UV-C robots in the clinic: what disinfection robotics can and cannot do

Hospital-acquired infections are a persistent and clinically serious problem in German hospitals. Reliable national case numbers are harder to come by than the figures circulating in vendor material suggest, and we deliberately do not quote one here: the widely repeated range attributed to the Robert Koch Institute cannot be traced to a primary RKI publication. What is undisputed is the operational reality behind it. Terminal disinfection after an isolation patient is discharged is labour-intensive, time-critical and highly dependent on the person doing it. Under staffing pressure, contact time and coverage vary. That variability, not a headline number, is the actual argument for automation.

This is where no-touch systems enter the discussion. A UV-C robot drives into a vacated room, positions itself at defined waypoints, irradiates for a defined period and logs the run. The appeal for hospital management is reproducibility and documentation, not a claim that the room becomes sterile. Facility management, hygiene officers and procurement should evaluate the technology on exactly that basis. Our overview of service robotics in healthcare places disinfection robots in the wider context of what ground robots currently do on a ward.

• Reproducibility: a robot delivers the same cycle at 3 a.m. as at 9 a.m., which manual processes under load do not.
• Documentation: every run produces a timestamped record, which is useful for hygiene audits.
• Staff relief: the cycle runs while cleaning staff are working elsewhere, provided the room can be locked.
• Hard boundary: none of this replaces the manual wipe step. It is added on top of it.

Germicidal UV-C is typically emitted at 254 nm by low-pressure mercury lamps. At that wavelength the radiation is absorbed by nucleic acids and induces pyrimidine dimers in microbial DNA and RNA, which blocks replication. The mechanism is purely physical and it is entirely governed by radiometry. There is no chemistry, no residual effect and no distribution mechanism inside the room. What the lamp does not reach, it does not affect.

The decisive quantity is dose: irradiance multiplied by exposure time, usually expressed in millijoules per square centimetre. Irradiance drops with the square of the distance from the emitter. Doubling the distance between lamp and surface leaves roughly a quarter of the irradiance, so a surface two metres away needs roughly four times the dwell time of one at one metre for the same dose. This is why cycle times in real rooms are measured in tens of minutes across several waypoints rather than in a single pass, and why claims based on lamp power alone say very little about what actually happens on a bed rail.

• Dose, not wattage: the relevant figure is what arrives at the surface, which depends on geometry and time.
• Distance dominates: the inverse-square relationship makes robot positioning the main design variable in a room.
• No cleaning effect: UV-C does not dissolve or remove soil. Blood, protein and dust layers shield organisms underneath them completely.
• No penetration: UV-C does not pass through fabric, packaging, drawer fronts or opaque plastics.

Photons travel in straight lines. Every object in a patient room casts a UV shadow, and everything inside that shadow receives no dose at all. In a typical room that includes the underside of the bedside table, the rear of the monitor, the inside of the drawer, the back of the door handle, the space between mattress and frame, and the wall side of any device on wheels. These are not marginal areas. They overlap substantially with the high-touch surfaces that matter most for transmission.

Robots mitigate this by moving. A cycle typically consists of several pre-calculated positions from which the room is irradiated from different angles, so that a surface shadowed from one position is exposed from another. This reduces the shadowed fraction, it does not eliminate it. Closed drawers, closed cabinets and anything under a mattress remain out of reach regardless of how many waypoints are added. Anyone specifying such a system should walk the actual room and identify which surfaces will realistically be reached, rather than accepting a generic room diagram from a datasheet.

• Multi-waypoint cycles: more positions mean better coverage and longer total cycle time. The trade-off is throughput per shift.
• Room clearance matters: the more equipment left standing in the room, the worse the coverage. Cycle quality depends on preparation by staff.
• Verification is difficult: dose indicators or radiometric measurement at defined points are the only way to show what a given room layout actually receives.
• Unreachable by design: enclosed volumes cannot be treated by a room system at all.

UV-C at germicidal intensity is hazardous to skin and eyes. Overexposure causes erythema and photokeratitis, with symptoms typically appearing hours after the exposure. Consequently the room must be unoccupied for the entire cycle, and that state has to be technically enforced rather than organisationally assumed. In German hospitals this falls under the employer's risk assessment for artificial optical radiation, so the safety concept is not optional and belongs in the procurement documentation from the start.

In practice this means layered safeguards. Door contacts or interlocks cut lamp power when a door opens. Motion sensing on the robot itself shuts down emission if presence is detected in the room. Signage, warning lights and audible pre-warnings cover the organisational side, and access control determines who can start a cycle at all. Each of these layers has failure modes, which is why the combination matters more than any single sensor specification. Fault logs and interruption records should be reviewable afterwards, both for hygiene documentation and for occupational safety.

• Door interlocks: emission stops before a person can physically enter the irradiated zone.
• Presence detection: onboard sensing as a second, independent layer rather than the only one.
• Warnings and access control: visible and audible signals plus a defined authorisation to start cycles.
• Auditable interruptions: every aborted cycle must be visible, because an aborted cycle is an incomplete dose.

This is the part of the discussion that vendor material usually compresses into a single percentage. The honest picture is considerably more restrained. Percentage figures for pathogen reduction on test surfaces are laboratory or field microbiology endpoints, not infection outcomes, and they do not answer the question a hospital actually cares about: does adding a UV-C robot reduce infections in patients?

The one large randomised trial on this question is the BETR-D study (Anderson et al., The Lancet 2017;389:805-814). It reported a relative risk of 0.70 for the UV arm, with a 95 percent confidence interval of 0.50 to 0.98. The study is, however, internally inconsistent: the arm combining bleach with UV showed the weakest and statistically non-significant effect, with a relative risk of 0.91. If UV-C added a robust independent benefit, one would not expect that pattern. Sun et al. (Epidemiology and Infection, 2023) found an incidence rate ratio of 0.90 for Clostridioides difficile and 0.72 for VRE, neither statistically significant, and concluded in their own words that they found no advantages for the use of UV-C in healthcare settings. Health Quality Ontario, in its 2019 health technology assessment, recommended against public funding. There is no Cochrane review on UV-C room disinfection and infection rates.

Vendor-specific evidence is weaker still. For UVD Robots, the platform behind roughly 40 German hospital deployments funded through the EU emergency programme in 2021 and 2022, there are no published controlled studies reporting infection rates. The frequently cited Xenex phrase about FDA authorisation refers to whole-room microbial reduction, which is a microbiological endpoint, not a claim about infection prevention. None of this makes UV-C useless. It means the technology should be procured as a documented supplementary process step with plausible microbiological effect, not as an intervention with proven outcome benefit.

• BETR-D 2017: the only large randomised trial, positive in the UV arm but inconsistent in the bleach-plus-UV arm.
• Sun et al. 2023: no significant effect for C. difficile or VRE.
• Health Quality Ontario 2019: recommendation against public funding.
• No Cochrane review on the question, and no published controlled infection-rate studies for the most widely installed platform in Germany.

The regulatory position in Germany is unambiguous. KRINKO recommendations on surface hygiene treat no-touch procedures as a supplementary measure. They do not replace cleaning and wipe disinfection. The reason is the one described above: pathogens in clinical settings are rarely sitting exposed on a clean surface, they sit in blood, secretions, protein residue and dust. Mechanical wiping removes that layer. Without it, the organisms underneath are shielded and the dose delivered to the surface above them is irrelevant.

The practical consequence for a workflow is a fixed order: clean and wipe-disinfect manually, then vacate and lock the room, then run the UV-C cycle, then document both steps. A UV-C run that is used to shorten or skip the manual step is not an optimisation, it is a downgrade. The St. Josef-Hospital in Bochum, which has been operating an ICA Traffic Hero21 system since October 2020, describes its process explicitly as a combination of scrub-wipe disinfection and robot, which is the correct framing.

A second distinction matters just as much in procurement, and it is regularly blurred: cleaning robots clean, they do not disinfect. Autonomous scrubber-driers remove soil from floors. They are not disinfection devices, and the two categories should never appear interchangeably in a specification. The BHUK practical guidance on scrubber-driers in hygienically critical areas, published in March 2024, places robotic machines primarily in areas without infection risk or with possible infection risk, advises against single-tank machines for hospitals, and flags data protection questions around camera and mapping data. Fraunhofer IPA put the state of the market plainly in November 2022, stating that no suitable cleaning robots are currently available on the market that can implement automated cleaning and disinfection of surfaces both technologically and economically. Nothing since then has overturned that assessment for surface disinfection. Our article on what ground robots actually do in hospitals works through the same distinction across the other robot categories.

UV-C robots are usually bought as appliances and then fail on the operational edges: rooms that cannot be locked, cycles that never fit the turnaround window, documentation that no one can export, and lifts or doors the robot cannot pass. The technical hygiene question is comparatively easy. The integration question is where projects stall, and it is worth putting into the tender explicitly rather than discovering it after delivery.

Two points deserve particular attention. First, dose verification: ask how the supplier demonstrates the dose actually delivered in your room geometry, not in a reference room, and what happens to the record when a cycle is interrupted. Second, building integration: no manufacturer in this market publicly documents a specific lift protocol or a lift manufacturer partnership, and fire door actuation is documented by nobody at all. If the robot has to change floors or pass a fire door, that is an integration line item in the project budget, not a product feature off the shelf. The same applies to transport platforms, as discussed in our piece on transport robots in hospitals.

• Process order in writing: the SOP must state that the UV-C cycle follows the completed manual wipe step, with both documented.
• Realistic cycle times: measured in your actual rooms with your actual furniture, not from a datasheet.
• Room availability: how many rooms per shift can genuinely be vacated and locked for the required duration.
• Safety concept: interlocks, presence detection, warnings and access control, assessed as a whole under the optical radiation risk assessment.
• Documentation export: can run records be exported into the existing hygiene documentation, and are aborted cycles visible.
• Lamp lifetime and service: lamp ageing reduces irradiance over time, so replacement intervals and service response belong in the contract.

A hardware-independent view helps here, because the sensible answer is sometimes that a UV-C robot is not the right instrument for the ward in question. Our robot catalogue covers the available platform categories, and the ROI calculator makes the cost side of a deployment comparable before a tender is written. werob works as a manufacturer-independent systems integrator for service robotics, based in Hamburg, which means the starting point is the ward workflow and the hygiene regime, not a single supplier's product range.