Roos van Andel has won the Best Thesis in Applied Mathematics Award 2026 in the master’s category. Her Business Analytics thesis uses mathematical models to help decide where to station mobile stroke units around Oslo and when to send them out.
Roos received the award on 4 November at Centrum Wiskunde & Informatica (CWI) in Amsterdam, where six finalists presented their research. The Best Thesis in Applied Mathematics Award recognises theses that apply mathematics to practical problems. It is organised each year by Math4NL and the Innovation Committee of Platform Wiskunde Nederland.
A mobile stroke unit is an ambulance equipped with a CT scanner and a specialist team. It can diagnose strokes and, where appropriate, start clot-dissolving treatment before a patient reaches hospital. These units are expensive and scarce, and some suspected strokes turn out to be something else. Sending a unit to one patient can leave it unavailable for another who might benefit more.
In her thesis, A Stochastic Optimization Model for Mobile Stroke Unit Deployment, Roos linked the choice of a base with decisions about individual emergency calls. Her model compares sending a conventional ambulance, sending the mobile stroke unit directly, or arranging for the two vehicles to meet on the way to hospital.
The model uses probabilities to account for the uncertain diagnosis and estimates how treatment delays affect the chance of recovery with little or no lasting disability. It balances the expected benefit to the current patient against the benefit future patients may lose while the unit is busy.
The Greater Oslo case study used historical records of 5,309 suspected-stroke calls in 2022 across 777 postal codes, together with population and stroke registry statistics, travel times and the locations of hospitals and ambulance stations. An earlier CWI and VU Amsterdam study, which placed a unit to maximise the number of patients it could reach within a set time, served as the benchmark.
For a single unit, Roos’ model recommends a conventional ambulance for almost half (47.5%) of the modelled calls, even when the mobile stroke unit is available. Accounting for future calls also changes the preferred base location, although both studies favour central Oslo. The location and the rules for sending out the unit therefore need to be planned together.
Roos also studied a model variant with several mobile stroke units, estimating how often a second unit would be free when the first was busy. The assumptions used to estimate this availability affected whether the model recommended a main unit with a backup or two units serving different areas. To compare the many possible deployments, she used optimisation algorithms that could rule out less effective combinations of locations. For some configurations, she could prove that no better arrangement existed within the model. Her final model examined how extra conventional ambulances could support a single mobile stroke unit.
The research was carried out during her graduation internship at CWI, supervised by Alessandro Zocca (VU Amsterdam) and Berend Markhorst (CWI), with René Bekker as second reader. Neurologist Karianne Larsen at Oslo University Hospital contributed data and medical expertise. Roos also presented her work at a practitioner symposium in Oslo.
The project reflects a strength of VU Amsterdam’s Master’s in Business Analytics: applying mathematics to problems with a direct societal purpose. Students learn optimisation and how to model uncertainty, then apply those methods during a six-month research internship at a company or research institute.
All three finalists in this year’s master’s category came from VU Amsterdam, two of them from Business Analytics. Fellow Business Analytics student Larson Beemster, also at CWI, was shortlisted for his thesis on hydrogen blending in natural gas networks.
Photo credits: Paul Roberts / CWI