Data Fabrication

Data fabrication is the intentional creation of data, results, or experimental observations that were never actually collected, measured, or observed, and their presentation as genuine research findings. It is classified as one of the most serious forms of research misconduct, alongside data falsification and plagiarism. Unlike honest errors or differences of interpretation, fabrication deceives by presenting invented values as real observations, whether the aim is to support a desired conclusion, to complete a dataset or to improve the results. Formal definitions depend on the jurisdiction or code. In the United States, the federal research misconduct policy and the Public Health Service (PHS) regulation, 42 CFR Part 93 (revised in 2024), define fabrication as making up data or results and recording or reporting them, and falsification as manipulating research materials, equipment, or processes, or changing or omitting data or results so that the research is not accurately represented in the research record. Under 42 CFR Part 93, a finding of research misconduct requires a significant departure from accepted practices of the relevant research community, committed intentionally, knowingly, or recklessly, and proven by a preponderance of the evidence. In Europe, the European Code of Conduct for Research Integrity (ALLEA, revised 2023) defines fabrication as making up data or results and recording them as if they were real.

Fabricated data corrupt the scientific record and can cause tangible harm when they influence clinical practice, public policy, or the direction of subsequent research. Patients may receive ineffective or dangerous treatments based on fabricated clinical trial results. Other researchers may waste years pursuing lines of inquiry built on a fraudulent foundation. Public trust in science erodes when fabrication cases become public, particularly when the fraudulent work was published in prominent journals and widely cited. The financial cost is hard to measure: one analysis estimated that papers retracted for research misconduct accounted for about $58 million in direct funding from the US National Institutes of Health between 1992 and 2012, less than 1% of the agency's budget over that period, before counting investigation expenses and the downstream effects on the research enterprise. Institutions, journals, and funding agencies have therefore invested heavily in systems to prevent, detect, and respond to fabrication.

Fabrication can take many forms. A researcher might generate an entire fictitious dataset, invent patients who were never enrolled in a study, or create laboratory results that were never obtained from actual experiments. In some cases, fabrication is partial: a researcher conducts some experiments but fills in missing data points with invented values to complete a dataset or improve the results.

Detection relies on a combination of statistical forensics, audits, and whistleblower reports. Statistical methods can identify data that are "too good to be true," such as distributions that are too uniform, standard deviations that are implausibly small, terminal digit distributions that deviate from expected patterns, or results that are perfectly consistent across subgroups in ways that real biological data almost never are. Tools such as GRIM (Granularity Related Inconsistency of Means) can detect reported means that are mathematically impossible given the reported sample size and scale, and SPRITE (Sample Parameter Reconstruction via Iterative Techniques) rebuilds sets of values that could have produced reported means and standard deviations, which shows whether the reported statistics are plausible.

When allegations of fabrication arise, institutions typically activate a formal investigation process. For research supported by the US Public Health Service, such as NIH grants, institutions follow 42 CFR Part 93, which the Office of Research Integrity (ORI) oversees: an initial assessment of whether the allegation falls within the definition of research misconduct and is credible and specific enough, then an inquiry to determine whether a full investigation is warranted. The 2024 revision applies to allegations that institutions receive on or after January 1, 2026. Research funded by other US agencies, such as the National Science Foundation, falls under those agencies' own rules, and other countries have their own procedures. The investigation committee, composed of researchers with relevant expertise, examines the evidence, interviews witnesses, reviews original data and laboratory notebooks, and determines by a preponderance of evidence whether misconduct occurred. Consequences for confirmed fabrication can include retraction of affected publications, termination of employment, and potential legal liability. In PHS-supported research, HHS can also impose administrative actions such as correction or retraction of the research record, supervision of the researcher's future PHS-funded work, or special review of their funding requests, and the HHS debarment official can exclude the researcher from federal grants and contracts; under the federal debarment rules, debarment generally should not exceed three years, but longer periods can be imposed when circumstances warrant.

This entry describes a prohibited practice, not a method to be applied. Understanding data fabrication is important for research ethics training, for reviewers and editors who must evaluate the integrity of submitted manuscripts, for research integrity officers who investigate allegations, and for all researchers who need to understand the boundaries of acceptable practice. Knowledge of fabrication indicators helps statisticians and peer reviewers identify suspicious data patterns.

While fabrication itself is the mistake, there are adjacent errors that researchers should avoid. Poor documentation of data collection (missing laboratory notebooks, absent audit trails) can create the appearance of fabrication even when none occurred, making the researcher unable to defend the legitimacy of their data. Delegating data collection to trainees without adequate oversight creates opportunities for fabrication that the principal investigator may not detect. Pressure to publish, particularly in competitive academic environments with "publish or perish" cultures, creates conditions where misconduct is more likely, and mentors who ignore warning signs share responsibility. Confusing fabrication with data imputation is another issue; statistical imputation of missing data using documented, transparent methods is a legitimate analytical technique, while inventing data to fill gaps and presenting them as observed values is misconduct.

An institutional investigation reveals that a postdoctoral researcher fabricated patient records for a multicenter cardiovascular outcomes study. Statistical reviewers noticed that the reported standard deviations were implausibly small and identical across all treatment subgroups, a pattern inconsistent with real patient data. Examination of the electronic data capture system showed that 84 patient records were entered from a single IP address on a weekend when the clinic was closed. The affected publications were retracted, following an ORI finding of research misconduct, HHS debarred the researcher from federal funding for seven years, longer than the three years the debarment rules generally allow, because of the seriousness of the fabrication, and the institution implemented mandatory source document verification for all clinical trials.