What digital triage support means
Triage prioritises care according to the estimated urgency of intervention. WHO distinguishes this function from simply arriving or registering. Its Interagency Integrated Triage Tool, developed with ICRC and MSF, is intended for triage within emergency units; it is not a general application enabling anyone to diagnose people at work.
A support system may organise information, present authorised clinical rules or help document decisions. The specific function needs describing before its value can be assessed. This entry concerns preventive integration into workplace medical emergencies, with competent professionals and connection to healthcare services. It does not offer a self-assessment algorithm or criteria for independently deciding that someone can wait. Software should support care rather than become a barrier to requesting it, particularly when available information is incomplete.
Sources: [1]
Recording, recommendations and clinical decisions
Distinguish three functions that a single screen may combine. Recording preserves data; a recommendation interprets information for a defined purpose; a clinical decision assigns an action within professional competence. Calling all three automation obscures what is being delegated and which evidence is needed.
As a design criterion, we propose that the interface identify entered data, missing information and outputs that remain suggestions awaiting review. A blank field should not silently become absence of symptoms. Clinical priority must also remain distinct from administrative order, cost of care or whether a person is directly employed or works for a contractor. The pathway should preserve this distinction from the first communication through the transfer of care. A neat record alone does not establish that the prioritisation was appropriate.
Who uses the tool
The official IITT frequently asked questions assign triage to healthcare workers trained in the tool, rather than security guards. They also place triage before administrative registration. This guidance concerns the stated healthcare setting; having a tablet does not confer new clinical competencies.
For workplaces, we propose explicitly defining the roles of reception, initial responders and healthcare professionals. A person can activate assistance and communicate observed facts without assigning a clinical category. First-aid training should not be presented as universal authorisation to apply any triage system. Design should help information reach the appropriate professional and prevent a need for care from being held up while awaiting information technology validation. Role definitions should remain clear during night shifts, absences and contractor attendance.
Sources: [2]
Evidence and the software’s intended purpose
The FDA’s January 2026 guidance distinguishes decision-support functions that may fall outside the US definition of a medical device from functions that remain within it. The label support does not settle regulatory status. This reference describes the US framework and does not replace regulatory assessment in another jurisdiction.
For procurement, we propose requesting intended purpose, users, population, use setting, version and validation evidence. A classifier developed from hospital data cannot be assumed to cover every workplace. Check which conditions were excluded and how errors were analysed. If a supplier reports overall accuracy alone, the consequences of priorities being assigned too low or too high remain unanswered. Acceptance should concern the actual function purchased, not a broader claim that the software uses artificial intelligence.
Sources: [4]
Validation within the actual care pathway
A useful evaluation examines both the system output and what happens afterwards. We propose observing whether professionals have enough information to review it, can disagree and can communicate their decision. A technically correct recommendation may lose value if it arrives late or reaches someone unable to activate resources.
A pilot should use cases and scenarios authorised by clinical leads, without delaying real care or experimenting on patients without appropriate safeguards. Record missing information, ambiguous language, identification errors and changes in condition. Review across relevant profiles and situations can reveal limitations hidden by an average. The objective is not to make every team member obey the algorithm, but to demonstrate safe assistance within the care process. A documented disagreement may be useful learning rather than a failure to comply.
Human autonomy, equity and confidentiality
WHO places human autonomy, safety, transparency, accountability, inclusion and equity among its principles for health AI. Maintaining human control and protecting privacy are governance conditions, not outcomes that can be assumed from installing an application.
As an organisational application, we propose reviewing accessibility, language and the person’s ability to communicate. A system should not penalise missing digital information or inability to respond as a form expects. Define access by role, retention and transfer channels. An employer does not need the complete clinical assessment to organise support or a preventive measure. Human review needs real time, information and authority; routinely clicking accept does not provide meaningful oversight. Review arrangements should be understandable to those using the system and those responsible for clinical governance.
Connecting with the site emergency plan
The pathway must fit available resources and external coordination. We propose documenting who activates a response, who transmits information and who confirms receipt. If a workplace has no healthcare service of its own, an application does not create that capacity or replace emergency services.
Site hazards also affect organisation: access for healthcare teams, the person’s location and availability of relevant information about the activity. Staff need to follow authorised procedures to avoid becoming exposed while helping. This entry does not prescribe rescue techniques, treatment or substance-specific measures. Digital support should facilitate existing coordination without creating an additional chain of approvals that delays care. A useful exercise checks the whole communication route rather than stopping when a notification appears on a screen.
Continuity when data or communications fail
We propose rehearsing what happens when there is no connection, the application stops responding or a person cannot be correctly identified. A known operational alternative should fit the workplace and be agreed with healthcare leads. Software failure cannot become a reason to postpone activating necessary assistance.
Change control matters too: a new version may alter questions, presentation or logic. Its effects on training and procedures need review before deployment. Retain versions, dates and relevant decisions without creating unnecessary duplicate clinical records. Later reconstruction of administrative records should reflect what actually happened, rather than implying that an automated recommendation preceded a human decision when it was produced afterwards. This distinction makes retrospective review more reliable and prevents the record from overstating the system’s contribution.
Hypothetical example of a drill
A site with its own healthcare service prepares a drill using role players. The hypothetical objective is to test how a request for help travels from a workshop to the clinical team. A tablet organises information, but the exercise reveals that it requires an employee identifier unavailable to contractor personnel.
The team removes this barrier and also tests the offline alternative. Healthcare professionals review clinical decisions in the scenario; administrative staff review transmission and timing. Passing one drill does not validate the system for every emergency. The outcome is a specific improvement in access to assistance and a list of outstanding questions requiring further assessment, with owners and closure criteria. The example illustrates process design, not evidence of a real product’s clinical performance.
Indicators that support learning
A monitoring dashboard can distinguish time until assistance is activated, information available on receipt, communication failures and use of continuity arrangements. Clinical quality requires professional review and appropriate criteria, rather than inference from clicks or time spent on a screen.
We propose analysing revised priorities, reasons for disagreement, missing data and corrective measures. Errors that underestimate urgency deserve specific attention; excessive prioritisation that may impede resource organisation should not be ignored either. Figures need context and information about the population served. Shorter registration time alone does not establish better care. Monitoring should reward finding and correcting problems, rather than hiding disagreements to create an appearance of perfect concordance with the system. The aim is reliable care, not a flawless-looking software report.
Relationship to OSH and integration limits
Triage support connects first aid, emergency planning, telemedicine, clinical safety and information governance. Its specific contribution to OSH is better coordination when a health problem occurs at work, without confusing the location with a diagnosis of occupational causation.
A management platform may organise training, contacts, drills and improvement actions where its functions allow. That does not mean it contains clinically validated software for assigning priorities. This entry attributes no undocumented automated triage function to Sabentis. Before integrating data or recommendations, confirm purpose, responsibility and boundaries. Success means accessible, coordinated care with traceable professional decisions, rather than replacing people with automatic classification. The software remains one component of a response whose staffing, resources and accountability must also be in place.
Frequently asked questions
Does an application replace a healthcare professional?
No. Competent users, meaningful oversight and authorised functions must be defined; software does not confer clinical competence.
Does triage mean diagnosis?
No. It prioritises urgency of care rather than establishing a definitive diagnosis.
Can assistance wait for registration?
Administrative processing or software failure should not block necessary care. A defined pathway and continuity alternative are needed.
Does a drill clinically validate an algorithm?
Not by itself. A drill can reveal organisational failures; clinical validation needs evidence matching the population, function and context.
