Who Reports Hospital Errors? Nurses, Doctors and Pharmacists Fill Out Safety Reports Differently

When a hospital patient is harmed—or nearly harmed—by a medication error, a fall, or a equipment malfunction, the incident is supposed to end up in a voluntary reporting database so the institution can learn from it. But a new study from a large tertiary hospital in China suggests that what ends up in that database depends heavily on who is doing the reporting. In a retrospective cross-sectional analysis of 7,218 patient safety incident reports submitted over a single year, researchers found that nurses, physicians, and pharmacists submitted strikingly different mixes of event types, and that nearly two-thirds of all reports arrived without any recorded severity classification at all.

The study, conducted at the Affiliated Hospital of Southwest Medical University and published in BMC Health Services Research, examined reports submitted between January and December 2025. Rather than treating all clinicians as a single homogeneous group, the research team led by Xia Tian, Yahui Liu, Mengli Chen, and Zaize Wang split the analysis along two axes: professional group—nurses, physicians, and pharmacists—and, within physicians, rank, ranging from junior to intermediate to senior. The goal was to see whether the composition of submitted reports and the completeness of a critical field, the severity of the event, varied systematically across these categories.

The headline numbers are telling. Nurses accounted for 57.23 percent of all submitted reports, with a 95 percent confidence interval of 56.09 to 58.37 percent. Physicians contributed 38.89 percent, and pharmacists just 3.88 percent. On its face, this suggests that nursing staff are the backbone of voluntary incident reporting, a pattern consistent with the long-standing observation in patient safety literature that nurses, who spend the most continuous time at the bedside, are often the most frequent reporters. But the authors are careful to note a crucial limitation: these figures describe the composition of submitted reports, not reporting rates. The study cannot estimate how often each profession actually encounters safety events, nor how likely members of each profession are to report an event when they do.

The most statistically striking finding concerns event-type composition. The association between profession and the type of event reported was very strong, with a Cramér’s V of 0.753—an effect size that, on a scale where 1.0 indicates a perfect association, indicates that knowing the reporter’s profession tells you a great deal about what kind of incident they reported. Severity-field completion showed a similarly strong association with profession, at Cramér’s V of 0.651. In practical terms, different professions were not merely reporting different numbers of incidents; they were reporting qualitatively different things, and they differed dramatically in whether they filled in the severity field at all.

That severity field matters more than it might first appear. Overall, 64.95 percent of reports—nearly two out of three—lacked a recorded severity category and were classified as Unclassified. Voluntary reporting systems are meant to support organizational learning: hospitals triage reports, investigate serious events, and feed lessons back into training and process redesign. When the severity of an event is missing, that triage becomes harder. A near-miss that could have caused catastrophic harm and a trivial inconvenience can look identical in the database, and safety officers cannot easily prioritize which reports deserve immediate attention. The authors argue that the high and profession-dependent proportion of Unclassified reports points to a need for clearer severity-classification guidance, training that is sensitive to both profession and career stage, and a review of reporting workflows—including who bears final responsibility for assigning severity.

The physician-rank analysis adds a subtler layer. Among physician-submitted reports, the associations between physician rank and both recorded severity distribution and event-type composition were statistically significant but small, with Cramér’s V values of 0.107 and 0.081 respectively. In other words, junior, intermediate, and senior physicians did differ in what they reported and how completely they characterized it, but those differences were modest compared with the chasm between professions. This is a notable departure from much of the existing literature, which has generally treated physicians as a single professional block. The findings suggest that if hospitals want to understand reporting behavior, the profession-level lens captures most of the variation, while career stage within medicine plays a secondary, though measurable, role.

Methodologically, the study is a single-center retrospective cross-sectional analysis, and the authors are explicit about what that design can and cannot support. Pearson’s chi-squared tests were used for group comparisons, with Cramér’s V reported as the effect-size measure and selected proportions accompanied by Wilson 95 percent confidence intervals, which perform better than standard intervals when proportions are near the extremes. The retrospective, anonymized nature of the database meant the institutional review board waived the requirement for informed consent, and the study was conducted in accordance with the Declaration of Helsinki. The authors received no external funding and declare no competing interests.

The limitations are as informative as the results. Because the analysis relies on submitted reports rather than an independent census of safety events, it cannot distinguish between differences in reporting propensity and differences in the underlying incidence of events. A hospital in which nurses report 57 percent of incidents might simply have a nursing culture that encourages reporting, or physicians might be underreporting, or both. The single-center design also limits generalizability: reporting culture, severity classification schemes, and workflow responsibilities vary widely across institutions and countries. The World Health Organization has long emphasized that voluntary reporting systems capture only a selected subset of patient safety events, and this study quantifies just how selective that subset can be.

For hospital administrators and patient safety officers, the practical implications are concrete. First, severity classification should not be left entirely to the reporter. Many reporting systems allow or require frontline staff to assign severity at the moment of submission, a task that requires judgment and training that vary across roles. Assigning final severity responsibility to trained reviewers, or building in a mandatory review step, could shrink the Unclassified category. Second, training on what and how to report should be tailored: a one-size-fits-all module may not address the specific event types and classification challenges that pharmacists, for instance, encounter when reporting medication-related near-misses. Third, the small but significant rank effects among physicians suggest that career-stage-sensitive approaches—recognizing that junior doctors may face different pressures and different event exposures than senior consultants—could refine reporting interventions.

Beyond the immediate operational lessons, the study speaks to a broader tension in patient safety science. Voluntary reporting is prized precisely because it is voluntary: reporters need psychological safety to disclose errors without fear of punishment. But voluntariness introduces selection effects at every step—who notices an event, who decides it is worth reporting, who completes the form, and who assigns severity. Each of those steps is filtered through professional identity, hierarchy, and workflow. By disaggregating physicians by rank and comparing three professions side by side, this analysis offers a template for other institutions to audit their own databases. If nearly two-thirds of reports at a well-resourced tertiary hospital arrive without a severity category, the question for every hospital running a reporting system is not whether its data are incomplete, but how incomplete—and what that incompleteness is hiding.

Subject of Research: Professional and physician-rank differences in hospital patient safety incident reporting

Article Title: Professional and physician-rank differences in submitted patient safety incident reports: a retrospective cross-sectional analysis of a hospital reporting database

Article References: Tian, X., Liu, Y., Chen, M., & Wang, Z. (2026). Professional and physician-rank differences in submitted patient safety incident reports: a retrospective cross-sectional analysis of a hospital reporting database. BMC Health Services Research. https://doi.org/10.1186/s12913-026-15757-8

Image Credits: AI Generated

DOI: 10.1186/s12913-026-15757-8

Keywords: patient safety, incident reporting, event severity, physician rank, nurses, pharmacists, hospital management, reporting database, health services research, near-miss, clinical governance, reporting propensity