by the model showed significant association between demographic attributes and recommendations for more expensive procedures as well as differences in patient perception. Our findings highlight the urgent need for comprehensive and transparent bias assessments of LLM tools such as GPT-4 for intended use cases before they are integrated into clinical care. We discuss the potential sources of these biases and potential mitigation strategies before clinical implementation. Priscilla Chan and Mark Zuckerberg.

A Qualitative Study of Racial, Ethnic, and Cultural Experiences of Minority Clinicians During Agitation Care in the Emergency Department

Racial and ethnic bias in health care has been documented at structural, organizational, and clinical levels, impacting emergency care, including agitation management in the emergency department (ED). Little is known about the experiences of racial and ethnic minority ED clinicians caring for racial and ethnic minority groups, especially during their agitated state. The objective of this study was to explore the lived experiences of racial and ethnic minority ED clinicians who have treated patients with agitation in the ED. We performed semistructured individual interviews of Black, Latino, and multiracial clinicians who worked at 1 of 3 EDs from an urban quaternary care medical center in the Northeast United States between August 2020 and June 2022. We performed thematic analysis through open coding of initial transcripts and identifying additional codes through sequential iterative rounds of group discussion. Once the codebook was finalized and applied to all transcripts, the team identified key themes and subthemes. Of the 27 participants interviewed, 14 (52%) identified as Black, 9 (33%) identified as Hispanic/Latino, and 4 (15%) identified as multiracial and/or other race and ethnicity. Three primary themes emerged from racial and ethnic minority clinician experiences of managing agitation: witness of perceived bias during clinical interactions with patients of color who bear racialized presumptions of agitation, moral injury and added workload to address perceived biased agitation management practices while facing discrimination in the workplace, and natural advocacy and allyship for agitated patients of color based on a shared identity and life experience. Our study found that through their shared minority status, racial and ethnic minority clinicians had a unique vantage point to observe perceived bias in the management of agitation in minority patients. Although they faced added challenges as racial and ethnic minority clinicians, their allyship offered potential mitigation strategies for addressing disparities in caring for an underserved and historically marginalized patient population.

Reducing Unconscious Bias in Health Care: Role of Nursing Education

No abstract available

Gender differences in horizontal strabismus: Systematic review and meta-analysis shows no difference in prevalence, but gender bias towards females in the clinic.

Strabismus is a misalignment of the visual axis that affects 2-3% of the population and can lead to loss of binocular vision. It is currently controversial whether there is a gender difference in the most common form of visual misalignment: horizontal strabismus. Some studies claimed that more females than males have an outward deviation (exotropia), while others concluded that there is no significant gender difference. No previous work has systematically explored gender differences in horizontal strabismus or has compared the results of population-based studies with those of clinic-based studies. We conducted a systematic review and meta-analysis of studies reporting the prevalence of horizontal strabismus. We included 73 population-based studies and compared their disclosed gender population with that in 141 comparable clinical-based studies. We analysed the data according to gender, strabismus type (esotropia, exotropia), and geographic region/ethnicity. Summary statistics showed a nearly identical prevalence of horizontal strabismus (2.558% for males, 2.582% for females), esotropia (1.386% males vs. 1.377% females), and of exotropia (1.035% males vs. 1.043% females). Meta-analysis results showed that these differences between males and females were not statistically significant (odds ratio (OR) = 1.01; 95% confidence interval (CI) = 0.97-1.10), but that females were significantly more frequent (by 7.50%) in clinic-based studies than males, with 5.00% more females for esotropia, and 12.20% more females for exotropia when adjusted for the population's sex ratio. The extent of the female gender bias differed between geographic regions/societies, with Asians having the lowest bias towards females and Latin American countries having the strongest bias. Males and females have the same prevalence of horizontal strabismus, including exotropia. Females with strabismus seek health care or are brought to clinics significantly more often than males. This is an example of gender bias in health care in favour of females rather than males, apparently because parents - erroneously fearing only cosmetic consequences - are more concerned about strabismus in their daughters than their sons. Societal attitudes towards females, as well as economic factors (insurance status), appear to be relevant factors that determine the magnitude of the gender bias in horizontal strabismus.

Algorithmic bias in artificial intelligence is a problem—And the root issue is power

BackgroundArtificial intelligence (AI) in health care continues to expand at a rapid rate, impacting both nurses and communities we accompany in care. PurposeWe argue algorithmic bias is but a symptom of a more systemic and longstanding problem: power imbalances related to the creation, development, and use of health care technologies. MethodsThis commentary responds to Drs. O’Connor and Booth’s 2022 article, “Algorithmic bias in health care: Opportunities for nurses to improve equality in the age of artificial intelligence.” DiscussionNurses need not ‘reinvent the wheel’ when it comes to AI policy, curricula, or ethics. We can and should follow the lead of communities already working ‘from the margins’ who provide ample guidance. ConclusionIts neither feasible nor just to expect individual nurses to counter systemic injustice in health care through individual actions, more technocentric curricula, or industry partnerships. We need disciplinary supports for collective action to renegotiate power for AI tech.

Weight Bias: A Narrative Review of the Evidence, Assumptions, Assessment, and Recommendations for Weight Bias in Health Care.

Weight bias is common in health care and needs to be addressed to provide equal and equitable care for people of all body sizes. This review aims to (a) describe the impact of weight bias on providers and patients, (b) identify the limitations of body mass index and suggest alternative screenings to determine health risk, (c) provide a review of self-assessment of weight bias, and (d) present evidence of an alternative weight-inclusive approach to health care. EBSCO Host, Google Scholar, and Pubmed were used to find peer reviewed articles. Keywords included the following: weight bias, weight stigma, weight inclusive, and weight discrimination. Projected weight bias is prevalent among health care providers and affects patient visit times, care provided, and communication given. Patients who experience weight bias are more likely to experience stress and depression, which impact chronic disease risk. Body mass index was also found to have numerous limitations as a metric for identifying accurately an individual's cardiometabolic risks, whereas, waist-to-height ratio is a more accurate indicator. There are also numerous screening tools that are available to self-assess for weight bias although more studies are needed to standardize results. An alternative approach to combat weight bias is a behavior-focused approach on health as opposed to weight-focused. Weight bias is a critical topic that health professionals need to consider to be aware of and begin to change as it has a strong impact on the public's health.

Artificial Intelligence Bias in Health Care: Web-Based Survey.

Resources are increasingly spent on artificial intelligence (AI) solutions for medical applications aiming to improve diagnosis, treatment, and prevention of diseases. While the need for transparency and reduction of bias in data and algorithm development has been addressed in past studies, little is known about the knowledge and perception of bias among AI developers. This study's objective was to survey AI specialists in health care to investigate developers' perceptions of bias in AI algorithms for health care applications and their awareness and use of preventative measures. A web-based survey was provided in both German and English language, comprising a maximum of 41 questions using branching logic within the REDCap web application. Only the results of participants with experience in the field of medical AI applications and complete questionnaires were included for analysis. Demographic data, technical expertise, and perceptions of fairness, as well as knowledge of biases in AI, were analyzed, and variations among gender, age, and work environment were assessed. A total of 151 AI specialists completed the web-based survey. The median age was 30 (IQR 26-39) years, and 67% (101/151) of respondents were male. One-third rated their AI development projects as fair (47/151, 31%) or moderately fair (51/151, 34%), 12% (18/151) reported their AI to be barely fair, and 1% (2/151) not fair at all. One participant identifying as diverse rated AI developments as barely fair, and among the 2 undefined gender participants, AI developments were rated as barely fair or moderately fair, respectively. Reasons for biases selected by respondents were lack of fair data (90/132, 68%), guidelines or recommendations (65/132, 49%), or knowledge (60/132, 45%). Half of the respondents worked with image data (83/151, 55%) from 1 center only (76/151, 50%), and 35% (53/151) worked with national data exclusively. This study shows that the perception of biases in AI overall is moderately fair. Gender minorities did not once rate their AI development as fair or very fair. Therefore, further studies need to focus on minorities and women and their perceptions of AI. The results highlight the need to strengthen knowledge about bias in AI and provide guidelines on preventing biases in AI health care applications.

Perspectives of Community Partners Involved in an Academic Training to Address Clinicians' Implicit Bias.

Community-academic partnerships are increasingly used in interventions to address health care disparities. Little is known about motivations and perceptions of participating community members. To elicit community members' perspectives of involvement in a community-academic partnership to address implicit bias in health care. With our partnering community organizer, we conducted one-on-one semistructured interviews and a follow-up group interview with participating community members to solicit experiences about involvement in an National Institutes of Health-funded clinician training; responses were organized using content analysis. Community members revealed that their participation was motivated by trust in our community organizer; they derived personal pride from participation in clinician training; the power differential between community members and clinicians in the training environment needed to be levelled. Our community organizer noted that the benefits of community-academic partnerships propagate to the larger community via community members' experiences. Community members note trust, pride, and power as important elements in community-academic partnership.

Evaluation of Nursing Coursework for Biased Content.

Bias-free content in nursing education prepares students to provide equitable care. The process of assessing content promotes the dismantling of systemic bias in health care, advancing social justice, diversity, equity, and inclusion. Unfortunately, there are no published studies to guide the evaluation of nursing curricula for the presence of bias. This innovative project aimed to identify bias in a course and develop a structure to guide the wider evaluation of the curriculum to identify and remove biased content. A modified version of the Upstate Bias Checklist was applied to a 15-week, three-semester hour, prelicensure, graduate-level nursing course.

Impact of a Comprehensive ST‐Segment–Elevation Myocardial Infarction Protocol on Key Process Metrics in Black Americans

Impact of a Comprehensive ST‐Segment–Elevation Myocardial Infarction Protocol on Key Process Metrics in Black Americans

Pharmacogenomics: Driving Personalized Medicine.

Personalized medicine tailors therapies, disease prevention, and health maintenance to the individual, with pharmacogenomics serving as a key tool to improve outcomes and prevent adverse effects. Advances in genomics have transformed pharmacogenetics, traditionally focused on single gene-drug pairs, into pharmacogenomics, encompassing all "-omics" fields (e.g., proteomics, transcriptomics, metabolomics, and metagenomics). This review summarizes basic genomics principles relevant to translation into therapies, assessing pharmacogenomics' central role in converging diverse elements of personalized medicine. We discuss genetic variations in pharmacogenes (drug-metabolizing enzymes, drug transporters, and receptors), their clinical relevance as biomarkers, and the legacy of decades of research in pharmacogenetics. All types of therapies, including proteins, nucleic acids, viruses, cells, genes, and irradiation, can benefit from genomics, expanding the role of pharmacogenomics across medicine. Food and Drug Administration approvals of personalized therapeutics involving biomarkers increase rapidly, demonstrating the growing impact of pharmacogenomics. A beacon for all therapeutic approaches, molecularly targeted cancer therapies highlight trends in drug discovery and clinical applications. To account for human complexity, multicomponent biomarker panels encompassing genetic, personal, and environmental factors can guide diagnosis and therapies, increasingly involving artificial intelligence to cope with extreme data complexities. However, clinical application encounters substantial hurdles, such as unknown validity across ethnic groups, underlying bias in health care, and real-world validation. This review address the underlying science and technologies germane to pharmacogenomics and personalized medicine, integrated with economic, ethical, and regulatory issues, providing insights into the current status and future direction of health care. SIGNIFICANCE STATEMENT: Personalized medicine aims to optimize health care for the individual patients with use of predictive biomarkers to improve outcomes and prevent adverse effects. Pharmacogenomics drives biomarker discovery and guides the development of targeted therapeutics. This review addresses basic principles and current trends in pharmacogenomics, with large-scale data repositories accelerating medical advances. The impact of pharmacogenomics is discussed, along with hurdles impeding broad clinical implementation, in the context of clinical care, ethics, economics, and regulatory affairs.

Algorithmic bias in health care: Opportunities for nurses to improve equality in the age of artificial intelligence

Artificial Intelligence (AI) consists of a range of sophisticated computational techniques, encompassing machine learning algorithms and natural language processing among others, that are lauded as a way to improve clinical decision making, patient care, and health service delivery. A recent systematic review of AI in nursing and midwifery found many clinical, managerial, and educational applications of these predictive algorithms over the last 20 years covering areas such as wound care, critical care, falls, infection control, emergency care, older adult care, and education among others (O'Connor et al., 2022).

Intersectionality Impacts Survivorship: Identity-Informed Recommendations to Improve the Quality of Life of African American Breast Cancer Survivors in Health Promotion Programming.

(1) Background: African American women breast cancer survivors face unique experiences that impact their quality of life as they transition beyond treatments. Experiences may be complicated by living at the intersection of systemically oppressed identities, including gender, race, social class, and cancer-related disability. Using the Black Feminist Thought (BFT) framework and the PEN-3 cultural model, this qualitative study sought to: (a) understand African American women breast cancer survivors’ lived experiences; (b) examine how the multiple intersecting factors of race, gender, social class/socioeconomic status, and cancer-related disability impact their quality of life; and (c) inform future health promotion programming that is culturally relevant to AAWBCS to improve their quality of life. (2) Methods: Seven focus groups were conducted with 30 African American breast cancer survivors in a Midwestern metropolitan region. Focus groups were audiotaped and transcribed verbatim. Framework analyses were conducted to identify themes with NVivo qualitative analysis software. (3) Results: Four themes emerged: (a) caregiving roles provide both support and challenges for survivors, (b) the “strong Black woman” is inherent in survivor experiences, (c) intersectionality impacts survivorship, and (d) African American women resist oppression through culturally specific supports and advocacy. (4) Conclusions: The intervention point of entry should be at the peer support group level and centered on family and provide community-based support and services. Future research should move upstream to address social determinants of health, including racism, sexism, and ableism; there is a critical need to discuss how structural racism affects health care and develop interventions to address racial discrimination and racial bias in health care.

Health Care Bias and Discrimination Experienced by Lesbian, Gay, Bisexual, Transgender, and Queer Parents of Children With Developmental Disabilities: A Qualitative Inquiry in the United States

This study explored the impact of health care (HC) bias and discrimination on lesbian, gay, bisexual, transgender, and queer (LGBTQ) parents and their children with disabilities in the United States, including the timing of developmental screening and diagnosis. We conducted semistructured interviews with 16 LGBTQ parents of children with developmental concerns or disabilities recruited through a prior national survey. Interviews were transcribed and analyzed using a combined inductive and deductive approach. Discrimination types reported included noninclusive forms, disclosure challenges, and providers dismissing nongestational parents and diverse families. Few parents reported screening and diagnosis delays. Parents' recommendations included: avoiding assumptions, honoring family diversity, increasing LGBTQ family support, improving HC forms, increasing antibias training, and convening a learning community. Our study advances the knowledge around HC bias and discrimination among LGBTQ parents of children with disabilities. Findings highlight the need for increased LGBTQ-affirming family support and research representing LGBTQ family diversity in U.S. health care.