Pacmed builds AI software for the intensive care unit

In an intensive care unit, decisions about discharge or continued treatment always carry the risk of being made too late or too early. Pacmed, founded in Amsterdam, develops software that supports ICU physicians with predictions based on patient data. The company's focus is not on replacing clinical judgment, but on supplementing it with objective, explainable insights.

The idea emerged in 2014 during the Nationale DenkTank, an annual programme in which young professionals tackle societal challenges. Wouter Kroese, who studied medicine at Utrecht University and also has a computational background, co-founded the company together with Hidde Hovenkamp and Willem Herter. Pacmed has grown into a team of more than thirty employees, spanning physicians, data scientists, software developers and compliance experts.

In 2021, the company raised a Series A round of €24 million. In the summer of 2023, a new investor came on board. Pacmed expects to require a further funding round before reaching profitability.

From fifteen years of ICU data to a certified medical device

The core product is called Pacmed Critical. The model was trained on fifteen years of data from the ICU at VUmc in Amsterdam, and was developed in collaboration with Amsterdam UMC. In 2020, Pacmed was among the first organisations worldwide to deploy AI based on electronic health record data as a medical device. The platform is MDR-certified and works with all commonly used electronic health record systems.

One of its features, Pacmed Critical Discharge, calculates the probability that a patient will die or be readmitted within seven days of being discharged from the ICU. The system indicates which clinical factors contribute most to that risk, enabling physicians to follow and assess the reasoning. This transparency is a deliberate design choice: hospitals working with medical AI are required to be able to explain its operation to patients and regulators.

Pacmed Critical is now in use at Maasstad Ziekenhuis in Rotterdam, OLVG and Amsterdam UMC in Amsterdam, Catharina Ziekenhuis in Eindhoven, and Medisch Spectrum Twente. OLVG and Maasstad took out their first full licences in 2023.

Capacity planning beyond the ICU

Alongside its ICU software, Pacmed is developing broader capacity tools. Pacmed Capacity Monitor predicts how many patients are expected in the coming days across departments such as the emergency department, acute admissions ward and the ICU, together with the associated care intensity. Pacmed Clinical Logistics Manager provides real-time insight into patient inflow and outflow to make logistical bottlenecks visible earlier.

This expansion aligns with partnerships Pacmed has formed in recent years. BovenIJ ziekenhuis in Amsterdam is working with Pacmed on capacity management outside the ICU. The Health AI en Data Alliantie Limburg aims to deploy AI more rapidly in the region with Pacmed. In November 2025, Gelre ziekenhuizen signed an agreement for implementation in the ICU. Earlier that year, Pacmed and software company ChipSoft announced a partnership to integrate the AI modules into HiX, the electronic health record used by many Dutch hospitals.

Pacmed also collaborates with health insurers CZ and Zilveren Kruis, and with the Santeon group, a collaborative of seven Dutch hospitals, to validate and more broadly deploy new technology.

Causal inference as the next step

Predicting what will happen is one thing; understanding what a treatment causes is a step further. Pacmed is working with the Department of Medical Informatics at Amsterdam UMC and the University of Amsterdam on research into causal inference. The aim is to model the effect of a specific treatment decision on a patient's outcome, independent of the confounding factors that are always present in observational data. This project started in 2022 and runs for at least three years, partly funded through the European Regional Development Fund.

Hospitals already using Pacmed provide a relevant testbed for this work: when the software runs daily in clinical workflows, it creates a controlled environment in which to validate new models. Whether and when causal models will make it into a production version has not yet been publicly announced.

Pacmed operates in a segment of the medical AI market that is relatively small in the Netherlands: software that is certified as a medical device and is actually operational in multiple hospitals. For other European founders and investors in health AI, Pacmed illustrates how long that journey takes. From the initial idea in 2014 to certified deployment at a handful of hospitals required nearly a decade of clinical collaboration, validation research and regulatory lead times. That timeline, and the associated funding requirements, is characteristic of anyone who wants to pursue AI seriously in the regulated healthcare sector.