Capacity planning with AI based on individual patient data saves OLVG team leaders and planners at least one hour per day

Steven Jonis - Senior Business Developer Pacmed

Date:

November 13, 2025

Reading time:

4 min

On August 14, Pacmed launched the Capacity Monitor: an AI application specifically developed for short-term (up to 5 days ahead) adjustments to staff planning in the Intensive Care Unit, based on real-time data from individual patients.
The application provides continuous and immediate insights into patient occupancy for the coming days, giving the ICU—and all departments within the acute care chain—greater control over staffing capacity. The application was developed and tested in close collaboration with OLVG.

In this article, we speak with Rosnie Jankipersadsing, working team leader of the Intensive Care Unit at OLVG, who played a key role in the development.

Rosnie Jankipersadsing – ICU Team Leader at OLVG

‍Rosnie Jankipersadsing has been with OLVG for nine years, seven of which as a working ICU team leader. During the COVID pandemic, Rosnie played a crucial role in large-scale ICU expansion. As a result, capacity planning and workforce management have become key priorities in her career.

‍

Article content — Rosnie on the Intensive Care department of the OLVG hospital

Complex context makes AI prediction especially valuable

‍Capacity planning in the ICU is complex due to unpredictable admissions, length of stay, care demand, and care intensity. To guarantee sufficient admission capacity, rosters are often scheduled generously. However, this frequently results in more staff being present than necessary.

Team leader Rosnie confirms this:

“Team leaders, planners, and coordinating nurses make countless decisions every day. We weigh factors such as the number of admitted patients, care intensity, scheduled surgeries, and available capacity for emergency admissions. On top of that, we must schedule three months ahead according to the collective labor agreement. By the time the month actually begins, many changes have already been made to the roster. Due to uncertainty about patient occupancy and ad-hoc changes in staff planning, we often choose staffing levels based on maximum bed capacity. While this helps prevent capacity issues, it also regularly results in unnecessary shifts being scheduled.”

Pacmed Capacity Monitor predicts over- and undercapacity 5 days ahead

With Pacmed’s Capacity Monitor for the ICU, teams can predict patient numbers and optimal nurse staffing per shift up to five days in advance. This enables planners and team leaders to look ahead and make timely adjustments.

The Capacity Monitor stands out by using real-time data from individual patients, enabling AI-based predictions of expected ICU length of stay for patients already admitted and those scheduled for elective surgery. It also predicts admissions from the Emergency Department and hospital wards to the ICU. This results in a complete and up-to-date picture of expected ICU occupancy and associated care intensity for the coming days.

Daily use after testing phase

The Capacity Monitor has been live at OLVG since late 2024. Rosnie has used it daily since April 2025. Especially last June, when the roster was already tight and became even tighter due to various circumstances, the application provided crucial support.

Examples of decisions supported by the application:

Canceling an internal shift
Moving a shift between days or locations
Choosing not to fill a shift
Sometimes deciding that doing nothing is the best option

The impact is visible not only in rosters but also in workflows. Planners can work more independently, and fewer coordination meetings are needed with team leaders and ICU managers. Rosnie estimates this saves at least an hour per day:

‍
“Several times a week, our capacity planner and I make decisions we wouldn’t have made without the Capacity Monitor. It creates clarity and calm, and helps me make faster choices. Moreover, the tool promotes uniformity because every user sees the same objective data on which decisions can be based.”

‍

Building trust in AI

For Rosnie, this was her first AI project. She describes herself as not a heavy user of ChatGPT or similar tools, so building trust in the predictions was crucial:

‍
“I make decisions that directly impact staff and patients, so at first it felt daunting. But I quickly noticed the predictions were accurate. The clear design of the application helped enormously.” — Rosnie Jankipersadsing

An example: When a senior nurse (coordinating nurse) was concerned about a weekend with unusually few scheduled shifts, Rosnie showed in one overview that the expected care demand would be low. Although it felt risky, the senior nurse trusted the prediction. After the weekend, the forecast proved completely correct.