The Operative Sentence

In the studies that looked, most operating-room blood exposures were not recognised by the people exposed

Five small surgical series tested the barrier after the operation, and the interesting number is not how often blood arrived but how often nobody knew.

In brief

• Five surgical series recovered the mask, visor or eyewear after the operation and tested it. Blood was present on 45% to 66% of the items examined. All five are small series at a small number of sites, and the authors of four of the five report from outside the United States, so blood may reach the face barrier in a large share of operations and the certainty is low.12345
• Two of the five asked whether the wearer noticed. In one, 2% of blood splatters were recognised during the case; in the other, contamination was noticed during the case on 87 of 174 contaminated lenses (50%). On two studies in two specialties, contamination may go unrecognised in a large share of cases; the certainty is low.23
• The federal duty to wear eye protection is written as a forward-looking test: it attaches whenever splashes may be generated and eye, nose or mouth contamination can be reasonably anticipated.6
• The federal sharps injury log records percutaneous injuries from contaminated sharps. A splash to the eye is not an entry on it.6
• We found no national system currently collecting blood and body fluid exposure data from US healthcare personnel. CDC’s NaSH collected through 2007, and NHSN retired its Healthcare Personnel Exposure Modules in June 2025 for low participation.1112

What these studies actually did

The question that gets asked on nursing and surgical boards — how likely is this really — is normally answered with an incident count, which is a count of the exposures somebody reported. The five studies below did something cruder: they took the mask, the visor or the eyewear off at the end of the case, held it up to the light, and then applied a chemical that finds blood the eye cannot.

The chemistry matters, because it is what separates these papers from surveillance. Endo and colleagues examined 600 face shields at six medical facilities, first by visual inspection and then by staining with leucomalachite green.1 Stacey and colleagues collected 331 protective eye shields from four surgeons and their assistants over 131 oculofacial plastic surgery cases at three separate locations and read them with a luminol blood detection system, with controls to verify the protocol.3 Davies, a single surgical trainee — a year-2 specialist registrar, working as operator or as assistant — held the lens of his own protective glasses against a plain sheet of white paper before and after every operation for a year and inspected the mask at the end of each case.2 Lakhani and colleagues examined visor masks from 102 ENT procedures in one department macroscopically and microscopically;4 De Silva and colleagues did the same with visors worn across several surgical disciplines.5

None of these is a survey. Each is a census of one unit’s or one surgeon’s barriers over a defined window, and the reason to read them is not that they estimate a national rate — they cannot — but that they measure something a reporting system structurally cannot measure, which is contamination that nobody filed a form about.

The recognition gap

Two of the five studies asked the people wearing the barrier whether they had noticed anything, and those two are the reason this article exists. Stacey’s team gave a questionnaire to every surgeon and assistant after every case, and reported that blood was detected on 61% of shields and in 80% of cases while only 2% of blood splatters were recognised intraoperatively by the participants — a figure the authors restate in their own conclusion as a 98% incidence of undetected intraoperative blood splatter.3 Davies recorded, for each of the 384 operations he took part in, whether the contamination was noticed at the time.

Blood/body fluid was only noticed intra-operatively on the lens of the protective glasses in 87 (50%) cases. In none of the laparoscopic cases was blood or body fluid splash noticed until inspection of the protective glasses postoperatively.

Read the full paperArchived copy (2026-03-13 capture)

The two numbers are not the same measurement and should not be averaged. Davies’ 50% is unambiguously a proportion of contaminated cases. Stacey’s 2% is printed as a proportion of blood splatters, and the same abstract records that on every positive shield “the total number of blood spots was counted”. We read that 2% as a proportion of marks rather than of cases, since the 80%-of-cases figure is reported separately; the abstract does not settle it and we did not obtain the full text. Either way, the two figures are counted in different units. They point the same way and they differ by a wide margin, and the honest statement of what two studies in two specialties support is that a large share of contamination — somewhere between half and nearly all of it, depending on which study and which unit of counting — was not recognised while it was happening. The certainty is low: two studies, two specialties, small numbers of surgeons, and in Stacey’s case a splatter rate that varied between the individual surgeons from 29% to 90%.3

Endo’s pair of figures is often quoted as a third recognition finding and it is not one. The comparison there is between two methods of examining a shield after the fact — 303 of 600 shields (50.5%) showed contamination on visual inspection and 396 of 600 (66.0%) showed it after leucomalachite green staining.1 That is a statement about what an unaided eye can find on a piece of plastic, not about what the wearer perceived during the operation. It bears on the same problem from a different direction: the two counts differ by 93 shields, which on our arithmetic is about a quarter of the stain-positive shields. Calling those 93 the shields that showed nothing to the unaided eye assumes every visually positive shield was also positive on staining, and the abstract does not report the overlap between the two methods.

The study where most people said no

De Silva and colleagues asked surgeons and assistants in several surgical disciplines to wear a facemask with a transparent visor and to hand it in afterwards. Fifty-nine per cent of them refused.5 The 45% splash incidence the paper reports is therefore the incidence among the minority who agreed, and the authors say so in the same abstract that reports it.

That refusal rate is usually cited as a finding about attitudes. In our assessment it is at least as important as a finding about the data: every one of these studies depends on a barrier being worn and returned, which means each of them is measuring the operations of people willing to be measured. Whether that biases the splash rate up or down is not something any of these papers establishes, and we are not going to guess.

What the federal rule turns on

The standard does not set a splash rate, and it does not need one. The personal protective equipment duty in the bloodborne pathogens standard is written as a forward-looking test, and the whole of it turns on a single phrase.

The published series bear on the phrase “reasonably anticipated” and on nothing else in the standard. What they can be cited for is that in the settings studied, contamination arrived often and was noticed rarely, so an anticipation test cannot be discharged by asking whether anyone has seen it happen. What they cannot be cited for is any statement about a particular service line in a particular hospital; a rate from 102 ENT procedures in one department, or from 131 oculofacial cases at three US sites, is not a rate for anywhere else.

Why there is no larger number to give you

The obvious next question is what the national data show, and the answer is that the systems built to collect it have closed. Start with what the federal standard requires a facility to write down.

The employer shall establish and maintain a sharps injury log for the recording of percutaneous injuries from contaminated sharps. The information in the sharps injury log shall be recorded and maintained in such manner as to protect the confidentiality of the injured employee. The sharps injury log shall contain, at a minimum:

(A) The type and brand of device involved in the incident,

(B) The department or work area where the exposure incident occurred, and

(C) An explanation of how the incident occurred.

Read the full sectionArchived copy (2025-08-19 capture; we quote the 2026-08-04 issue)

The log is for percutaneous injuries from contaminated sharps.6 Blood in an eye is not a percutaneous injury and does not go on it. In our assessment that scoping phrase is much of why the numbers in national circulation about occupational blood exposure are needlestick numbers, and why a reader looking for data about splashes finds sharps data instead.

The voluntary systems that did collect splash data are in worse shape. CDC’s National Surveillance System for Healthcare Workers ran from 1995 and, in its own words, “collected surveillance data through 2007”; participation grew from 5 hospitals in 1995 to 64 in 2000 and fell to 18 by 2007, across nearly 130,000 reported adverse events, with at least one facility in 28 states and the District of Columbia.11 Of the 30,881 blood and body fluid exposures with a recorded location, operating rooms accounted for 29%.11 CDC’s own summary report is unusually candid about what that dataset can carry.

There are significant limitations of the data from NaSH including the absence of denominator data preventing reliable estimates of risk, and the variable number of reporting hospitals preventing meaningful trend analysis. Analyses of trends, particularly to identify reductions in injuries after the implementation of key OSHA regulations, were not feasible due to changes in healthcare facility participation in NaSH from year to year.

… additional limitations include a lack of representativeness of NaSH participants due to an overrepresentation of facilities from the eastern half of the US, over-representation of larger teaching hospitals, incomplete follow up after occupational exposures, and a small sample size of US facilities.

Read the full reportArchived copy

NaSH’s successor was supposed to be the Healthcare Personnel Safety Component of the National Healthcare Safety Network, and the NaSH report closes by looking forward to NHSN assuming that role.11 It did, and then it stopped.

The Healthcare Personnel Exposure Modules are used to monitor data on healthcare personnel for blood/body fluid exposures, assess the impact of preventive measures, and characterize antiviral medication use for exposures to influenza. Participation in the Healthcare Personnel Exposure Modules has declined over time, and NHSN has decided to retire the modules in June 2025 due to low participation. Facilities can continue to access their historical exposure data for analysis after the modules are retired.

Read the full newsletterArchived copy

The documentary trail matches the announcement. The NHSN page for the Healthcare Personnel Exposure module carried protocols and data collection forms as late as an Internet Archive capture dated 2025-06-10, with a “Last Reviewed” stamp of 25 March 2025; the capture dated 2025-08-08 is a redirect stub pointing at the component index, and every later capture we read is the same stub.12 The component index itself, captured 2026-07-25, lists two modules, both of them vaccination reporting.12

What is left, and what it is worth

EPINet is what remains as a multi-hospital US dataset. It is a convenience sample: hospitals choose to install the software and choose to submit, there is no sampling frame, the denominator is average daily census rather than full-time equivalents, and the set of contributing hospitals changes from year to year. Nothing in an EPINet report supports a year-over-year comparison, and we do not make one below.

One arithmetic operation on that data is ours rather than EPINet’s, and it is the one closest to this article’s subject. In the 2024 blood and body fluid report, among 453 records of barrier garments worn at the time of exposure, protective eyewear or goggles appear 19 times (4.2%), a face shield 16 times (3.5%) and eyeglasses with side shields 23 times (5.1%).10 That question accepts multiple answers, so a worker wearing goggles under a shield would be counted twice; the sum of 58 of 453, or 12.8%, is therefore an upper bound on the share of these exposures where any real eye protection was in place, not an estimate of it. The same arithmetic on the 2022 report gives 39 of 362, or 10.8%.8 Both bounds inherit everything wrong with the underlying data — a self-selected set of hospitals, no sampling frame, and a different set of contributors in each year — so the difference between those two numbers is not interpretable and we are not interpreting it.

The one US dataset we retrieved in this area that is not a convenience sample is a state one, and it covers sharps only.

Since 2001, hospitals licensed by the Massachusetts Department of Public Health (DPH) have been required to report data on sharps injuries among workers to the Department annually (MGL/Chapter 111 s 53D) and to use sharps with engineered sharps injury prevention features (SESIPs). Annual data have been collected from all DPH licensed hospitals since 2002.

Read the full data briefArchived copy

Because reporting is a licence condition rather than a volunteer effort, the Massachusetts figures are a count rather than a sample: 2,878 sharps injuries in 2021 and 2,743 in 2022, at rates of 16.4 and 15.7 per 100 licensed beds, which the Department describes as similar to rates since 2010 and as demonstrating a plateau of more than a decade.13 Operating rooms accounted for 902 of the 2021 injuries (31%) and 825 of the 2022 injuries (30%), and the wider category of operating and procedure rooms for 1,229 (43%) and 1,111 (41%).13 That is one state, and it counts injuries that pierce skin or mucous membrane; it does not count a splash to the eye that pierced nothing.

The number everybody quotes is about something else

Any page about occupational blood exposure eventually arrives at 385,000 sharps injuries a year. The figure is real and it has a document behind it: Panlilio and colleagues estimated 384,325 percutaneous injuries annually among US hospital-based healthcare workers, with a 95% confidence interval of 311,091 to 463,922.14 Three things travel with it and almost never do.

The first is the vintage. The estimate was built by combining data collected in 1997 and 1998 at 15 NaSH hospitals and 45 EPINet hospitals, adjusting for under-reporting, and weighting each hospital’s injuries by national admissions.14 It is a 1997–98 measurement, published in 2004, and it therefore predates the Needlestick Safety and Prevention Act and the revision to the bloodborne pathogens standard that the Act directed, which CDC records as becoming effective in 2001.11 A number collected before an intervention cannot describe the state of affairs after it.

The second is the scope. It is an estimate of percutaneous injuries. It says nothing about mucocutaneous exposure, which is the subject of every study in Table 1, and quoting it in an argument about eye protection is quoting the wrong number.

The third is how much of this rests on people saying so. Panlilio’s estimate was adjusted for under-reporting,14 and under-reporting is itself measured by asking. NaSH separately surveyed healthcare personnel about how many of their percutaneous injuries they had reported: across roughly 53,000 surveys distributed in 30 hospitals between 1996 and 2007, respondents said an average of 46% of injuries were reported, varying from 66% among technicians to 53% among nurses and 30% among surgeons.11 Neither document we retrieved says that this 46% was the adjustment Panlilio applied; the two exercises drew on different hospital sets over different periods. The survey figure is itself self-reported and is now at least nineteen years old, and we hold no current replacement for it.

In our assessment, what a reader can take from this is narrower than a rate and more durable. The published evidence does not support a sentence beginning “X% of operating-room staff are exposed each year”, because we found nobody in the United States counting mucocutaneous exposures at a national scale and the systems that tried have stopped. What it does support is that in every setting where somebody physically checked the barrier, blood was on it more often than anyone had said, and that the federal duty is written to be discharged in advance rather than in response to a report.

What would change this answer

• A US general-surgery series that recovers and tests the barrier. The authors of four of the five write from outside the United States, and the one US series is oculofacial plastic surgery. A US series in a high-volume general or vascular theatre would settle how far these rates travel, in either direction.
• A study that measures recognition rather than contamination. Only two of the five asked the wearer, and they counted in different units. A third study designed around the question would replace the range in this article with a figure.
• A restarted national system. If NHSN or another body resumed collecting healthcare personnel blood and body fluid exposure data, the central limitation of this page — that EPINet is the only multi-hospital US source we found, and is a convenience sample — would go away.
• Your own facility’s occupational health record. The federal log records percutaneous injuries from contaminated sharps and nothing else, but a facility’s own occupational health record is not confined to that. Whatever your own service has logged is better evidence about your own theatres than anything on this page.
• Another state with mandatory reporting. We describe Massachusetts because we retrieved its data brief. We have not surveyed the other forty-nine states for comparable mandates, and a second census-based state dataset would test whether the Massachusetts distribution across departments is unusual.

Where this could be wrong

The strongest objection is to the word “most” in the headline. It is carried mainly by Stacey, where 2% of splatters were recognised, and Davies puts the figure at exactly half of contaminated cases; a reader who thinks two studies in two specialties cannot support “most” is making a reasonable argument, and the body of this article gives the range rather than the headline figure for that reason. The second objection is to the detection chemistry. Luminol and leucomalachite green find blood at concentrations far below what a person would see or feel, and a positive shield is not the same thing as an exposure that mattered; none of these papers establishes that the material found would have transmitted anything, and this page makes no claim about transmission risk. Third, the 12.8% eye-protection figure is our own arithmetic on a multiple-answer question, and we have labelled it an upper bound rather than an estimate for that reason. Fourth, we read abstracts rather than full texts for four of the five series, and one of those abstracts is internally odd: Lakhani reports both a “median number of splash spots per mask” of 4.7 and an average of 8.2 splash marks per mask, which we cannot reconcile without the paper and have therefore not used. Fifth, and most general: studies that recovered barriers and found nothing are less likely to have been written up, and we cannot quantify that.