Health equity assessment of machine learning performance (HEAL): a framework and dermatology AI model case study.

Background: Artificial intelligence (AI) has repeatedly been shown to encode historical inequities in healthcare. We aimed to develop a framework to quantitatively assess the performance equity of health AI technologies and to illustrate its utility via a case study. Methods: Here, we propose a methodology to assess whether health AI technologies prioritise performance for patient populations experiencing worse outcomes, that is complementary to existing fairness metrics. We developed the Health Equity Assessment of machine Learning performance (HEAL) framework designed to quantitatively assess the performance equity of health AI technologies via a four-step interdisciplinary process to understand and quantify domain-specific criteria, and the resulting HEAL metric. As an illustrative case study (analysis conducted between October 2022 and January 2023), we applied the HEAL framework to a dermatology AI model. A set of 5420 teledermatology cases (store-and-forward cases from patients of 20 years or older, submitted from primary care providers in the USA and skin cancer clinics in Australia), enriched for diversity in age, sex and race/ethnicity, was used to retrospectively evaluate the AI model's HEAL metric, defined as the likelihood that the AI model performs better for subpopulations with worse average health outcomes as compared to others. The likelihood that AI performance was anticorrelated to pre-existing health outcomes was estimated using bootstrap methods as the probability that the negated Spearman's rank correlation coefficient (i.e., "R") was greater than zero. Positive values of R suggest that subpopulations with poorer health outcomes have better AI model performance. Thus, the HEAL metric, defined as p (R >0), measures how likely the AI technology is to prioritise performance for subpopulations with worse average health outcomes as compared to others (presented as a percentage below). Health outcomes were quantified as disability-adjusted life years (DALYs) when grouping by sex and age, and years of life lost (YLLs) when grouping by race/ethnicity. AI performance was measured as top-3 agreement with the reference diagnosis from a panel of 3 dermatologists per case. Findings: Across all dermatologic conditions, the HEAL metric was 80.5% for prioritizing AI performance of racial/ethnic subpopulations based on YLLs, and 92.1% and 0.0% respectively for prioritizing AI performance of sex and age subpopulations based on DALYs. Certain dermatologic conditions were significantly associated with greater AI model performance compared to a reference category of less common conditions. For skin cancer conditions, the HEAL metric was 73.8% for prioritizing AI performance of age subpopulations based on DALYs. Interpretation: Analysis using the proposed HEAL framework showed that the dermatology AI model prioritised performance for race/ethnicity, sex (all conditions) and age (cancer conditions) subpopulations with respect to pre-existing health disparities. More work is needed to investigate ways of promoting equitable AI performance across age for non-cancer conditions and to better understand how AI models can contribute towards improving equity in health outcomes. Funding: Google LLC.

Audio-Recorded Discharge Instructions for Limited English Proficient Parents: A Pilot Study.

BACKGROUND: Parents with limited English proficiency (LEP) demonstrate lower comprehension of discharge instructions. A study was conducted to (1) determine the feasibility of providing a greeting card with language-specific, audio-recorded discharge instructions to LEP parents; (2) describe use of and satisfaction with the cards; and (3) evaluate card effect on instruction comfort with home care and comprehension. METHODS: LEP parents of children undergoing day surgery from April to September 2016 were eligible. Participants were randomized to usual discharge instructions, or usual instructions plus a three-minute card with language-specific audio instructions that could be replayed repeatedly. Parents were surveyed by telephone two to seven days postdischarge to assess card use and satisfaction, comfort with home care, and discharge instruction recall (medications, home care, follow-up, and return precautions). Parent-reported instructions were compared to instructions in the medical record; concordance was determined by two blinded reviewers. Due to difficulty achieving recruitment goals, analysis focused on feasibility and acceptability. RESULTS: Of 83 parents enrolled, 66 (79.5%) completed the follow-up survey. Most had not completed high school (61.0%) and spoke Spanish (89.2%). Parents reported high satisfaction with the card (4.5/5 for ease of use, helpfulness, and understandability). Ninety-four percent shared the card with others, and 45.2% reported listening > 5 times. Besides reviewing the care instructions generally, parents reported using the card to review medications and engage others in the child's care. CONCLUSION: Providing language-concordant, audio-recorded discharge instructions was feasible, and parents reported high satisfaction with and frequent use of the cards with multiple caregivers.

Children, Families, and Disparities: Pediatric Provisions in the Affordable Care Act.

The Affordable Care Act has caused and continues to cause sweeping changes throughout the health system in the United States. Poorly explained, complex, controversial, confusing, and subject to continuous legal and regulatory definition, the law stands as a hallmark piece of legislation that will change the health sector in America forever. This article summarizes the Affordable Care Act with a focus on children, families, and disparities. Also provided is the context of the current system of health care coverage in the United States.

Gender and generational influences on the pediatric workforce and practice.

In response to demographic and other trends that may affect the future of the field of pediatrics, the Federation of Pediatric Organizations formed 4 working groups to participate in a year's worth of research and discussion preliminary to a Visioning Summit focusing on pediatric practice, research, and training over the next 2 decades. This article, prepared by members of the Gender and Generations Working Group, summarizes findings relevant to the 2 broad categories of demographic trends represented in the name of the group and explores the interface of these trends with advances in technology and social media and the impact this is likely to have on the field of pediatrics. Available data suggest that the trends in the proportions of men and women entering pediatrics are similar to those over the past few decades and that changes in the overall ratio of men and women will not substantially affect pediatric practice. However, although women may be as likely to succeed in academic medicine and research, fewer women than men enter research, thereby potentially decreasing the number of pediatric researchers as the proportion of women increases. Complex generational differences affect both the workforce and interactions in the workplace. Differences between the 4 generational groups comprising the pediatric workforce are likely to result in an evolution of the role of the pediatrician, particularly as it relates to aspects of work-life balance and the use of technology and social media.

African-American parents' trust in their child's primary care provider.

OBJECTIVE: Patients' trust in their primary care providers has important implications in terms of health outcomes and, among minority patients, mitigating racial health disparities. This study aims to identify family, provider, and health care setting characteristics that predict African American parents' trust in their child's primary care provider and whether provider partnership-building communication style explains this association. METHODS: Data were collected via retrospective telephone interviews completed 2 weeks after a child's health care visit to 1 of 7 pediatric primary care clinics in Washington, DC (3 community health centers, 3 private practices, and 1 hospital-based clinic). Four hundred twenty-five self-identified African American parents of children 0 to 5 years of age participated. Parents completed several standard survey instruments about trust and provider communication style as well as demographic questionnaires about their family and their child's provider. RESULTS: A step-wise linear regression revealed significant independent effects of having a previous relationship with the provider and seeing a provider in a community health center (CHC) on higher trust. There was also evidence of mediation by provider communication style, suggesting that parents who take their child to a CHC report greater trust in their child's provider because they have higher perceptions of provider partnership building. CONCLUSIONS: African American parents' trust in their child's provider may be enhanced by continuity of care and greater use of a partnership-building communication style by providers.

African-American parents' perceptions of partnership with their child's primary care provider.

OBJECTIVE: To identify family, provider, and healthcare setting characteristics associated with African-American parents' perceptions of partnership with their child's primary care provider. STUDY DESIGN: Data were collected via a telephone survey of 425 African-American parents of 0- to 5-year-old children who had presented for a health visit 1 to 2 weeks earlier at participating pediatric primary care practices in Washington, DC. Parents' perceptions of the level of partnership building by their child's provider were assessed using the Street Provider Communication Style instrument. RESULTS: Multivariate logistic regression models indicated that, after adjusting for other family and provider/setting characteristics, parents seen in community health centers were more likely to report high partnership building compared with parents seen at private or hospital-based practices. Parents with at least a college education and those who described their child's provider's race as "other" were most likely to report moderate partnership building. CONCLUSIONS: Future studies should examine elements of care delivery at community health centers that may lead to better partnerships between parents and providers in private and hospital-based practice settings.

Beliefs about the appropriate age for initiating toilet training: are there racial and socioeconomic differences?

OBJECTIVE: To examine racial and socioeconomic differences in parental beliefs about the appropriate age at which to initiate toilet training. STUDY DESIGN: A cross-sectional survey of 779 parents visiting child health providers in 3 clinical sites in Washington, DC and the surrounding metropolitan area completed a self-report survey. The main outcome variable was parental beliefs about the appropriate age at which to initiate toilet training. Using multiple linear regression, differences in beliefs were assessed in relation to race, family income, parental education, parental age, and age of the oldest and youngest children. RESULTS: Among respondents, parents felt that the average age at which toilet training should be initiated was 20.6 months (+/-7.6 months), with a range of 6 to 48 months. Caucasian parents believed that toilet training should be initiated at a significantly later age (25.4 months) compared with both African-American parents (18.2 months) and parents of other races (19.4 months). In the multiple regression model, factors predicting belief in when to initiate toilet training were Caucasian race and higher income. CONCLUSIONS: Race and income were independent predictors of belief in age at which to initiate toilet training. More research is needed to determine what factors contribute to toilet training practices in diverse populations.

The use of AAP-recommended disciplinary practice guidelines among African American caregivers of children in Head Start programs.

OBJECTIVE: The American Academy of Pediatrics (AAP) recommends that parents not use harsh disciplinary practices. Previous studies have characterized the disciplinary practices of African American parents as harsh, with reliance on more aggressive techniques not currently recommended by the AAP. However, recent research has indicated more disciplinary practice diversity among African Americans. This study describes factors associated with the use of AAP-recommended disciplinary practices among lower-income African American caregivers of children in Head Start. METHODS: Subjects were caregivers of children at three Head Start sites. Participants were eligible for inclusion if the biological mother, biological father, or target child was identified as African American. Using consensus methods, responses to the Parental Discipline Methods Interview (PDMI) were described as consistent or inconsistent with AAP guidelines regarding use of negative disciplinary practices (e.g., spanking, yelling). Caregivers avoiding any of these inconsistent methods were referred to as "adherent." RESULTS: "Adherent" caregivers were older (32.5 years vs. 30.4 years) and had more education (86.0% vs. 75.4% high school graduates). They were also less likely to report that their child had behavioral problems (12.9% vs. 25.2%) or deficient social skills (1.7% vs. 8.0%). CONCLUSIONS: Lower-income African American caregivers were more likely to use disciplinary practices consistent with AAP guidelines if they had higher levels of education and were living in an urban setting. Caregivers describing their child as having fewer behavior problems, better social skills, or themselves as less stressed were also more likely to be "adherent."

Discipline in the African American community: the impact of socioeconomic status on beliefs and practices.

OBJECTIVE: To describe and compare disciplinary beliefs and practices among African American parents from diverse socioeconomic backgrounds. METHODS: A cross-sectional survey was conducted of self-identified African American parents of children <48 months of age at 2 ambulatory teaching clinics, 2 community health centers, and 3 private practices in Washington, DC, and the surrounding metropolitan area. Disciplinary beliefs and practices of African American parents were measured. RESULTS: A total of 175 of the 189 parents who were approached for the study completed the survey for a participation rate of 92.5%. Middle/upper socioeconomic status (SES) parents in this study were more likely to be married (60.9% vs 14.7%), older (31.4 years vs 25 years), and more educated (80% having attended at least some college vs 34.4%) than lower SES parents. There were no significant differences between middle/upper and lower SES parents with regard to their belief in a preferred disciplinary method (teaching, spanking, removing) or approach (positive, negative). Lower SES parents were more likely to endorse spanking a 1- to 3-year-old child if they were doing something that was not safe (90.5% vs 78.3%). Middle/upper SES parents were significantly more likely to reward their child for positive behavior than lower SES parents (66.1% vs 47.1%). CONCLUSIONS: Lower and middle/upper SES parents in this study population were reasonably similar with respect to disciplinary beliefs and practices. Exceptions to this generalization were that lower SES parents were more likely to endorse spanking as a response to an unsafe behavior on the part of the child, and middle/upper SES parents reported higher levels of reward for positive behavior.