I have worked in medical device engineering long enough to expect these stories. They still bother me. Hospital-born device ideas are almost always grounded in real, observed problems. A surgeon watches a suction device clog mid-procedure. An ICU nurse silences the same false alarm fourteen times in one hour, then misses the fifteenth, the one that matters. These people know what is broken. What tends to kill those ideas is not a lack of funding or regulatory knowledge. It is a consistent underestimation of what it takes to make a device function inside a hospital as a part of clinical operations. Teams assume that if the device works and the budget is there, the rest will follow. In practice, it does not.
Early Decisions, Late Consequences
Hospital teams tend to frame early development as a clinical question: does the device work, and does it address the need? Both matter. But they are not enough. Engineering decisions that get deferred in those early months are the ones that derail the project a year or two later.
A team we know of tested a polymer housing material for six months. It performed well in every bench test. During sterilisation validation, the material degraded after three autoclave cycles. Hospital central sterile processing departments typically require a minimum of twenty. That was not a delay. The project went back to material selection.
Tolerances are a similar problem. A device works in the lab because its inventor assembles it with a particular technique, a way of seating a connector, a step that never made it into the documentation. Someone else follows the written procedure and gets a different result. Without defined tolerances, there is no design transfer, and without design transfer, there is no path to manufacturing under a quality system. Both EU MDR and FDA expect this discipline early, not at the point of submission.
The working prototype is where teams get into the most trouble. It demonstrates the clinical concept. It attracts pilot funding. It convinces internal stakeholders. But it works because the inventor selected a specific sensor, applied heat to a housing that would not close, and adjusted a connector by hand until it fit. None of that appears in the design history file. None of it can be repeated by someone who was not present during the assembly.
Manufacturing
Hand assembly works for one unit, perhaps five. It does not work for a submission batch where each device must perform identically. We have seen the first batches where each unit takes two days to assemble, and each behaves differently. One needs a firmware adjustment. Another has a connector issue that the team has learned to compensate for. A third becomes the unit reserved for demonstrations because it happens to work reliably. That is not a manufacturing process.
3D printing follows a familiar pattern. A desktop printer produces a functional prototype. The concept is proven. Then the team moves toward medical-grade production and encounters material validation requirements, process qualification, and batch-to-batch consistency standards. The distance between a working print and a device that can be manufactured under ISO 13485 is large, and it catches teams off guard.
Component sourcing is less visible. A sensor sourced from a general electronics supplier performs well through twenty prototypes over two years. Then the manufacturer discontinues the part. No stock, no direct replacement. We have seen lead times reach 52 weeks for individual components. Projects with single-source parts and long lead times in their bill of materials do not have a realistic production path.
Care Pathway Integration
This is where the most expensive failures happen, and where I think the industry has the largest blind spot.
FDA clearance establishes safety and efficacy under controlled conditions. It does not establish that a device can function in a hospital on a Wednesday afternoon when the unit is short-staffed, and three other things are going wrong at the same time. Clearance is the beginning of commercialisation, not the end of it. Adoption is a separate problem, and it has little to do with clinical merit.
A device that cannot be placed within a specific point in the patient journey, used by whom, triggered by which clinical event, replacing or augmenting which existing step, will not be adopted regardless of its performance data. The French health authority HAS now lists "integration in the care pathway and organisational impact" as a formal evaluation criterion alongside interoperability and user acceptance. This reflects a broader recognition that technical validation alone does not predict whether a device will actually be used.
Reprocessing is one of the clearest examples. ECRI has listed inadequate or difficult device cleaning instructions among the top ten health technology hazards, citing direct links to hospital-acquired infections, device damage, and staff injuries. New devices regularly arrive at central sterile processing departments with instructions that require fourteen steps, specialised tools, or custom solutions. When the instructions are unrealistic, the device is not reprocessed. It goes back in the box. If it was expensive, it sits in storage. If it were cheap, it would be discarded.
Alarm fatigue is another. The FDA MAUDE database contains 566 reports of alarm-related patient deaths over five years. A Joint Commission review of 98 alarm-related sentinel events found that 80 resulted in death. Most involve monitors, ventilators, and infusion pumps, equipment in daily use. The alarms work. They fire constantly, including for non-actionable events, until the nurse who has responded to the same signal sixteen times stops responding. The devices were never designed with the reality of a twelve-hour clinical shift in mind.
Training follows the same pattern.
A new monitoring system requires four hours of training per nurse. The hospital employs two hundred nurses. The training budget covers twenty. The device ends up in one unit, used by a handful of staff, for three months, until rotation reassigns them, and no one remaining knows how to operate it.
We worked on a project where the development team spent months redesigning a disposal mechanism after receiving complaints from nursing staff. The actual problem turned out to be a cross-contamination risk at a different step in the workflow. The solution had nothing to do with disposal. Da Vinci surgical robot adoption illustrates how these dynamics play out at scale. The underutilization that many hospitals experience is organisational: insufficient case volume to reach breakeven, annual maintenance costs exceeding $100,000, dependence on a small number of trained surgeons, absence of dedicated OR scheduling blocks, and nursing staff who were never given structured training. When the trained surgeon leaves, the program often ends. A Johns Hopkins analysis estimated that major US hospitals collectively discard at least $15 million annually in unused but sterile surgical supplies. Capital equipment under active service contracts sits in storage rooms. This is what happens when purchasing decisions are made without utilisation targets or disposition plans.
Fragmented Ownership
Inside hospitals, device development responsibility is distributed across groups that operate independently.
The clinical team builds around efficacy. The innovation office tracks pilot metrics. Biomedical engineering reviews safety months after the design is set. Procurement becomes involved at the purchase order stage. Regulatory is consulted when someone raises the question of FDA classification. By then, there may be a form factor that cannot be sterilised, components without lot traceability, or assembly methods that depend on individual skill rather than documented procedure.
These problems are not caught because the people who would recognise them are not present when the decisions are made. A material selected in month two causes a sterilisation failure in month eighteen. An assembly method that works in a lab does not transfer. A component disappears from the supply chain before production begins.
We see a recurring pattern at the regulatory submission stage. A team has spent years on development. FDA requests process validation. The team discovers that their assembly procedure relies on operator judgment at several steps and cannot be validated as written. Redesign takes six months and costs approximately $200,000. By the time it is complete, the budget is exhausted, and the clinical champion who drove the project has moved to another institution.
What Changes the Outcome
Material and component selection should account for manufacturing requirements from the prototype. That means medical-grade traceability, qualified alternate suppliers, and validated supply chains. We worked with a team that selected automotive-grade sensors for a monitoring device instead of general-purpose equivalents. The cost per unit was higher, but the sensors came with batch documentation and ten-year availability commitments. When the device moved to production, nothing in the design needed to change.
Central sterile processing staff, procurement, and end users should be consulted during design, not after. Their operational constraints are design inputs. A surgical tool team that works with sterile processing before finalising design geometry avoids the reprocessing failures that derail projects months or years later.
The care pathway should be mapped as an engineering input. Who will use this device? At what point in the clinical workflow? Triggered by what event? What does it replace? How long does the procedure take with this device compared to the current method? How many staff require training, and what does that training cost relative to the hospital’s budget for it? If these questions do not have answers, the project has a technology but not yet a product.
Documentation should be maintained continuously. A team that records design decisions as they are made spends a manageable amount of time each week. A team that attempts to reconstruct the design history file later will spend months and will still have gaps.
The Pattern
Teams come to us regularly with devices that have clinical validation, sometimes regulatory clearance, and occasionally committed funding. The device is not moving forward. The reason is rarely that the technology does not work. It is that the device was developed without sufficient attention to how it would operate inside the system it is intended to serve.
A hospital is a set of interlocking operational layers: clinical workflows, sterile processing, information systems, financing, procurement, and governance. A device that does not account for any one of these layers will fail regardless of its clinical performance.
The problems described in this article are preventable. They require treating hospital operational reality as a design constraint from the beginning of development rather than as a commercialisation problem to be addressed later. When that happens, devices reach patients.
References
[1] MDCG 2023-1. Guidance on the health institution exemption under Article 5(5) of Regulation (EU) 2017/745 and 2017/746. European Commission, 2023.
[2] FDA. Quality Management System Regulation (QMSR) final rule, 2025.
[3] Quaregia. In-house devices under MDR/IVDR, 2024.
[4] Arterex Medical. Design verification and validation in medical devices, 2024.
[5] Complizen. FDA design controls for medical devices: 21 CFR 820.30 guide, 2025.
[6] FDA. Quality and compliance resources for medical devices, 2024.
[7] MatrixOne Health. Design verification vs. design validation in medical devices, 2024.
[8] Worksure. Medical device development: The crucial role of claim validation, 2024.
[9] European Commission. Study on reprocessing and reuse of single-use devices in the EU, 2024.
[10] Complizen. FDA warning letters for medical devices: Complete guide, 2025.
[11] FDA. Guidance for reprocessing single-use medical devices, 2024.
[12] ECRI. Top 10 health technology hazards for 2024.
[13] ECRI. Top 10 health technology hazards for 2025.
[14] Sterile Processing Technology. Case studies: Solving equipment issues in sterile processing departments, 2024.
[15] Med Equipment. FDA guidance for reprocessing single-use medical devices, 2024.
[16] NHS England. Medical technology strategy: One year on, 2024.
[17] Frontiers in Medical Technology. A comprehensive roadmap for MedTech innovations uptake in healthcare systems, 2023.
[18] Global Counsel. Unleashing innovation in the NHS: Report on procurement barriers, 2024.
[19] NHS England. The MedTech funding mandate, 2024.
[20] Health Innovation Network. Impact report 2023-24.
[21] Medicept. Guidance on the health institution exemption under Article 5(5) of EU MDR and EU IVDR, 2024.
[22] Artemia. MedTech commercialization: What it takes to beat the odds, 2024.
[23] PMC. Systematic analysis of medical device innovation barriers, 2024.
[24] Nature Digital Medicine. AI medical devices regulation and deployment challenges, 2024.
[25] Healthcare Guys. Why healthcare innovation fails: The $150M lesson, 2024.
[26] PMC. Medical device development and regulatory pathways, 2024.
[27] MDI Europa. New and updated MDCG guidance documents, 2024.
[28] Complizen. FDA design controls for medical devices: 21 CFR 820.30 guide, 2025.
