The scale of the underlying problem is global. A 2023 national survey in the United States found that 60 percent of Black adults and 42 percent of Asian adults reported preparing themselves for potentially unfair treatment before medical visits—psychological armor that patients should never need. Nurses, as the largest group of health professionals in continuous contact with patients, sit at the exact point where such inequities are either challenged or reproduced. Yet most research on the topic has relied on surveys measuring prevalence, leaving the how and why of discriminatory behavior largely unexplored, particularly outside Western healthcare systems with their distinct power dynamics and resource constraints.

The Iranian context sharpens the stakes. The country’s hospitals operate under severe economic sanctions, a depreciating currency, and chronic shortages of supplies, with nurse-to-patient ratios in public general wards often reaching 1:10 to 1:15—far beyond international standards. Extreme workloads, burnout, and the emigration of experienced staff create conditions in which rapid prioritization decisions are made constantly, and in which, the study suggests, discriminatory practices can take root whether or not anyone intends them. Against this backdrop, the research team conducted a secondary qualitative analysis of interviews originally collected between June 2023 and May 2024 in public and private hospitals across Tehran, using conventional inductive content analysis to let categories emerge from the data rather than imposing a pre-existing framework.

The participants—seven women and six men with a mean age of 34.7 years and roughly a decade of clinical experience each—worked in settings ranging from emergency departments and intensive care to surgery, psychiatry, and neonatal units. Saturation was confirmed after eleven interviews, with two additional interviews verifying that no new codes emerged. Rigor was reinforced through member checking with four participants, independent peer review of the coding, an external audit by a qualitative expert, and maximum variation sampling across departments, hospital types, and experience levels. The analysis yielded three overarching categories—patient-related, organizational, and nurse-related factors—comprising ten subcategories in total.

The patient-related findings are the most socially charged. Nurses consistently described how socioeconomic status shaped care: wealthy patients, celebrities, and government officials received faster medications, quicker physician visits, and the best rooms in the ward. One participant put it bluntly, saying such patients are prioritized in almost everything. At the opposite pole, prisoners, people with substance use disorders, homeless patients, and those lacking social support were described as receiving care delivered merely to fulfill a duty—fewer visits, less communication, a lower overall level of attention. Age and gender added further layers: young patients, especially young women, were monitored more carefully and given more explanation, while older adults, despite closer basic monitoring, sometimes saw aggressive treatment quietly scaled back on the assumption that they were old and had underlying diseases anyway.

Clinical condition itself became a vector of inequity. Nurses admitted that end-stage patients—those near death with little hope of recovery—could trigger emotional distancing, with care sliding into a minimal, task-focused routine. One nurse described care for such patients as provided just to keep them alive, not to make them comfortable. Stigmatized infectious diseases acted independently: when HIV or hepatitis appeared in a chart, some colleagues unconsciously kept their distance, delaying even routine tasks in ways that went well beyond standard universal precautions. Patient behavior mattered too—aggressive or demanding patients received less time and attention, while proactive family members who constantly followed up extracted more care for their relatives, a dynamic nurses themselves acknowledged was unfair.

Perhaps the study’s most striking contribution is its demonstration that much discrimination is organizational rather than personal. Nurses in private hospitals reported direct calls from managers flagging socially influential or ‘VIP’ patients as important to the hospital’s reputation and brand, instructing staff to provide everything without delay even when the patient’s clinical condition was no different from anyone else’s. Informal referrals from senior physicians and administrators—conveyed through a single phone call saying a patient is an acquaintance—were enough to reorder care priorities across an entire ward, and nurses who ignored such recommendations risked warnings, lost support, or damaged performance evaluations. In these situations, the researchers argue, nurses function less as agents of discrimination than as enforcers of institutional directives, a dynamic closely aligned with the concept of moral distress: knowing the ethically right action but being constrained from taking it.

Structural scarcity compounded everything. Facing staff shortages, insufficient equipment, and limited beds, nurses described triaging not by who deserved care more but by which patient they could manage faster. Complex, high-risk, or time-consuming patients drifted to the bottom of the list—not out of intent, participants stressed, but because systemic pressure made them the path of least resistance. Compounding this, the study found a striking educational void: nurses reported receiving in-service training on CPR, infection control, and new equipment, but never on how to treat all patients equally. Ethics instruction stayed abstract, with no real clinical examples, so discriminatory habits were absorbed unconsciously from ward culture and senior colleagues, normalized as an unavoidable byproduct of an overburdened system.

The nurse-related factors add a final, human dimension. Tips, gifts, and implicit promises from patients or families fostered a sense of reciprocal obligation, producing faster care and closer follow-up even when nurses knew it was professionally improper. Personal connections and fear of reputational damage when a patient was introduced by someone important created further sensitivity. Emotional responses cut both ways: one nurse described doing everything possible for a child whose condition resembled her own child’s, while another admitted following up less on terminal cancer patients because whatever she did would not make a difference. A third realized, after the fact, that a patient who reminded her of her deceased father had received extra time and attention at the implicit expense of others.

The authors’ conclusion is a call to shift the frame from individual blame to systemic design. Because discriminatory care is embedded in power relations, managerial logics, structural constraints, and educational gaps—not merely personal belief—sustainable change requires transparent triage protocols based on clinical urgency, anonymous reporting systems, ethics rounds where moral distress can be discussed without reprisal, and curricula that teach implicit bias and social determinants of health through case-based learning rather than abstract principle. Simple practices such as self-auditing time spent per patient could help nurses detect unconscious prioritization patterns. As one of the first in-depth qualitative investigations of this phenomenon in the Middle East, the study offers a multi-level framework that is directly actionable for resource-constrained health systems everywhere—and a sobering reminder that fairness at the bedside depends on fixing the system around it, not just the hearts within it.

Subject of Research: Factors contributing to discriminatory nursing care in Iranian hospitals

Article Title: Factors Contributing to Discriminatory Nursing Care: A Qualitative Study From Nurses' Experiences

Article References: Karimian, A., Sadooghiasl, A., Khoobi, M., Mohammadi, E., & Kazemnejad, A. (2026). Factors Contributing to Discriminatory Nursing Care: A Qualitative Study From Nurses' Experiences. Nursing Open, 13(10), Article e70890. https://doi.org/10.1002/nop2.70890

Image Credits: AI Generated

DOI: 10.1002/nop2.70890

Keywords: nursing, healthcare discrimination, qualitative research, Iran, health equity, moral distress, implicit bias, hospital management, ageism, stigma, nursing ethics, workload

The research, led by J. Christopher Polanco-Santana of Beth Israel Deaconess Medical Center and Harvard Medical School, together with Alessandra Storino of UT Southwestern Medical Center, Daniel Wong, and senior author Tara S. Kent, builds on the team’s earlier work. In a previous study published in the Journal of Surgical Education, the same group demonstrated that ethnic and racial bias existed in the MSPEs of general surgery residency applicants when measured by the differential use of two categories of descriptive language: agentic and communal terms. Agentic words, such as assertive, confident, independent, and ambitious, describe traits associated with leadership, self-direction, and achievement. Communal words, such as caring, compassionate, helpful, and warm, describe traits associated with relationships and interpersonal connection. Decades of research in social psychology and organizational behavior have shown that agentic language tends to be associated with professional success, particularly in competitive and high-status fields like surgery, while communal language, though positive, is less often linked to perceptions of leadership and excellence.

What remained unknown after that earlier study was which portion of the MSPE was responsible for the biased word use. The MSPE is not a monolithic document. It contains multiple distinct sections, including a summary or evaluation section, often referred to as the structural portion, in which the writer distills the applicant’s overall performance into a synthesis and often a comparative statement about where the student stands relative to peers. Other portions include the academic history, professional performance evaluations organized by clerkship, and the applicant’s own noteworthy characteristics. Each section is written under different circumstances and may draw on different sources of information, which means that bias could, in principle, be concentrated anywhere in the document. Identifying the specific source matters enormously, because targeted interventions, whether in writer training, template design, or automated language screening, can only be effective if they are aimed at the right section.

To answer the question, the researchers conducted a retrospective study of all MSPEs submitted by applicants to categorical general surgery residency positions at a single academic institution across two consecutive Match cycles. The final dataset comprised 1,314 evaluations drawn from 146 different medical schools, giving the analysis a broad national sampling of how these documents are written across the United States. The two application cycles were comparable at baseline, and the gender distribution of the applicant pool was nearly even, with women making up 51.6 percent of applicants. Most applicants identified as white, accounting for 52.7 percent, or Asian, accounting for 25.3 percent, while 17 percent of applicants self-identified as underrepresented in medicine, a designation that includes racial and ethnic groups whose representation in the medical profession is disproportionately low relative to their share of the general population.

The methodological core of the study was the calculation of a bias score for each distinct portion of the MSPE. Rather than treating the document as a single block of text, the researchers computed separate scores reflecting the relative frequency of agentic versus communal terms in each section. They then used multivariable regression, a statistical technique that allows researchers to assess the association between an outcome and one predictor while accounting for the influence of other variables, to test whether each section-specific bias score differed between underrepresented-in-medicine applicants and non-underrepresented applicants. This approach isolates where in the document the differential language use arises, rather than merely confirming that it exists somewhere. The use of regression also helps guard against the possibility that observed differences are explained by confounding factors such as application cycle, medical school, or applicant characteristics.

The results were striking in their specificity. Applicants who identified as underrepresented in medicine had lower median overall bias scores, at 9.09, compared with white applicants, whose median overall bias score was 13.04. The same pattern appeared in the structural or summary portion of the document, where the median bias score for underrepresented applicants was 13.33 compared with 20 for white applicants. Because higher scores reflect a greater relative use of agentic language, these numbers indicate that the MSPEs of underrepresented-in-medicine applicants were written with systematically different word choices, leaning more heavily on communal descriptors and less on agentic ones. When the researchers examined the other sections of the document, the differential word use was concentrated in the summary portion, indicating that the structural or summary section is the primary source of the biased writing.

This finding carries real weight for the residency selection process. The summary section is arguably the most consequential part of the MSPE, because it is the portion that program directors most often read closely and cite when deciding whom to invite for interviews. Survey research from the National Resident Matching Program has shown that the MSPE is among the most influential documents in residency selection, and studies of program director behavior have found that the narrative evaluation can change decisions about whether to invite an applicant at all. If the language in the summary section systematically frames underrepresented-in-medicine applicants in more communal and less agentic terms, then even well-intentioned reviewers may absorb an implicit impression that these candidates are warm and dedicated but less commanding or leader-like than their peers. Such impressions, formed in seconds during application review, can compound across thousands of programs and contribute to the persistent underrepresentation of certain racial and ethnic groups in surgery.

The study also situates itself within a broader and rapidly growing literature on bias in evaluative language across medicine. Prior work has documented racial and ethnic disparities in clinical grading during medical school, gender-based differences in the language students use to describe their own noteworthy characteristics, and agentic and communal differences in letters of recommendation for applicants to radiology, urology, and surgical training programs. Researchers have shown that implicit racial bias among health care professionals influences clinical outcomes, that implicit bias affects admissions decisions, and that word embeddings trained on large text corpora can quantify decades of shifting gender and ethnic stereotypes in language. Some studies have explored whether redacting identifiers from applications can reduce biased scoring, while others have asked whether artificial intelligence tools could help mitigate bias in recommendation writing. The new study adds a crucial layer of granularity to this literature by demonstrating that bias is not diffusely spread through evaluative documents but is concentrated in the section where writers synthesize and summarize.

Why would the summary section be particularly vulnerable? The authors’ findings are consistent with what psychologists know about how stereotypes operate. When writers compile detailed evaluations from discrete clerkship assessments, they are largely transcribing and organizing specific observations, a task that anchors them to concrete performance data. When they write the summary, however, they must distill an entire career into a general characterization, and it is precisely in this act of abstraction that implicit associations are most likely to surface. The summary is also the section most likely to include comparative statements, ranking the applicant against peers, and comparative judgments are known to be especially sensitive to stereotype-driven expectations. Whatever the underlying mechanism, the practical implication is clear: interventions aimed at reducing bias in the MSPE should focus their attention on the summary and structural portion of the document, whether through structured templates, writer training that highlights agentic and communal language, or automated screening tools applied specifically to the final synthesis before the letter is released.

The researchers caution that their data, drawn from applicants to a single academic institution over two Match cycles, reflect one program’s applicant pool, and that the underlying data, obtained through the Electronic Residency Application Service and containing protected applicant information, are not available directly from the authors. Nevertheless, the scale of the sample, spanning 146 medical schools and more than 1,300 evaluations, lends considerable weight to the conclusion. The study was approved by the Institutional Review Board at Beth Israel Deaconess Medical Center and received no specific external funding. As surgery and medicine more broadly confront the persistent gap between the diversity of the population they serve and the homogeneity of their workforce, this research offers something rare in the field: a precise anatomical map of where bias hides in one of the most important documents in a physician’s career. Fixing the summary section, the study suggests, may be the single most efficient place to start.

Subject of Research: Ethnic and racial bias in the language of medical student performance evaluations for general surgery residency applicants

Article Title: Source of ethnic/racial bias in medical school performance evaluation of general surgery residency applicants

Article References: Polanco-Santana, J. C., Storino, A., Wong, D., & Kent, T. S. (2026). Source of ethnic/racial bias in medical school performance evaluation of general surgery residency applicants. Global Surgical Education – Journal of the Association for Surgical Education, 5(1), Article 130. https://doi.org/10.1007/s44186-026-00531-5

Image Credits: AI Generated

DOI: 10.1007/s44186-026-00531-5

Keywords: MSPE, dean's letter, general surgery residency, ethnic bias, racial bias, agentic language, communal language, underrepresented in medicine, residency match, implicit bias, medical education, surgical workforce diversity

The study, published in Nature Machine Intelligence by Messi H. J. Lee of Washington University in St. Louis and Calvin K. Lai of Rutgers University, introduces what the authors call the reasoning-model implicit association test, or RM-IAT. Rather than measuring reaction times in milliseconds, as the human IAT does, the RM-IAT measures reasoning-token counts — the number of tokens a model produces in its internal reasoning process before arriving at an answer. The central insight is an analogy of computational effort: just as humans take longer to respond when a task conflicts with their implicit associations, a reasoning model may expend more computational effort — more thinking tokens — when a task is incompatible with the associations embedded in its learned representations.

The technical setup mirrors the classic two-block design of the human IAT. In association-compatible blocks, the pairing of a social category with a stereotypical attribute aligns with the associations the model is presumed to have absorbed from its training data. In association-incompatible blocks, those pairings are crossed, requiring the model to reason against the grain of its learned statistics. The researchers then compared the reasoning-token counts across the two conditions. If the model’s internal associations lean in a stereotypical direction, the incompatible condition should demand more elaborate reasoning, and therefore more tokens, to reach an equally acceptable answer.

Across four reasoning models — OpenAI’s o3-mini, DeepSeek-R1, gpt-oss-20b and Qwen3-8B — the researchers found consistent evidence for exactly that pattern. Association-incompatible tasks reliably required greater computational effort than association-compatible tasks, producing effect sizes analogous to the latency differences observed in human IAT studies. In other words, when these models were forced to reason in ways that cut across stereotypical associations, their chain-of-thought processes grew measurably longer, as though the models needed to work harder to override a default tendency. The consistency of the effect across four architecturally different models, trained by different organisations with different pipelines, suggests the phenomenon is not an idiosyncrasy of a single system but a general property of reasoning models trained on human language.

One model, however, broke the pattern in a fascinating way. Claude 3.7 Sonnet exhibited reversed effects: it expended more computational effort on association-compatible tasks rather than incompatible ones. A thematic analysis of its reasoning traces revealed why. Unlike the other models, Claude 3.7 Sonnet’s internal reasoning frequently turned inward, scrutinising the possibility of bias and stereotypes in the task itself before answering. This self-monitoring — an apparent product of its safety training — meant that stereotype-consistent pairings triggered extra deliberation about whether responding in line with a stereotype was appropriate, inflating token counts precisely where the other models were fastest. The finding offers a vivid illustration that the RM-IAT does not merely detect raw statistical associations; it detects the shape of a model’s processing dynamics, including the compensatory reflexes instilled by alignment training.

A crucial question for any such measure is whether it captures something meaningful about behaviour or is merely a statistical curiosity. The researchers addressed this by testing convergent validity: they examined whether RM-IAT effects predicted biases in downstream model outputs on tasks known to elicit biases in large language models. They found evidence that models showing stronger RM-IAT effects also displayed measurable biases in word association tasks and in decision-making scenarios — two domains where LLM bias has been extensively documented in prior research. This predictive relationship echoes the meta-analytic literature on the human IAT, where implicit measures show modest but reliable correlations with judgement and behaviour. In the machine context, the analogy suggests that the effortful override seen in reasoning tokens is not decoupled from what the models ultimately produce.

The theoretical framing draws heavily from the psychology of automaticity. Implicit biases are characterised as automatic mental processes that shape perception, judgement and behaviour — processes that are efficient, unintentional and often outside awareness. The human IAT operationalises this through the principle that compatible associations facilitate processing while incompatible ones impede it. Lee and Lai’s contribution is to argue that the same logic can be applied to reasoning models, where chain-of-thought generation provides a visible, countable trace of processing. Previous studies of bias in language models have focused almost exclusively on outputs — whether the model’s answers are discriminatory, whether generated text contains stereotypes, whether word embeddings encode prejudicial associations. The RM-IAT shifts the analytical lens upstream, to the reasoning process itself, opening a window on bias-like dynamics that output-level audits can miss.

The findings also connect to a growing body of evidence that surface-level debiasing does not eliminate deeper associations. Prior work has shown that explicitly unbiased language models can still form biased internal associations, and that alignment techniques such as reinforcement learning from human feedback may suppress stereotypical outputs without eradicating the underlying statistical tendencies learned from training corpora. The RM-IAT results are consistent with this picture: the models studied presumably produce carefully hedged, often non-stereotypical answers on sensitive topics, yet their reasoning processes still show the signature of effortful override when tasks run against stereotypical associations. Bias, in this sense, appears to persist as a property of the models’ computational dynamics even when it is masked at the output layer.

For AI developers and auditors, the practical implications are significant. Reasoning-token counts are cheap to measure, require no special access to model weights, and can be collected at scale, making the RM-IAT a potential screening tool for bias-like processing in deployed systems. Because the measure is sensitive to training interventions — as the Claude 3.7 Sonnet reversal demonstrates — it could serve as a diagnostic for whether safety training changes not just what models say but how they think. At the same time, the authors are careful with the interpretation: the measure captures bias-like processing differences, not proof of subjective experience or attitudes in the human sense. The analogy is structural, drawing on a shared signature of processing efficiency, rather than a claim that models possess minds with unconscious prejudice.

The broader significance of the work lies in the deepening dialogue between cognitive science and machine learning. Tools built to illuminate the automatic workings of the human mind are proving unexpectedly informative about statistical systems trained on human-produced text, and the patterns they reveal — effortful override, association-driven processing costs, dissociations between expression and underlying tendency — are recognisably familiar. As reasoning models become the dominant interface through which people interact with artificial intelligence, understanding the hidden dynamics of their chains of thought becomes as important as auditing their answers. The RM-IAT offers a rigorous, replicable method for doing so, and its central finding — that the hardest thinking for a reasoning model is often the thinking that goes against its associations — is likely to shape how researchers study machine bias for years to come. The authors have released their data openly on Figshare and their code on GitHub, with archived versions on Zenodo and a reproducible capsule on Code Ocean, ensuring that other laboratories can immediately extend this new bridge between the psychology of implicit bias and the engineering of reasoning machines.

Subject of Research: Measurement of implicit-bias-like processing patterns in large language model reasoning systems

Article Title: Implicit-bias-like patterns in reasoning models

Article References: Lee, M. H. J., & Lai, C. K. (2026). Implicit-bias-like patterns in reasoning models. Nature Machine Intelligence. https://doi.org/10.1038/s42256-026-01300-1

Image Credits: AI Generated

DOI: 10.1038/s42256-026-01300-1

Keywords: implicit bias, reasoning models, large language models, RM-IAT, computational effort, stereotypes, AI fairness, chain-of-thought reasoning, psychology of AI, algorithmic bias, Nature Machine Intelligence, implicit association test

The research, led by Sarah N. Kunz of Harvard Medical School and Beth Israel Deaconess Medical Center together with colleagues at Stanford University School of Medicine and the California Perinatal Quality Care Collaborative, examined data from California on very low birth weight, preterm infants born before 37 weeks of gestation who were less than 28 days old. The study period spanned 2012 through 2018, a window that captures a mature era of regionalized perinatal care in the nation’s most populous state. California offers a particularly valuable setting for this kind of analysis because of its sheer scale and the richness of its linked birth cohort records, which allow researchers to follow infants from birth through any subsequent acute transport between hospitals with unusual precision.

The design of the study was a retrospective cohort analysis, meaning the investigators looked backward at records that had already been generated by routine clinical care. Their outcome of interest was acute inter-hospital transport, the urgent movement of a newborn from one hospital to another, typically because the receiving institution can provide a level of neonatal intensive care that the discharging hospital cannot. These transports are among the highest-stakes events in neonatal medicine. A tiny infant, often weighing less than a carton of milk, is placed in a portable incubator, connected to a transport ventilator, monitored continuously, and driven or flown across traffic and distance to a destination intensive care unit. Every minute of that journey carries physiologic risk, and the quality of the stabilization before departure can determine whether the infant arrives in stable condition or in crisis.

To isolate the effect of race and ethnicity on transport likelihood, the team calculated odds ratios comparing each racial and ethnic group against non-Hispanic White infants. The comparison groups included non-Hispanic Black infants, infants classified as Asian American, Native Hawaiian, and Pacific Islander, often abbreviated AANHPI, and Hispanic infants. Critically, the models did not stop at this simple comparison. The investigators controlled for a battery of confounding factors organized into four domains: the characteristics of the hospital network in which the birth occurred, hospital-level factors such as the level of the neonatal intensive care unit, maternal sociodemographic characteristics, and infant clinical characteristics that reflect how sick the baby was at the time a transport decision would be made.

The central finding was striking in its specificity. Asian American, Native Hawaiian, and Pacific Islander infants were significantly less likely to be acutely transported than non-Hispanic White infants, with an odds ratio of 0.85 and a p-value of 0.01 after full risk adjustment. An odds ratio of 0.85 translates to roughly a 15 percent lower odds of transport for this group compared with their White counterparts, once all measured sources of confounding have been removed from the equation. In other words, this was not a difference explained by where these infants happened to be born, by the capabilities of their birth hospitals, by their gestational age or birth weight, or by the illness severity recorded in their charts. Something in the system itself appeared to be operating differently for these infants.

Equally instructive was what the analysis revealed about the strongest predictors of transport overall. Hospital-level factors, most notably the level of the neonatal intensive care unit at the birth hospital, were the variables most significantly associated with whether a transport occurred. This makes biological and organizational sense. An infant born at a community hospital without a Level III or Level IV neonatal intensive care unit is far more likely to require transfer to a regional center than an infant born already inside such a center. This mechanism is the entire logic of perinatal regionalization, the organized system through which states route high-risk mothers and infants toward hospitals with the resources to care for them. Studies stretching back to the 1980s have consistently demonstrated that very low birth weight infants delivered at appropriate levels of care experience lower mortality, and meta-analyses have confirmed the survival advantage of regionalized systems.

Yet the system does not always function as designed. Prior work by some of the same investigators has documented the phenomenon of deregionalization, in which an increasing share of very low birth weight infants are born at hospitals that lack the highest levels of neonatal capability, eroding the protective effect of the regional model. Other research from the California Perinatal Quality Care Collaborative has shown that racial and ethnic disparities extend deep into the quality of care itself, with infants from minoritized groups receiving care in neonatal intensive care units that systematically deliver lower-quality services, a pattern described as racial segregation and inequality within the neonatal intensive care landscape. The new transport findings fit into this larger mosaic, suggesting that inequities are not confined to what happens inside intensive care units but also shape the very pathways by which infants move between them.

The authors’ interpretation of their findings is deliberately measured. They conclude that the differing likelihood of acute transport by race and ethnicity may reflect underlying inequities and implicit biases in the system of care. This framing is important because it locates the problem not in the decisions of any single clinician but in the accumulated, often invisible patterns of how referrals are initiated, how transport teams are dispatched, and how risk is assessed across different patient populations. Implicit bias in clinical decision-making is well documented across medicine, and neonatal transport involves rapid judgments under time pressure, exactly the conditions in which unexamined assumptions about patients are most likely to influence behavior. Whether the lower transport rate among AANHPI infants reflects under-triage, differences in referral relationships, communication barriers, or other mechanisms is a question the study was not designed to answer, but the statistically robust association demands investigation.

The implications for policy and practice are concrete. Because hospital-level factors dominate the transport equation, strengthening adherence to regionalization principles, ensuring that high-risk deliveries occur at appropriately equipped hospitals, and auditing transport decisions for racial and ethnic equity are all actionable levers. The study also underscores the value of the granular data infrastructure maintained by quality collaboratives, which made it possible to detect a disparity that would be invisible in national aggregates. For the AANHPI category in particular, the finding adds urgency to calls for disaggregated data, since this grouping bundles together populations with widely divergent risk profiles and outcomes. What the study ultimately delivers is a measurable signal that the ambulance is not the same ambulance for every baby, a finding that should resonate far beyond California and into every perinatal system that claims, as its founding promise, that the sickest infants will reach the highest level of care regardless of who they are.

The dataset underpinning the analysis came from the California Perinatal Quality Care Collaborative, a statewide initiative that collects detailed clinical information from neonatal intensive care units across California. Because these records are gathered under a data use agreement rather than released publicly, the authors note that the underlying data are available only upon reasonable request with the collaborative’s permission. This governance model is common among quality collaboratives, which balance the research value of granular clinical data against the privacy protections owed to patients and participating hospitals.

The study’s statistical approach deserves emphasis. By adjusting simultaneously for network, hospital, maternal, and infant characteristics, the investigators sought to ensure that the observed difference in transport odds was not an artifact of clustering, since infants born at the same hospital share equipment, staffing, and referral practices. This kind of multilevel risk adjustment is essential in perinatal research, where the hospital an infant is born in is itself shaped by maternal residence, insurance, and patterns of segregation that precede any clinical decision.

The findings also connect to a long line of evidence on neonatal outcomes by race and ethnicity. Prior studies have documented disparities in very preterm neonatal morbidities, differences in mortality among very preterm infants across hospitals in large cities, and variation in outcomes for infants born at less than 30 weeks of gestation over time. Systematic reviews of neonatal intensive care have catalogued racial and ethnic differences spanning access, quality, and outcomes, indicating that no single point in the care pathway is immune.

Acute transport occupies a distinctive position in this pathway because it sits at the junction of multiple institutions. A transport decision requires coordination between the referring hospital, the transport team, and the receiving center, each with its own protocols and thresholds for escalation. Disparities arising at this junction may be harder to detect than disparities within a single unit, which makes the statistical signal reported here particularly valuable for quality improvement efforts aimed at ensuring equitable access to regionalized care.

Subject of Research: Racial and ethnic disparities in acute inter-hospital neonatal transport among very low birth weight infants in California

Article Title: Differential transport patterns by race and ethnicity in very low birth weight infants

Article References: Kunz, S. N., Zitnik, M., Helkey, D., Razdan, S., Gould, J. B., Profit, J., & Zupancic, J. A. F. (2026). Differential transport patterns by race and ethnicity in very low birth weight infants. Journal of Perinatology. https://doi.org/10.1038/s41372-026-02896-3

Image Credits: AI Generated

DOI: 10.1038/s41372-026-02896-3

Keywords: very low birth weight, neonatal transport, race and ethnicity, health disparities, neonatal intensive care, perinatal regionalization, odds ratio, California, Journal of Perinatology, AANHPI, NICU level, implicit bias