Using AI and ML in predictive analytics for bed demand forecasting
Bed capacity management is of critical importance to health systems, impacting patient care and safety, operational efficiency, system sustainability and financial performance. Efforts to improve and streamline management often are isolated to regions within the center and may lead to suboptimal resource utilization, inconsistent patient care, and inefficiencies between care units for transfers and other care coordination.
Assessment of end-to-end bed demand management globally from admission to discharge eliminates many of the unintended consequences of localized optimization efforts. Froedtert Health identified improving capacity management as an important and targetable goal that could be achieved through AI, machine learning and data analytics approaches.
Understanding and dissecting patient flow and its sources allowed the team to create a suite of predictive tools designed specifically for the care coordination center. Froedtert Health was able to improve patient care, operationalize key performance indicators and streamline operations through more effective staff deployment and utilization and by pre-emptively responding to anticipated changes in patient bed demand.
This led to optimized allocation of resources, improved patient flow, better coordination between departments and cost savings.
Ravi Teja Karri is a machine learning engineer at Froedtert ThedaCare Health. He and two colleagues will be speaking on these achievements at HIMSS25 in a session titled “Improving Capacity Planning and Bed Demand Forecasting Using Machine Learning.” We interviewed Karri to get a sneak peek at what he plans to discuss in March at HIMSS25 during his session.
Q. What is the overarching theme of your session, and why is it especially relevant to healthcare and health IT today?
A. The overarching theme of our session is focused on improving hospital capacity management and bed demand forecasting through the application of artificial intelligence and machine learning techniques. This topic is increasingly relevant in healthcare as hospitals face unpredictable changes in patient volume.
Seasonal surges, unplanned admissions and fluctuating patient needs make it challenging to maintain an optimal allocation of resources. Leveraging AI and ML to predict bed demand and patient flow enables hospitals to optimize staffing, allocate beds and streamline operations, resulting in enhanced patient care and overall efficiency.
Our session also will explore how healthcare organizations can leverage AI and ML to transform processes into anticipatory workflows rather than reactive ones. This proactive approach enables more accurate forecasting of patient volumes and better interdepartmental coordination, ultimately enhancing patient experience through more efficient resource allocation and timely care delivery.
Integrating these predictive models into daily operations enables healthcare organizations to better anticipate demand fluctuations, minimize overcrowding risks and enhance interdepartmental coordination.
Q. You are focusing on AI and ML, important technologies in healthcare today. How are they being used in healthcare in the context of your session’s focus and content?
A. Our session focuses on artificial intelligence and machine learning technologies, specifically their application in predictive analytics for bed demand forecasting and capacity management in hospitals. ML models are designed to analyze large datasets, including historical patient admissions, discharge trends, seasonal illness patterns and other factors, to forecast future hospital capacity needs.
We will explore how these models can predict patient flow and bed demand, enabling healthcare organizations to make more informed decisions about resource allocation, staffing and patient care management.
These predictive models use algorithms to identify patterns and trends in patient admissions, length of stay and discharge rates, enabling hospitals to forecast fluctuations in demand with a high degree of accuracy. ML integrates data from multiple sources, including emergency departments, surgical units and outpatient care, to provide a comprehensive view of organizational capacity.
This analysis helps hospital leadership and care coordinators to anticipate surges in bed demand – like those experienced during flu seasons or following natural disasters – and plan effectively to ensure that resources are available when needed most. By implementing these technologies, healthcare institutions can transition from a reactive approach to a more proactive and anticipatory model of patient flow management.
In our session, we will examine how machine learning can be effectively applied in healthcare to predict bed demand and enhance capacity management. By analyzing historical data such as patient admission rates, discharge patterns and seasonal trends, ML models can forecast hospital capacity needs.
These predictions enable healthcare organizations to optimize resource allocation, plan staffing requirements and deliver improved patient care, enabling a proactive rather than reactive approach to operations.
We also will discuss how these ML models can be integrated into healthcare workflows, transforming predictions into action for hospital staff. Rather than remaining in experimental environments or isolated tools, the predictions are processed, stored and made available for decision making through business intelligence platforms.
These BI tools enable healthcare staff to access insights for effective planning, such as allocating beds, managing staffing and coordinating patient discharges, ultimately improving operational efficiency and patient outcomes.
Q. What is one of the various takeaways you hope attendees will leave your session with and be able to apply when they return home to their organizations?
A. A key takeaway we hope attendees will gain from our session is the knowledge to implement machine learning-based predictive analytics tools to enhance their own hospital’s capacity management.
Attendees will discover how predictive models can accurately forecast bed demand and identify potential bottlenecks in patient flow before they occur. These insights will empower leaders to make data-driven decisions, allocate resources more efficiently, and avoid overburdening units or staff during peak periods.
By using this toolkit, healthcare providers can minimize last-minute staffing adjustments, optimize bed utilization, and ensure patient care remains uninterrupted during periods of high demand. Predicting patient flow across the entire hospital, rather than in isolated units, allows for optimized resource allocation across departments and minimization in delays caused by mismatches between patient demand and available resources.
This will foster better communication between clinical teams and operational leaders, resulting in smoother transitions between patient care stages and improved overall patient experience.
Ravi Teja Karri’s session, “Improving Capacity Planning and Bed Demand Forecasting Using Machine Learning,” is scheduled for Tuesday, March 4, at 10:15 a.m. at HIMSS25 in Las Vegas.
Follow Bill’s HIT coverage on LinkedIn: Bill Siwicki
Email him: [email protected]
Healthcare IT News is a HIMSS Media publication
link