Background: While artificial intelligence (AI) offers possibilities of advanced clinical prediction and decision-making in healthcare, models trained on relatively homogeneous datasets, and populations poorly-representative of underlying diversity, limits generalisability and risks biased AI-based decisions. Here, we describe the landscape of AI in clinical medicine to delineate population and data-source disparities.Methods: We performed a scoping review of clinical papers published in PubMed in 2019 using AI techniques. We assessed differences in dataset country source, clinical specialty, and author nationality, sex, and expertise. A manually tagged subsample of PubMed articles was used to train a model, leveraging transfer-learning techniques (building upon an existing BioBERT model) to predict eligibility for inclusion (original, human, clinical AI literature). Of all eligible articles, database country source and clinical specialty were manually labelled. A BioBERT-based model predicted first/last author expertise. Author nationality was determined using corresponding affiliated institution information using Entrez Direct. And first/last author sex was evaluated using the Gendarize.io API.Results: Our search yielded 30,576 articles, of which 7,314 (23.9%) were eligible for further analysis. Most databases came from the US (40.8%) and China (13.7%). Radiology was the most represented clinical specialty (40.4%), followed by pathology (9.1%). Authors were primarily from either China (24.0%) or the US (18.4%). First and last authors were predominately data experts (i.e., statisticians) (59.6% and 53.9% respectively) rather than clinicians. And the majority of first/last authors were male (74.1%).Interpretation: U.S. and Chinese datasets and authors were disproportionately overrepresented in clinical AI, and almost all of the top 10 databases and author nationalities were from high income countries (HICs). AI techniques were most commonly employed for image-rich specialties, and authors were predominantly male, with non-clinical backgrounds. Development of technological infrastructure in data-poor regions, and diligence in external validation and model re-calibration prior to clinical implementation in the short-term, are crucial in ensuring clinical AI is meaningful for broader populations, and to avoid perpetuating global health inequity.

PURPOSE: Aggregation of Asian Americans (AAs) with Native Hawaiians and Other Pacific Islanders (NHPIs) masks significant health disparities. We evaluated overall survival (OS) and surgery-to-radiation intervals (STRIs) among AA and NHPI women with early-stage breast cancer. METHODS: This National Cancer Database study included women with stage 0-II breast cancer diagnosed between 2004 and 2017. STRI was defined as days from surgery to radiation. Patients were stratified by adjuvant treatment. AAs were disaggregated into geographically relevant subpopulations: East, South, and Southeast Asians. Kaplan-Meier estimates and log-rank tests assessed survival. Cox proportional hazard and linear regression were adjusted for clinical and sociodemographic factors. RESULTS: In total, 578,927 women were included (median age 61 years, median follow-up 65 months, and 10-year OS 83%). AA and NHPI 10-year OS was 91% overall; subpopulation 10-year OS was 92% for East Asian, 90% for South Asian, 90% for Southeast Asian, and 83% for NHPI. On multivariable analysis, compared with non-Hispanic White, NHPI women had worse survival (adjusted hazard ratio [aHR] = 1.38; 95% CI, 1.09 to 1.77); all AA subpopulations had improved survival: East Asian (aHR = 0.57; 95% CI, 0.48 to 0.69), South Asian (aHR = 0.66; 95% CI, 0.51 to 0.84), and Southeast Asian (aHR = 0.78; 95% CI, 0.65 to 0.94). The AA and NHPI median STRI for was 73 days overall; the disaggregated median STRI was 68 days for East Asian, 80 days for South Asian, 77 days for Southeast Asians, and 81 days for NHPI. On adjusted analysis, compared with non-Hispanic White, Southeast Asians and NHPI had longer STRI by 6.6 (95% CI, 4.3 to 8.9) and 10.0 (95% CI, 5.8 to 14) days, respectively. CONCLUSION: Breast cancer disparities exist among disaggregated AA and NHPI subpopulations. Data disaggregation insights may lead to interventions to overcome these disparities, such as optimizing time-to-treatment for select populations.

PURPOSE: We identified (1) differences in localized prostate cancer (PCa) risk group at presentation and (2) disparities in access to initial treatment for Asian American, Native Hawaiian, and Pacific Islander (AANHPI) men with PCa after controlling for sociodemographic factors. METHODS: We assessed all patients in the National Cancer Database with localized PCa with low-, intermediate-, and high-risk disease who identified as Thai, White, Asian Indian, Chinese, Vietnamese, Korean, Japanese, Filipino, Hawaiian, Pacific Islander, Laotian, Pakistani, Kampuchean, and Hmong. Multivariable logistic regression defined adjusted odds ratios (AORs) with 95% CI of (1) presenting at progressively higher risk group and (2) receiving treatment or active surveillance with intermediate- or high-risk disease, adjusting for sociodemographic and clinical factors. RESULTS: Among 980,889 men (median age 66 years), all AANHPI subgroups with the exception of Thai (AOR = 0.84 [95% CI, 0.58 to 1.21], P > .05), Asian Indian (AOR = 1.12 [95% CI, 1.00 to 1.25], P > .05), and Pakistani (AOR = 1.34 [95% CI, 0.98 to 1.83], P > .05) men had greater odds of presenting at a progressively higher PCa risk group compared with White patients (Chinese AOR = 1.18 [95% CI, 1.11 to 1.25], P < .001; Japanese AOR = 1.36 [95% CI, 1.26 to 1.47], P < .001; Filipino AOR = 1.37 [95% CI, 1.29 to 1.46], P < .001; Korean AOR = 1.32 [95% CI, 1.18 to 1.48], P < .001; Vietnamese AOR = 1.20 [95% CI, 1.07 to 1.35], P = .002; Laotian AOR = 1.60 [95% CI, 1.08 to 2.36], P = .018; Hmong AOR = 4.07 [95% CI, 1.54 to 10.81], P = .005; Kampuchean AOR = 1.55 [95% CI, 1.03 to 2.34], P = .036; Asian Indian or Pakistani AOR = 1.15 [95% CI, 1.07 to 1.24], P < .001; Native Hawaiians AOR = 1.58 [95% CI, 1.38 to 1.80], P < .001; and Pacific Islanders AOR = 1.58 [95% CI, 1.37 to 1.82], P < .001). Additionally, Japanese Americans (AOR = 1.46 [95% CI, 1.09 to 1.97], P = .013) were more likely to receive treatment compared with White patients. CONCLUSION: Our findings suggest that there are differences in PCa risk group at presentation by race or ethnicity among Asian American, Native Hawaiian, and Pacific Islander subgroups and that there exist disparities in treatment patterns. Although AANHPI are often studied as a homogenous group, heterogeneity upon subgroup disaggregation underscores the importance of further study to assess and address barriers to PCa care.

Since the start of the COVID-19 pandemic, Asian Americans have been subjected to rising overt discrimination and violent hate crimes, highlighting the health implications of racism toward Asian Americans. As Asian Americans are the only group for whom cancer is the leading cause of death, these manifestations of anti-Asian racism provoke the question of the impact of racism across the cancer continuum for Asian Americans. In this Commentary, we describe how the myth of the “model minority” overlooks the diversity of Asian Americans. Ignoring such diversity in sociocultural trends, immigration patterns, socioeconomic status, health behaviors, and barriers to care masks disparities in cancer risk, access to care, and outcomes across Asian American populations. We recommend cancer epidemiologists, population science researchers, and oncology providers direct attention toward: (i) studying the impacts of structural and personally mediated racism on cancer risk and outcomes; (ii) ensuring studies reflect the uniqueness of individual ethnic groups, including intersectionality, and uncover underlying disparities; and (iii) applying a critical race theory approach that considers the unique lived experiences of each group. A more nuanced understanding of cancer health disparities, and how drivers of these disparities are associated with race and differ across Asian American ethnicities, may elucidate means through which these disparities can be alleviated.