The scale of the problem the team addressed is well documented. People from ethnic minority backgrounds frequently experience worse outcomes and poorer experiences than the general population when they arrive at an emergency department. Previous strategies to close this gap have included strengthening self-management of health issues, improving communication between patients and carers on one side and emergency department staff on the other, and building workforce capacity. Yet the researchers note that details about such interventions are scant, and, crucially, intervention design has typically proceeded without involving the patients and carers whose lives those interventions are meant to improve. That omission is what Experience-Based Co-Design, the methodology at the heart of this study, is intended to correct.

Experience-Based Co-Design is a participatory approach that treats patients, carers, and clinicians as equal experts in the experience of care. Rather than asking people to respond to pre-formed proposals, it gathers their lived experiences, extracts needs and ideas from that material, and then brings all parties together to prioritise and shape concrete solutions. In this study, the team conducted five workshops with a total of 25 patients and carers from ethnic minority backgrounds, alongside five clinical staff from the emergency departments of the three participating hospitals serving the area. The workshops formed a deliberate pipeline: early sessions generated raw material, middle sessions refined and prioritised it, and the final session produced finished intervention concepts.

The yield from those early workshops was striking. The first two sessions alone gathered 87 distinct needs and 162 ideas from patients and carers for improving their experience in the emergency department. That volume reflects the density of friction points that people encounter when they arrive acutely unwell, often in a second or third language, in an unfamiliar and intimidating institutional environment. Importantly, the needs and ideas were not confined to communication and health systems literacy, the study’s original focus. They also addressed the physical environment of the department and patients’ emotional wellbeing, suggesting that the burden of a difficult emergency department visit is experienced holistically rather than as a single technical problem to be solved.

After the refinement and prioritisation that took place in the third and fourth workshops, the fifth session converged on four concrete interventions. The first is a waiting room digital display, designed to give people waiting for care better information about what is happening and what to expect. The second is a bilingual and visual pain scale, a tool that speaks directly to one of the most fundamental barriers in emergency medicine: if a clinician cannot accurately assess how much pain a patient is in, every downstream decision, from triage category to analgesia, is compromised. A bilingual, visually anchored scale allows patients to communicate pain intensity across language boundaries, reducing the risk that linguistic disadvantage translates into clinical disadvantage.

The third intervention is an informational video about emergency department processes, aimed at demystifying the sequence of events, triage, waiting, assessment, and treatment, that patients often find opaque and anxiety-provoking. The fourth is an information kiosk, a physical point in the department where patients and carers can access guidance. Taken together, the four interventions form a coherent package: they target the waiting period, the clinical encounter, and the general orientation to the system, the three moments where the researchers’ evidence suggested communication and literacy gaps do the most harm.

One of the study’s most sobering findings concerns the relationship between public feedback and institutional memory. Some of the ideas presented by ethnic minority participants were already in place in the departments, or had been tried in the past, without the participants knowing. This is not a trivial observation. It indicates that health services can implement improvements and then lose visibility of them, so that the public continues to experience the gap the improvement was meant to close. The researchers conclude from this that a regular feedback loop between health services and the public is essential, both to improve awareness of what already exists and to keep interventions current as circumstances change. Co-design, in other words, is not a one-off event but a continuous relationship.

The study also offers a candid technical assessment of the limits of innovation in a complex system. Several ideas were inspired by participants’ experiences in other sectors, such as retail or hospitality, where customer-facing information systems work smoothly. The researchers found that these ideas were not easily translated into the emergency department context because of system constraints and the complexity of the environment. An emergency department is not a shop: it operates under clinical governance, legacy information technology, fluctuating demand, and safety requirements that resist simple analogies. This finding matters for anyone designing patient-facing tools in healthcare, because it suggests that the feasibility filter must be applied early, with people who understand the system’s architecture, rather than after concepts have already been embraced by participants.

From that assessment flows one of the study’s clearest recommendations for future practice. Involving emergency department staff from other regions, and support services professionals such as information technology specialists, in the co-design workshops could increase innovative thinking and enable the selection of ideas that are more easily integrated into current systems. The logic is that a wider professional mix at the table would allow the group to distinguish between ideas that are imaginative but unimplementable and ideas that are both imaginative and compatible with the department’s technical and operational reality. It is a reminder that co-design’s promise of inclusivity extends beyond patients and clinicians to the often-invisible professionals who maintain the systems those interventions must run on.

The researchers are equally honest about the social dynamics inside the workshops themselves. Managing knowledge and power imbalance within co-design groups, they report, can prove challenging. When patients and carers sit alongside clinicians and researchers, differences in professional status, technical vocabulary, and confidence can quietly shape whose ideas get airtime and whose get refined. Acknowledging that difficulty, rather than assuming that bringing people together automatically equalises them, is itself a contribution to the methodology. The study was approved by the Local Health District Human Research Ethics Committee, with informed consent obtained from all participants, and it forms part of a broader MyED partnership that includes Western Sydney Local Health District, the NSW Agency for Clinical Innovation, the University of the Sunshine Coast, the University of New South Wales, the Department of Social Services, the NDIS Quality and Safeguards Commission, and Health Consumers NSW. For health services everywhere grappling with ethnic disparities in emergency care, the message is that the tools exist to be built, but they must be built with, not merely for, the communities they serve, and the building must never stop.

Subject of Research: Co-design of emergency department interventions with ethnic minority patients and staff to improve communication and health system literacy

Article Title: Experienced-based co-design interventions in the emergency department for ethnic minority adults

Article References: Goirand, M., Francis-Auton, E., Cheek, C., Hayba, N., Richardson, L., Ransolin, N., Dominello, A., Austin, E., Murphy, M., Mitchell, R., Carrigan, A., Harrison, R., Braithwaite, J., & Clay-Williams, R. (2026). Experienced-based co-design interventions in the emergency department for ethnic minority adults. BMC Health Services Research. https://doi.org/10.1186/s12913-026-15822-2

Image Credits: AI Generated

DOI: 10.1186/s12913-026-15822-2

Keywords: emergency department, co-design, ethnic minority, health equity, health communication, health system literacy, patient experience, Experience-Based Co-Design, health services research, bilingual pain scale, health disparity, Australia

The study, published in the journal Cancer Causes & Control, drew on records from the North Carolina Central Cancer Registry covering the years 2003 through 2021. The investigators constructed two analytic cohorts: a diagnosis cohort comprising 111 non-Hispanic American Indian and Alaska Native patients and 7,685 non-Hispanic white patients, and a treatment cohort of 97 American Indian and Alaska Native patients and 6,040 white patients for whom treatment information was available. By restricting the analysis to North Carolina and comparing the two groups directly, the team aimed to provide the first baseline assessment of the gastric cancer burden borne by American Indian communities in the state, a population that has historically been undercounted and understudied in cancer surveillance research.

The demographic and clinical profile of gastric cancer differed markedly between the two groups. American Indian and Alaska Native patients were diagnosed at a median age of 65 years compared with 68 years for white patients, a difference that was statistically significant. The anatomical distribution of their tumors also diverged: 80 percent of American Indian and Alaska Native patients had non-cardia gastric cancer, tumors arising in the lower portions of the stomach, whereas the corresponding figure among white patients was 62 percent. This distinction matters clinically because cardia and non-cardia gastric cancers have different risk factor profiles, different patterns of occurrence across populations, and different associations with infection by the bacterium Helicobacter pylori, a well-established cause of gastric carcinoma.

Stage at diagnosis also differed, with 36 percent of American Indian and Alaska Native patients diagnosed at a later stage compared with 32 percent of white patients, a modest but statistically significant difference. The two groups differed just as sharply in where they lived. Fully 76 percent of American Indian and Alaska Native patients resided in areas with low residential population density, compared with only 26 percent of white patients. This rural concentration is more than a descriptive footnote; rural residence is associated with longer travel distances to specialized cancer centers, reduced access to gastroenterologists and oncologists, and documented delays in cancer diagnosis and treatment across many malignancies. Any explanation of the treatment gaps observed in this study must contend with the geography of the affected communities.

To quantify the treatment disparity, the researchers analyzed the risk of receiving no documented treatment at all and the timing of the first treatment after diagnosis. Using modified Poisson regression, a statistical technique well suited to modeling common binary outcomes while estimating risk ratios directly, they estimated the multivariable-adjusted association between race and failure to initiate treatment. The adjusted risk ratio was 1.13, with a 95 percent confidence interval spanning 0.87 to 1.47 and a p-value of 0.34. In other words, American Indian and Alaska Native patients had a point estimate suggesting a thirteen percent higher risk of receiving no treatment, but the wide confidence interval means the data cannot rule out effects ranging from a modest reduction to a nearly fifty percent increase in risk. The authors were careful to present this result as inconclusive rather than as evidence of no difference.

The timing analysis told a more consistent story. The team estimated race-specific cumulative incidence functions for treatment initiation, a method that accounts for the possibility that some patients die before starting treatment, and refined these estimates using inverse probability weighting to adjust for key covariates such as age, stage, and socioeconomic characteristics of the patients’ residential areas. Confidence intervals were obtained through bootstrapping, a resampling approach that makes minimal assumptions about the underlying distributions. The cumulative incidence of treatment initiation was consistently lower for American Indian and Alaska Native patients throughout follow-up, and at thirty days after diagnosis the difference was statistically significant, with a cumulative incidence difference of minus 0.09 and a 95 percent confidence interval of minus 0.17 to minus 0.01. Nine percentage points fewer American Indian and Alaska Native patients had begun treatment within the first month.

Why does a thirty-day delay matter? In gastric cancer, as in many solid tumors, time to first treatment is a widely used quality-of-care indicator, and delays are associated with disease progression, reduced eligibility for curative surgery, and worse survival. A nine-percentage-point gap in the first month, sustained across the first year after diagnosis, suggests a systematic difference in how quickly patients move from diagnosis to oncologic care, whether because of barriers in referral pathways, insurance and authorization hurdles, transportation and lodging challenges in rural areas, patient-provider communication, or differences in tumor biology and comorbid illness that complicate treatment planning. The study was not designed to isolate which of these mechanisms dominates, but the consistency of the timing gap, in contrast to the imprecise estimate for receiving no treatment at all, points toward delays rather than outright withholding of care as the central disparity.

The findings fit into a broader national picture. Previous research has documented elevated gastric cancer incidence and mortality among American Indian and Alaska Native populations in the United States, with particularly high rates in Alaska and certain regions of the country, and prior work has linked these patterns to higher prevalence of Helicobacter pylori infection and its virulent strains in some Native communities. The North Carolina study extends this literature by focusing on a state with a large, diverse Native population that includes both federally and non-federally recognized tribes, and by adding the dimension of treatment timeliness, which had been examined nationally but not previously at the state level for this population. The authors also situate the work within a series of recent North Carolina analyses of cancer burden and end-of-life care quality among American Indian patients conducted with the Cancer Information and Population Health Resource at the UNC Lineberger Comprehensive Cancer Center.

The researchers acknowledge important limitations inherent to registry-based studies of Native populations. Racial misclassification is a persistent problem in state cancer data; earlier work in North Carolina has shown that American Indian race is sometimes misrecorded among members of non-federally recognized tribes, which could bias incidence and outcome estimates in either direction. Registry treatment information may also be incomplete, particularly for care delivered outside reporting facilities, and the modest number of American Indian and Alaska Native patients, 111 in the diagnosis cohort, limits statistical power and explains the wide confidence interval around the risk ratio for no treatment. The investigators addressed the treatment-data concern by noting prior validation work on registry-based treatment information and by using documented receipt of treatment as the outcome, while the small-sample problem was handled through robust variance estimation and bootstrapped intervals rather than ignored.

For public health practitioners and oncologists in North Carolina and beyond, the study delivers a clear message: gastric cancer in American Indian and Alaska Native communities presents differently, is caught later, and reaches treatment more slowly than in the white population, and these differences are measurable with the surveillance systems already in place. The authors frame their results as a baseline against which future interventions can be judged, whether those interventions target earlier detection of non-cardia tumors in younger Native patients, streamline the diagnostic-to-treatment pathway in rural counties, or expand culturally grounded patient navigation. Given that gastric cancer is often preventable through detection and eradication of Helicobacter pylori and that survival depends heavily on stage at diagnosis, closing the thirty-day treatment gap identified here could translate into concrete survival gains for a population that has carried a disproportionate share of this disease for far too long.

Subject of Research: Racial disparities in gastric cancer diagnosis and time-to-first-treatment between non-Hispanic American Indian Alaska Native and white patients in North Carolina

Article Title: Characterizing gastric cancer and time-to-first treatment among non-Hispanic American Indian Alaska Native and white patients in North Carolina

Article References: Uhrin, M. T., Jackson, B. E., Baggett, C. D., Spees, L. P., Begay, J. A., Epplein, M., Wheeler, S. B., Bell, R. A., & Emerson, M. A. (2026). Characterizing gastric cancer and time-to-first treatment among non-Hispanic American Indian Alaska Native and white patients in North Carolina. Cancer Causes & Control, 37(10), Article 173. https://doi.org/10.1007/s10552-026-02254-w

Image Credits: AI Generated

DOI: 10.1007/s10552-026-02254-w

Keywords: gastric cancer, American Indian, Alaska Native, health disparity, time to treatment, North Carolina, cancer registry, Helicobacter pylori, non-cardia gastric cancer, rural health, cancer epidemiology, Cancer Causes & Control

The research, led by investigators examining hospital-level adoption patterns across the United States, maps the diffusion of AI and robotic technologies through the American hospital system and connects those patterns to longstanding measures of access inequality. Rather than treating innovation as a rising tide that lifts all boats, the study treats each hospital’s adoption decision as the outcome of financial capacity, workforce readiness, regulatory environment, and patient demand. When those variables are mapped geographically, a stark pattern emerges: adoption clusters in large, urban, teaching-affiliated hospitals with high operating margins, while rural and safety-net institutions lag dramatically behind.

The technical logic behind this clustering is straightforward, and the authors unpack it in detail. Deploying a clinical machine learning model is not simply a matter of purchasing software. Hospitals must maintain the digital infrastructure to feed models with clean, standardized electronic health record data; they need data science personnel to validate, calibrate, and monitor algorithms over time; and they require the regulatory and governance frameworks to manage model drift, bias audits, and liability. Robotic surgical platforms add capital costs that can exceed two million dollars per system, plus recurring maintenance contracts and the need for surgeons trained on high procedural volumes. Each of these requirements scales with hospital size and revenue, giving well-resourced institutions a compounding advantage.

The study’s analysis of access inequality draws on the demographic and socioeconomic characteristics of the communities served by adopting and non-adopting hospitals. Patients in regions with early, intensive adoption tend to be wealthier, more likely to hold private insurance, and more likely to live in metropolitan counties with dense specialist networks. By contrast, rural hospitals, which serve roughly one in five Americans, frequently operate on thin or negative margins and cannot justify the capital expenditure or recruit the technical staff that AI-driven medicine demands. The result is a two-tier landscape in which the benefits of predictive analytics, automated diagnostics, and robot-assisted intervention accrue disproportionately to populations that already enjoy superior health outcomes.

What makes the finding more consequential is the mechanism by which early adoption generates future advantage. AI systems improve with data, and hospitals that deploy them early accumulate larger, better-labeled clinical datasets, refine their workflows sooner, and build institutional expertise that late adopters cannot easily replicate. Surgical outcomes for robot-assisted procedures are known to improve with surgeon and team experience, meaning hospitals with early robotic programs simultaneously achieve better results and attract more patients, further increasing volume and revenue. The authors characterize this as a potential cumulative-advantage dynamic, in which technological gaps do not merely persist but widen over time, the healthcare analogue of the winner-take-all economics seen in other data-driven industries.

The study also documents disparities in the types of technology being adopted. General administrative AI, such as scheduling optimization and billing automation, has diffused relatively broadly because its returns are immediate and its technical demands modest. Clinical AI, including diagnostic imaging support and risk prediction models, shows a much steeper socioeconomic gradient. Robotic surgical systems show the steepest gradient of all, concentrated overwhelmingly in high-volume urban centers. This stratification matters because clinical and surgical technologies are where the direct health benefits lie; administrative automation may improve a hospital’s finances without improving a single patient’s outcome.

Policy implications flow directly from the analysis. The authors point out that federal incentive programs, including the multibillion-dollar push toward electronic health records in the 2010s, succeeded partly because they tied payments to adoption, effectively subsidizing the transition. No comparable mechanism currently exists for clinical AI and robotics. Without deliberate intervention, market forces alone will continue to route innovation toward institutions that can afford it, a pattern the study suggests could entrench existing inequalities in mortality, disease detection, and surgical access. Potential remedies discussed include targeted grants and loan programs for rural and safety-net hospitals, shared-service models in which regional networks pool AI infrastructure, and reimbursement structures that reward outcomes rather than technology ownership.

The research also adds a cautionary note to the national conversation about AI in medicine, much of which has focused on algorithmic bias within individual models. A biased model deployed at a single hospital can harm that hospital’s patients, but a deployment gap between hospitals harms entire populations by denying them access to the technology at all. The study frames this second form of inequity, which the authors analyze at the system level rather than the algorithm level, as underexamined in the literature. Fairness auditing of individual models, the work implies, is necessary but not sufficient if the models themselves never reach the communities that need them most.

For clinicians, hospital administrators, and policymakers, the message of the study is that the window for shaping equitable adoption is now. Technological diffusion patterns harden as standards settle, vendor markets mature, and training pipelines consolidate around early adopters. The United States has already lived through versions of this story with MRI machines, positron emission tomography, and minimally invasive surgical platforms, each of which arrived in wealthy urban institutions years before reaching rural America. Whether AI and robotics follow the same trajectory or bend toward broader access depends on choices being made today, in state legislatures, federal agencies, and the boardrooms of hospital systems deciding where their next million-dollar investment will go. The evidence assembled here makes clear that leaving those choices to the market alone carries a predictable cost, and that the cost will be paid by the patients with the least capacity to bear it.

Subject of Research: Adoption of artificial intelligence and robotics in United States hospitals and its relationship to healthcare access inequality

Article Title: Hospital AI and robotics adoption and access inequality in the United States

Article References: Johnson, A., Gefen, D., & Harrison, T. D. (2026). Hospital AI and robotics adoption and access inequality in the United States. Scientific Reports. https://doi.org/10.1038/s41598-026-70027-1

Image Credits: AI Generated

DOI: 10.1038/s41598-026-70027-1

Keywords: hospital AI, medical robotics, healthcare inequality, health disparity, rural hospitals, health policy, clinical AI, surgical robots, digital health, healthcare access, technology diffusion, United States healthcare