Word bans and the integrity of real-world data
In 2025 words and health data disappeared from US federal websites and scientific documents after White House executive orders. Real-world evidence depends on accurate, inclusive and standardised terminology and on large-scale health data, so the removals bear on the evidence itself, on the terms its data standards, demographics and translated patient-reported outcomes rest on, and on how scientists, writers and publishers respond.
Executive orders since 2016
An executive order is an official act of the President of the United States, and although it is not a law, it is a primary tool by which the President can direct the operations of the federal government. Executive orders can significantly shift federal research priorities, funding allocations and public health directives, and so influence which medical topics receive attention and resources. They also affect the transparency, availability and regulation of medical data.
Since 2016, US executive decisions ranging from memoranda to executive orders have had a significant impact on real-world evidence. They span Trump's first term, from January 2017 to January 2021, Biden's four years in between, and Trump's return to office in January 2025. Executive Order 13924 of 2020, on regulatory relief to support economic recovery, prioritised deregulation and reduced data and reporting burdens, which may have limited the development of real-world evidence infrastructure. In 2021 Executive Order 13985, on racial equity and underserved communities, promoted equity in data collection and health research, which enables inclusive real-world evidence. A memorandum on scientific integrity in the same year stated that federal policy must be guided by the best available science and data, and that scientific findings should never be distorted or influenced by political considerations. Executive Order 14081 of 2022, on biotechnology and biomanufacturing, encouraged data innovation and evidence generation, including the use of real-world evidence for regulatory and clinical applications. Executive Order 14094, on modernising regulatory review, promoted evidence-based decision-making and encouraged agencies to use modern data approaches, potentially including real-world evidence.
In 2025 Executive Order 14151 targeted diversity programmes that are essential to equitable real-world evidence, and it may roll back inclusive data strategies. Executive Order 14236 rescinded Executive Order 14081, and Executive Order 14094 was also rescinded, deregulatory moves that reduce support for real-world evidence, particularly where it is rooted in diversity, equity and inclusion, data modernisation or government health innovation. A further 87 orders, 14147 to 14233, commanded that documents, official communications and federal employees' email signatures be revised to remove references to diversity, equity and inclusion.
Words removed from federal documents
In 2025 certain words disappeared from scientific documents and official US government websites, as they had to a lesser extent in 2017, but with much greater intensity. These are the so-called word bans that followed the White House's executive orders. The White House denied there was a list of prohibited words, and no executive order banning specific words exists. In March 2025 The New York Times compiled, from the orders' texts in the Federal Register, a list of 197 words or concepts that agencies had flagged to limit or avoid. The list runs from "accessible" to "women and underrepresented" and includes many terms on diversity, equity and inclusion and on climate science, and these words were being removed from websites and replaced with terms the administration deemed acceptable. In one example, a memo about Head Start, a US programme of early childhood education for children in lower-income families, had its passages on the pandemic's disparate impact on under-resourced communities, on racial injustice and on the workforce's ethnic and linguistic make-up struck out. These policies have shaped the reporting of clinical and epidemiological information, real-world evidence included.
Specific uses of language were also reviewed to decide which health grants should be cancelled. A decision tree sent to all programme officers at the National Science Foundation, among others, and posted on social media by a university professor, asked whether a grant's title, abstract, project summary or description contained keywords and context implicating the orders. Operators of individual agencies were tasked with deciding, depending on the context, whether each term should be removed, replaced or retained. Because .gov sites are interconnected through hyperlinks and linked open data, a change in the terms of one .gov site can indirectly influence how terms are interpreted or used on other linked .gov sites, and although it does not update them, it can cause misalignment or misunderstanding or require reinterpretation downstream, so there are many opportunities for inconsistent or contradictory terminology across agencies. Roughly 8,000 federal web pages disappeared from public view, some returning later with warning banners such as "CDC's website is being modified to comply with President Trump's Executive Orders" and some, such as reproductiverights.gov, still unavailable in September 2025.
Health data taken offline
Real-world evidence depends fundamentally on the availability, transparency and integrity of large-scale health data, domains in which the US has historically been a global leader. If data are selectively removed, censored or altered for ideological or political purposes, the reliability of real-world evidence as a scientific tool is called into question, which affects both the credibility of US data sources and the trustworthiness of any evidence derived from them.
The CDC's Behavioral Risk Factor Surveillance System, one of the most widely used national health surveys, had run for about 40 years by 2025. It has been used for decades to inform policymakers, the media and the public on obesity rates, access to breast cancer screening, vaccination rates and the proportion of people with pre-existing conditions, and its sampling in every state makes its data particularly helpful for understanding health issues in low-population states and rural areas. It was briefly taken offline and later returned without its questionnaires or codebooks, and without that documentation researchers cannot verify how variables were measured or replicate analyses, which undermines any real-world evidence derived from the data.
Other federal health data taken offline at least temporarily include:
- the CDC's AtlasPlus, with about 15 years of surveillance data on HIV, viral hepatitis, sexually transmitted diseases and tuberculosis and data on the social determinants of health
- PEPFAR's data dashboards of programme budgets and expenditure by country and service category
- the USAID-supported Demographic and Health Surveys, with population, health, HIV and nutrition data from more than 90 countries
- foreignassistance.gov, with all US foreign assistance data
- the Area Health Resource Files, on health professionals and hospitals
- the CDC's Social Vulnerability Index, census-based socioeconomic data used for disaster planning, response and recovery
Why the terms carry the data
Real-world evidence depends on accurate, inclusive and standardised terminology, and banning certain words undermines data integrity and scientific utility. Terminological accuracy is essential to understandable and meaningful real-world evidence, and the loss of that accuracy or the elimination of context-specific terms can deprive decision-makers of vital information. Data standardisation relies on terminology, which the NIH defines as a systematically organised set of terms, concepts and codes that describe clinical conditions, procedures, medications and other health-care topics in a consistent and uniform manner, and a term is the human-readable text that can act as the anchor meaning for a concept. A term is not the concept, but the two are inextricably bound: losing the term starves the concept anchored to it, and free use of the term lets the concept circulate and be elaborated on.
The word bias, one of the banned words, is not a mark of diversity, equity and inclusion topics only, but a foundational concept in knowledge and science. The medical community knows that its attempts to understand reality are biased, and it accounts for those biases by routinely implementing corrective measures. In one view, banning the word sabotages efforts to understand reality as it is, and real-world evidence as its most appropriate measure.
A ban on "woman" and "women" would make it impossible to present data by gender, failing to reflect reality for either men or women. Removing the terms for ethnic groups, as the US administration commanded, would yield data that do not accurately describe reality. Removing the terms race and ethnicity, as the word bans state, implies that neither concept matters in medicine, which is known to be untrue for genetic, environmental, social or cultural reasons, often several combined.
Translated patient-reported outcomes
The bans will strip patient-reported outcome materials developed in the US of a wide range of commonly used terms that patients and health care providers understand unambiguously, making the source texts less comprehensible to patients and their translations non-viable or invalid. The number of viable source texts for patient-reported outcome translations will drop, and this will affect patients in real-life practice worldwide.
The response of scientists, publishers and writers
US scientists reacted to preserve the integrity of real-world evidence, including through the reversal of language bans, the republication of vital datasets and the reaffirmation of evidence-based standards in agency guidelines. Scientists, advocates and institutions mobilised to protect data and defend the principles of evidence-based research, and to save federal health websites and databases researchers downloaded datasets, scraped websites and archived them with the Internet Archive's Wayback Machine. The Association of Health Care Journalists protested the removal of public health data "at a time when the rise in chronic illnesses and harmful behaviors among young people is at the top of the national agenda", and the American Medical Writers Association reaffirmed its values and mission on diversity, equity and inclusion in a message to members.
Outside the United States, scientists and researchers showed solidarity. A Lancet commentary co-signed by editors worldwide called the policies part of a global assault on evidence, inclusion and truth, and urged that scientists, publishers and editors must resist silence. PLOS reaffirmed its commitment to open, rigorous science. A Nature Medicine commentary warned that banning words in medical research is bad news for everyone. A Tobacco Control editorial warned of savage cuts to health research, agencies and programmes, attempts to prevent, retract or amend publications, and deleted health databases.
In one view, scientists and writers should be free to use every term developed across disciplines that has been shown sound and valid for its purpose. The censorship policies underline the political vulnerability of health data systems, their implications for the reliability of research worldwide, and the need for international standards of data governance and independent audits. The medical writing community can contribute to safeguarding the ethical use of real-world evidence by building such standards and audits.