Hospital del Mar, a research hospital in Barcelona, spent the past year working with Qilimanjaro Quantum Tech, a Barcelona-based quantum computing company known for its multi-modal platform – combining digital and analog quantum processors with classical supercomputing. Together, they tackled one of healthcare’s hardest logistics problems: surgical scheduling.
At a hospital in Barcelona, the bottleneck is not always an operating room.
Sometimes it is a bed.
Hospital del Mar has 11 operating rooms but only 15 beds in its post-anaesthesia care unit. Every operation that leaves an operating room creates demand for recovery capacity. Schedule too many procedures requiring lengthy recovery at the wrong time and the unit fills up. Patients wait. Operating rooms sit idle. A schedule that looks efficient on paper can unravel as the day progresses.
For decades, hospitals have dealt with this kind of problem through a combination of expertise, rules and careful planning.
Now Hospital del Mar is asking whether a machine can do something humans struggle to do: consider thousands of interacting decisions at once.
As part of Q-CARE, a project funded by CDTI, Spain’s public business entity for promoting innovation and technological development, the hospital has spent the past year working with Qilimanjaro Quantum Tech on surgical scheduling. The immediate goal was not to put a quantum computer in the operating theatre.
It was to find out whether the problem could be formulated precisely enough to become a candidate for quantum optimisation in the future.
So far, yes.
Using hospital data and simulated scenarios, the team built an optimisation model that increased potential weekly surgical capacity from 267 procedures to 331 — a 27% increase — without adding operating rooms, staff or other resources.
That result is interesting in its own right. But for the quantum-computing industry, the more important achievement may be the model itself.
The scheduling problem hiding in plain sight
A surgical schedule sounds like a straightforward spreadsheet exercise.
There are patients waiting for procedures. There are operating rooms. There are surgeons and other clinical staff. Put the patients into the available slots and make the most of the week.
The result is a combinatorial problem: changing one decision can alter the feasibility and value of dozens of others.
Hospital del Mar already had a well-established manual planning process, designed around the robustness required in real clinical operations. The Qilimanjaro model did not simply attempt to automate that process. Instead, it expanded the number of possible combinations planners could consider.
Rather than focusing only on patients at the top of the waiting list, the model could evaluate a wider pool of candidates while simultaneously considering clinical, staffing, operating-room and recovery constraints.
That is where optimisation can expose opportunities that are difficult to see from inside a spreadsheet — or even from years of operational experience.
First, forget quantum
The team had to define the problem mathematically, assemble usable data and demonstrate that solving it could produce a meaningful result.
Working with data covering 24 types of surgery, including procedure duration, recovery requirements and waiting-list guarantees, the researchers built a stochastic optimisation model capable of planning an entire week.
It accounts for uncertainty in operation and recovery times while coordinating patient selection, room allocation and start times. Most importantly, it keeps the recovery unit within capacity throughout the schedule.
The resulting plan reached 331 procedures per week, compared with 267 under the hospital’s existing process.
The model also identified something less obvious: the dominant constraint was not necessarily the number of operating rooms or cleaning crews. It was the timing and coordination of recovery-bed capacity.
If the bottleneck is misunderstood, adding infrastructure may not solve the problem. An additional operating room, for example, is of limited value if the recovery unit cannot absorb the resulting patients.
“The preliminary results, obtained in simulated scenarios using the data available, indicate potential improvements of close to 30% in planning capacity compared with traditional approaches, and provide us with a much richer quantitative basis to analyse and evolve our resource-management methodology, although still within an exploratory setting,” said Rafa Luque, Nursing Coordinator of the Surgical Process & Sterilisation, Hospital del Mar.
The quantum question comes later
Today’s classical optimisation tools can handle the Hospital del Mar problem at its current scale.
So why quantum?
Classical computers can solve many optimisation problems remarkably well, but eventually there are cases where finding a sufficiently good solution within an operationally useful timeframe becomes difficult.
Quantum computers are being developed partly with such optimisation problems in mind. Whether they will deliver a practical advantage for this particular application – and when – remains an open research question.
They cannot wait for the hardware to arrive before figuring out what to do with it.
“Projects like this are essential to understanding where quantum technologies can create real value in optimization tasks. Since the scale and timing of this impact are still uncertain, we believe it is crucial to start early and collaborate closely with domain experts,” said Jordi Riu, Algorithm Portfolio PO, Qilimanjaro Quantum Tech.
A quantum strategy that starts without a quantum computer
Instead of starting with the technology and searching for a problem, Hospital del Mar started with a problem it already had.
Hospital del Mar and Qilimanjaro are now exploring how the optimisation framework could move closer to day-to-day clinical planning, including integration with real workflows and live patient data.
But the experiment has already established something valuable: the hospital knows what it wants to optimise, has a mathematical representation of the problem, has identified its key constraint and has a way to measure the result.
Not a futuristic machine sitting in a data centre.
A well-defined problem, backed by good data, waiting for a computer powerful enough to solve it.
