CMMS for Healthcare and Hospitals

Most CMMS advice written for healthcare makes one of two mistakes. Either it treats a hospital as a slightly stricter office building, or it treats healthcare maintenance as a single discipline, which ignores the fact that the people maintaining the air handling units and the people maintaining the infusion pumps answer to different managers, different standards and often different regulators. If you are selecting or configuring a CMMS for a hospital, the first thing to get straight is that you are supporting two professional communities with genuinely different requirements, and the second is that the regulatory framework you must satisfy depends entirely on where you are.

The message up front: healthcare maintenance is two disciplines, estates and biomedical, sharing a building and a patient population. A healthcare CMMS has to serve both without forcing either to work in a way its profession does not recognise. The hardest requirement is not scheduling, it is evidence: proving to an accreditation surveyor or a regulator, years later, that a specific device or system was maintained correctly by a competent person on the date the record claims.

1. Two disciplines, one building: why healthcare is different

In a commercial building, maintenance is one function. In a hospital it is at least two, and they are structurally distinct in a way that a CMMS configuration has to respect.

Estates and facilities looks after the building and its engineering services: medical gas pipelines, electrical distribution and emergency power, HVAC with its pressure regimes and filtration, water systems, sterilisation plant, lifts, fire and life safety, the building fabric. The staff are mechanical, electrical and building trades, the standards are engineering and building-code standards, and the failure mode that frightens them is a system-wide outage: the theatre suite losing air, a chiller failing in summer, a generator not picking up the load.

Biomedical engineering, also called clinical engineering or medical physics depending on the country, looks after the medical devices: monitors, ventilators, infusion pumps, defibrillators, anaesthesia machines, dialysis machines, imaging equipment, laboratory analysers. The staff are biomedical engineers and technicians, sometimes with vendor-certified specialisms per modality. The standards are device standards, manufacturer instructions and medical-device regulation. The failure mode that frightens them is a single device delivering the wrong therapy or the wrong reading to one patient, a quieter failure than a chiller going down and potentially a more direct harm.

Both groups do preventive maintenance, corrective work, inspections and compliance testing, which is why the temptation to treat them as one function is strong. But the asset models, intervals, evidence, competency rules and regulatory owners differ enough that a configuration built for one will irritate the other. If you are new to the underlying platform concepts, the CMMS buyer's introduction and the core modules explainer are the right starting points, and the general-purpose view is in CMMS for facilities management.

2. Estates versus biomedical: a side-by-side comparison

This is the table I put in front of a hospital steering group early, because it settles a lot of arguments about whether one configuration can serve both.

Look at the location row in particular, because it is the single configuration issue that causes the most trouble. Estates assets have a location. Biomedical assets have a location right now, and the record of where they have been matters for infection tracing, for utilisation analysis and for finding them when a recall lands. A CMMS that models location as a static asset attribute will fight biomedical every day.

3. Estates: the constraints that make hospital maintenance hard

Hospital estates work is technically similar to any other complex building, and operationally much harder. Four constraints do most of the damage to a naive maintenance plan.

There is no downtime window. A commercial tower has evenings, weekends and holidays; a hospital has none of those. Work that elsewhere would simply be scheduled out of hours has to be planned around clinical activity, so maintenance planning becomes a negotiation rather than a calendar exercise. Critical systems often need designed-in redundancy specifically so that maintenance is possible at all, and the plan has to be written against that redundancy design: which chiller, which transformer, which air handling unit can be taken out, under what conditions, with what fallback.

Infection control constrains the physical work. This is the constraint that most surprises engineers arriving from other sectors. Dust generation, water disturbance, disruption of airflow and temporary openings are all infection risks in a clinical environment, particularly near immunocompromised patients, theatres, intensive care and sterile services. In many hospitals, work in or adjacent to clinical areas requires an infection-control risk assessment and specified precautions before it starts: containment, negative pressure on the work area, agreed routes for materials and waste, cleaning and verification afterwards. The CMMS implication is that the permit and risk-assessment step is a gating condition on the work order, not optional metadata, and the evidence that it was done belongs in the record.

Ward access has to be negotiated, not assumed. A planned task on a ward is only executable when the ward agrees, so the CMMS needs to represent the difference between "due", "offered to the ward", "accepted for a date" and "deferred by clinical request". If your system only has "scheduled" and "overdue", the maintenance team will be recorded as failing when in fact they were refused access, and that distortion poisons both the KPI reporting and the relationship with clinical teams. Capturing deferral reason as structured data is one of the highest-value small configuration decisions in a hospital CMMS.

The evidence burden is continuous. Accreditation and licensing regimes generally expect documented evidence that critical systems are inspected and tested, that the people doing the work are competent, and that deficiencies are tracked to closure. That evidence is produced in the ordinary course of maintenance work or it is not produced at all. Hospitals that treat accreditation as a project undertaken before a survey end up reconstructing history under time pressure, which is both expensive and unconvincing.

4. The critical engineering systems and what drives their maintenance

Every hospital has a set of systems whose failure has direct clinical consequence. These are the systems where criticality classification is not a paper exercise, and where the maintenance regime is usually set by a mix of code, manufacturer requirement and local health-authority guidance.

For the deeper treatment of the life-safety end of this list, see elevator, fire and life safety system PM. For ranking these systems against each other in a defensible way, asset criticality classification is the method I would use, and it earns its keep in healthcare more than in any other sector because criticality here is a clinical argument, not a financial one.

5. Biomedical: the device inventory is the whole foundation

On the biomedical side, everything rests on the completeness and accuracy of the medical device inventory. A device that is not in the inventory is not scheduled, not tested, not tracked when a safety notice lands, and invisible when someone asks how many of a given model the hospital owns. Inventory gaps are the most common finding I would expect on a biomedical CMMS review, and they arrive through predictable routes: devices bought on departmental budgets without going through clinical engineering, devices arriving with a project or a research grant, loan and trial devices that never left, and donated equipment.

What a usable device record needs, beyond the obvious identifier and model:

• Manufacturer, model and serial number recorded exactly as the manufacturer writes them, because recall and field safety notices are matched on model and serial ranges. Sloppy model naming makes recall matching a manual trawl.
• A device classification that lets you group by type consistently. Whatever nomenclature your region uses, pick one and enforce it, because free-text device naming destroys any ability to report by device family.
• Risk or criticality class according to your own agreed model, typically weighing the consequence of malfunction, whether the device is life-supporting, and how it is used.
• Owning department and current location, with current location treated as changeable state and historical location retained.
• Service arrangement: in-house, vendor contract, shared-risk or per-incident, with the contract reference and entitlements attached. This is where warranty management matters, because a device under warranty repaired in-house may lose its cover.
• Accessories and consumable dependencies, since a monitor without its correct cable or cuff is not a functioning device.
• Software version and network status: whether the device connects to the clinical network, holds patient data or feeds a clinical record. This has become a joint responsibility with IT and information security rather than a purely biomedical concern.

The control I would put in place first is a procurement gate: no medical device enters clinical use without being registered by clinical engineering and receiving acceptance testing. It is an administrative discipline rather than a technical one, and it is worth more than any feature in the CMMS.

6. Scheduled performance assurance, safety testing and calibration

Biomedical planned work is usually a bundle of related but distinct activities, and one reason biomedical teams get frustrated with generic CMMS configurations is that these get collapsed into a single undifferentiated "PM" task.

• Acceptance or incoming inspection: performed before first clinical use, establishing that the device arrived undamaged, is correctly configured, performs to specification and is safe. It is also where the baseline measurements for later comparison are captured.
• Performance assurance: periodic verification that the device still performs within specification. What this means is device-specific: a defibrillator energy check, an infusion pump flow-rate check, a ventilator volume and pressure check, an analyser control run. It normally requires its own calibrated test equipment.
• Electrical safety testing: verification of protective earth continuity, insulation and leakage currents against the limits in the applicable standard for medical electrical equipment, with limits differing by applied-part classification. The principle is stable across regimes; the exact limits, test conditions and intervals come from the standard and guidance that apply to you, so take them from your own standard rather than from an article.
• Calibration: adjustment or verification against a traceable reference where the device produces a measurement clinicians rely on. The point people miss is traceability of the test equipment itself. A pump analyser or safety tester that is out of calibration invalidates every test performed with it, so the test-equipment register needs the same rigour as the device register, and the CMMS should link each test record to the instrument used.
• Manufacturer-specified planned maintenance: parts replacement, kit-based servicing and internal adjustments specified in the service manual, often with hour or cycle triggers on higher-end equipment.
• Functional checks by clinical users: pre-use checks that belong to the clinical team rather than engineering. Worth modelling, because a device failing a pre-use check should generate a work request quickly, and the fastest route from bedside to work order is a QR or barcode scan rather than a phone call.

The structural lesson is that these need distinct work order types with distinct forms, acceptance criteria and required fields, not one PM type with a long free-text notes box. The framework for that is in work order types in a CMMS, and the scheduling mechanics are in the complete guide to preventive maintenance.

7. Risk-based intervals versus manufacturer recommendations: the live debate

This is the most contested question in clinical engineering, and it deserves to be presented as a genuine professional disagreement rather than resolved with a confident answer.

The traditional position is that you follow the manufacturer's recommended schedule for every device, without deviation. The argument is straightforward: the manufacturer designed and tested the device, their instructions form part of the conditions under which it is deemed safe, deviating exposes the organisation legally, and no hospital has better failure data on a given model than the company that makes it.

The risk-based position is that manufacturer schedules are set conservatively for a global population of users and environments, that many devices show no evidence of benefit from the prescribed interval, and that finite biomedical capacity spent on low-yield inspections of low-risk devices is capacity taken from higher-risk work. Advocates argue that an evidence-led programme, adjusting intervals using the hospital's own failure and inspection-finding data, device risk class and utilisation, delivers better overall safety than uniform adherence. The clinical engineering literature has supported variants of this view for years, and a number of health systems operate this way.

Where I land: the risk-based argument is intellectually strong and operationally attractive, and it is only defensible if three things are true.

• Your regulator and accreditation body permit it, explicitly. Some regimes allow an alternative equipment maintenance approach with documented justification; others do not, and some exclude categories such as life-support, imaging, laser and radiological equipment from any deviation. This is the first question to answer, not the last, and the answer comes from your own authority in writing.
• Your data is good enough to support the decision. A risk-based interval is a conclusion drawn from inspection findings, failure history and utilisation. If your records say "PM completed" with no findings captured, you have no evidence base and any interval change is a guess wearing an evidence-based label.
• The decision is documented, governed and reviewed. Who approved the deviation, on what evidence, with what clinical involvement, and when it is reviewed. A risk-based programme without that governance trail is indistinguishable, to a surveyor, from one that simply stopped doing the work.

The part of the debate that is not contested: deferring work because you are short-staffed is not risk-based maintenance, and calling it that is the failure mode I would watch for hardest. Risk-based maintenance is a documented engineering decision made in advance. Capacity-driven deferral is a resourcing problem. A CMMS that lets the second masquerade as the first is doing the organisation real harm.

8. Recalls, field safety notices and alerts

A capability unique to biomedical and almost always underbuilt is the handling of external device safety communications. Manufacturers and regulators issue recalls, field safety corrective actions, safety notices and advisory bulletins, and the hospital has to answer three questions quickly: do we own any affected devices, where are they right now, and what action is required. The workflow I would build is more a discipline than a feature:

• A single monitored intake point for safety communications, so notices arriving by email, through a regulator subscription or via a vendor account manager all land in one place with an owner.
• A logged assessment per notice, recording the date received, the affected model and serial criteria, the search performed against the inventory, and the result including a documented nil return. A nil return is evidence and needs recording as such.
• Work orders generated per affected device, linked back to the notice, so completion is tracked unit by unit rather than as a departmental assurance that it was handled.
• A quarantine mechanism marking affected devices as withdrawn from service, visibly at the device and as a hard status in the CMMS, so a device cannot be issued from a store or loan pool while quarantined.
• A closure record showing the full population, the action taken on each unit, and the devices that could not be found. Unlocatable devices are the uncomfortable output of this process and precisely why location tracking matters.

The honest observation: the recall process is where inventory quality gets tested in public. A hospital with a clean inventory and current locations closes a notice in days. A hospital without one spends weeks walking corridors, and can never fully prove it found everything.

9. Loan pools, mobile equipment and the location problem

A large share of medical devices are mobile, and many sit in shared pools: infusion pumps, syringe drivers, monitors, feeding pumps, specialist beds and pressure-relieving surfaces, and rented equipment brought in for a single patient. This creates three problems a generic CMMS does not solve out of the box.

Finding the device when it is due. A due date on a pump that has sat in a store cupboard on the fourth floor for six weeks is not actionable. The workable answers are a library or issue-and-return model where devices are scanned out to a ward and back to a decontamination and check point, which gives a natural interception opportunity for planned work, or a real-time location system, which is effective, expensive, and only worth it on high-value or high-churn equipment.

Rented and loaned-in equipment. Devices belonging to a supplier but used on your patients still need to be safe, tracked and returned. They should be in the system with a distinct ownership flag, an expected return date, and clarity about who is responsible for their maintenance. Equipment that arrives on loan and quietly becomes permanent, with nobody owning its maintenance, is a recurring finding.

Pool sizing and the maintenance connection. Biomedical teams are frequently asked to keep devices in service because there is no spare, which is a fleet-sizing problem pushed onto the maintenance function. The useful move is to make it visible: report devices out of service and the reason, and the case for pool depth or contract improvement makes itself. On the parts side, the pool-and-spares thinking in spare parts and MRO inventory in a CMMS transfers directly, with the addition that many biomedical parts are single-source and contract-controlled.

10. One CMMS or two: the arguments each way

This is the question that usually brings me into a hospital conversation, and the honest answer is that both models work and both fail, for reasons that have more to do with governance than software.

The pattern that works best in my experience is a single system for the request channel, the location and department masters, work order lifecycle and consolidated reporting, with biomedical configured as a genuinely first-class discipline inside it rather than as estates plus a few extra fields. That means device-specific work order types, test-record structures holding measured values and the instrument used, a device risk class that actually drives scheduling, and recall handling built as a real process. If the platform cannot do those things, the case for a specialist clinical-engineering system becomes strong, and it is a legitimate choice rather than a failure of integration nerve.

What does not work is a single system configured by an estates-led project with biomedical consulted late. The biomedical team will keep a spreadsheet, the spreadsheet will become the real record, and the CMMS will hold an incomplete device inventory nobody trusts. If you cannot resource genuine biomedical involvement in the configuration, two systems with a clean integration is the more honest plan. The broader platform-category question is covered in CAFM vs CMMS vs EAM vs IWMS.

11. The evidence and audit layer that dominates everything

If you remember one operational point from this article, it should be that healthcare maintenance record-keeping is judged on evidential quality, not on completion percentages. A surveyor or regulator is generally asking a version of: show me that this was done, correctly, by someone competent, on the date you claim, and show me what you did about what you found. What that means for configuration:

• Capture findings, not completion. A closed work order with no recorded readings, observations or pass/fail outcomes is administratively complete and evidentially worthless. Structured findings are also the only route to risk-based intervals later.
• Record measured values as data. Numbers in a notes field cannot be trended, checked against limits or reported. Numbers in typed fields with limits can flag out-of-tolerance results automatically.
• Link the test instrument used. This is the traceability chain that makes a measurement credible, and it is routinely missing.
• Attach competency to the person doing the work. Whether a trade licence, an authorised-person appointment for a permit regime, or device-specific training, the system should be able to show the individual who signed the record was qualified on that date, and expiring competency needs to be visible before it expires.
• Track deficiencies to closure separately from the parent task. A finding raised during an inspection has its own life, risk rating and closure date. Surveyors follow findings, and a system that buries them inside a closed inspection cannot show that trail.
• Make deferrals explicit and reasoned. Clinical access refusal, parts unavailable, capacity, and planned risk-based change are four different things and should never collapse into one "overdue" bucket.
• Protect the audit trail. A system where completion dates can be quietly altered after closure is not an evidence system.
• Plan for retention. Records often need to be retrievable for many years, potentially beyond the life of the device and of the software. Ask during selection how you would export a complete, readable history if you changed platform.

The measurement frame that supports this is covered in the FM KPI framework. In healthcare I would add one KPI that is rarely reported and always revealing: the proportion of completed compliance tasks containing complete structured evidence. It is usually lower than anyone expects, and improving it is worth more than improving raw completion.

12. Platforms, and what to actually check

Healthcare is served by three broad groups of products, and being clear about which you are buying from avoids a lot of disappointment.

• Enterprise asset management platforms such as IBM Maximo, SAP PM, Hexagon EAM and Infor EAM. Strong on estates, deep hierarchies, work management at scale and integration with finance and procurement. Device-specific workflows may need configuration or an add-on, and implementation effort is substantial.
• Workplace and facility platforms such as Planon, coming from the property and facilities direction and strong on space, compliance and service management. Biomedical depth again varies.
• Modern cloud CMMS products such as MaintainX, Limble, Fiix, UpKeep and eMaint. Fast to deploy, strong mobile experience, genuinely good at getting frontline staff to raise and close work. The questions to press are structured test-record capture, evidence integrity, competency control and whether the audit trail meets a surveyor's expectations.
• Specialist clinical-engineering systems, which exist precisely because device management has requirements generalist tools treat as edge cases. If biomedical is the dominant scope, these belong on the shortlist.

Whichever group you look at, the questions that actually differentiate in healthcare are narrow: structured findings and measured values with limits, test-equipment traceability, recall search by model and serial, device location history, competency linkage, deferral reasons, audit-trail immutability, long-term retention and export, and mobile working where phones and networks may be restricted. Feature lists rarely answer these. Scripted demos with your own data do.

Two external references worth having open while you work through requirements, as starting points rather than answers: the World Health Organization publishes medical device and health technology management guidance aimed at exactly this problem, and ISO is the source for the medical electrical equipment and quality management standards your local rules will reference. For accreditation expectations, go to your own accreditation body's current published standards, such as those from Joint Commission International where that is the scheme you are surveyed against, rather than to a secondary summary.

13. A sequence I would follow

A healthcare CMMS implementation that starts with software configuration is already behind. The sequence that works:

• Establish the regulatory and accreditation baseline in writing. Which authority regulates devices, which regulates the estate, which accreditation scheme you are surveyed against, and what each expects in terms of records. Get this from the authorities and your compliance lead, and record it as the requirement source for everything that follows.
• Give both disciplines equal standing in the project. A named biomedical lead and a named estates lead with equal say on the data model and work order types. This is the decision that most determines whether one system works.
• Build one location and department hierarchy, agreed by both sides and by clinical services, because it is the spine of scheduling, reporting and access negotiation.
• Verify the inventories physically. A walked device inventory and a walked critical-systems asset register. Unglamorous, expensive, and the foundation of everything. Do it before go-live, not after a survey.
• Agree the criticality and device risk model with clinical involvement. Documented scoring rules, signed off, reviewable.
• Design work order types and evidence forms before touching schedules. Decide what evidence each activity must produce, then build the task to capture it.
• Load compliance-critical schedules first. Medical gas, emergency power, clinical HVAC, water safety, fire and life safety, and the highest-risk device classes. Prove the evidence chain there before expanding coverage.
• Put request raising in clinical hands. QR or barcode on the device or room, minimal form, clear feedback on status. Adoption by clinical staff is what keeps the record complete.
• Run a mock survey against the system. Ask someone to pull evidence as a surveyor would, on a sample of assets and devices. Every gap that exercise finds is one you would otherwise find under pressure.

For the general PM-programme mechanics underneath this, preventive maintenance examples across industries gives comparable patterns from other sectors that transfer reasonably well to the estates side.

14. Where this approach costs you, and where it does not work

Everything above has a price, and it is worth naming it.

Structured evidence capture takes longer at the point of work. A technician recording six measured values, an instrument reference and a pass/fail outcome spends more time documenting than one ticking a completion box. That is a real productivity cost, and if you impose it without either mobile tooling that makes it fast or an honest adjustment to workload expectations, technicians will find shortcuts and evidence quality will quietly degrade back to where it started.

Physical inventory verification is expensive and politically unpopular. It produces no visible improvement on the day it finishes, and it is the first thing cut when a project runs late. Cutting it is the most common reason healthcare CMMS implementations underdeliver, and there is no software substitute.

Location tracking rarely justifies itself across the whole estate. Real-time location systems work, and their cost only makes sense on high-value, high-churn or frequently-lost equipment. Applied to every device, the business case collapses. A library and issue-return discipline achieves most of the benefit at a fraction of the cost, for the subset of devices that genuinely move.

Small facilities may not need any of this structure. A single-site clinic with a few hundred devices can be run well on a simple system and a competent lead, and imposing a hospital-grade configuration produces administrative overhead without safety benefit. Scale the rigour to the risk.

And a CMMS cannot fix understaffing or underinvestment. If the biomedical team is half the size the device population needs, or the estate has a large backlog of ageing critical plant, a better system will document the problem more clearly and will not solve it. That documentation has value, it is often what finally makes the case for investment, but nobody should expect software to close a resourcing gap.

The idea to walk away with

A healthcare CMMS succeeds or fails on two things that have nothing to do with feature lists. The first is whether it treats estates and biomedical as two equally legitimate disciplines with different needs, rather than configuring for one and asking the other to adapt. The second is whether it produces evidence as a natural by-product of daily work, so that the record which satisfies a surveyor is the same record the technician created at the device, years earlier, without extra effort.

Everything else follows from those two. A verified inventory, a location model that matches reality, structured findings with traceable measurements, competency attached to the signature, deficiencies tracked in their own right, and deferrals recorded with honest reasons. Get those in place and the choice between one system and two becomes a straightforward practical judgement rather than an argument. Skip them and no platform on the market will save the programme.

Final thoughts

The hardest thing about healthcare maintenance is not the engineering. It is that two professions with different standards, different regulators and different fears have to operate in the same building, on the same patients, often through the same system, while the building never closes. A good CMMS does not resolve that tension, it makes it workable: one request channel, one location truth, shared reporting, and enough specialisation inside the configuration that neither discipline feels like a guest.

And because this is safety-critical work, I will end where I started. Nothing here is a substitute for your own regulator, your own accreditation standards, your own manufacturers' service documentation and your own competent persons. Use this as a structure for the questions and a description of the principles. Take every interval, limit and acceptance criterion from the authority that will actually hold you to it.

Related reading: CMMS for facilities management, What is a CMMS: a complete buyer's introduction, Elevator, fire and life safety system PM, Asset criticality classification, Work order types in a CMMS, CAFM vs CMMS vs EAM vs IWMS.