Predictive Patient Flow: How AI and a Modern Data Stack Can Optimize Hospital Operations

By Nogamy's Architecture Team

In today's rapidly evolving healthcare landscape, hospitals are facing unprecedented operational challenges. A surge in patient demand, coupled with persistent staffing shortages and the complexities of managing chronic diseases, has pushed many healthcare systems to their limits. One of the most critical, yet often overlooked, areas impacted by these pressures is patient flow– the movement of patients through a healthcare facility from admission to discharge. Inefficient patient flow leads to overcrowded emergency departments (EDs), long wait times, and a cascade of negative consequences that affect patient outcomes, staff morale, and the financial health of the organization.

But what if we could predict patient demand before it happens? What if we could anticipate bottlenecks and proactively allocate resources to ensure a smoother, more efficient patient journey? This is no longer a far-fetched vision. By harnessing the power of Artificial Intelligence (AI) and a modern data stack, hospitals can transition from a reactive to a proactive operational model. This blog post will explore how a combination of cutting-edge technologies- including Boomi, Rivery, dbt, and AWS– can be used to build a powerful predictive patient flow solution, transforming hospital operations and delivering a higher standard of care.

The Challenge: Why Patient Flow is a System-Wide Problem

For years, the issue of overcrowded Emergency Departments (EDs) has been a persistent headline and a major source of concern for both patients and healthcare providers. However, the problem is not confined to the ED alone. As a comprehensive review in BMC Health Services Research highlights, "ED overcrowding is not solely an ED problem but rather reflects dysfunction throughout the entire patient journey" [1]. This system-wide dysfunction, often referred to as poor patient flow, creates a domino effect with far-reaching consequences.

At its core, the patient flow problem is a classic case of a mismatch between supply (hospital resources, including beds, staff, and equipment) and demand (patient needs). This mismatch is exacerbated by a lack of real-time visibility and predictive insight into patient arrivals, bed availability, and discharge readiness. Without the ability to anticipate demand and proactively manage resources, hospitals are left in a constant state of reaction, fighting fires instead of preventing them. Most hospitals today face this challenge, with supply-demand imbalances resulting in delays, staffing gaps, and inefficient hospital ward utilization [1].

The Solution: A Modern Data Stack for Predictive Analytics

To effectively tackle the patient flow challenge, hospitals need to move beyond traditional, siloed data systems and embrace a modern, integrated approach. A modern data stack provides the foundation for building a robust predictive analytics solution that can deliver the real-time insights and forecasting capabilities needed to optimize hospital operations. Here's how we can architect such a solution using a combination of best-in-class technologies:

This architecture is designed to be modular, scalable, and cloud-native, allowing for flexibility and cost-effectiveness. Let's examine the role of each component and how they work together to create a powerful predictive analytics platform.

Boomi: The Enterprise Integration Backbone

Boomi serves as the foundational integration layer that connects the disparate systems across the hospital enterprise. As a leading Integration Platform as a Service (iPaaS), Boomi excels at connecting various healthcare systems, including Electronic Health Records (EHR), Admission/Discharge/Transfer (ADT) systems, laboratory information systems, radiology systems, and bed management platforms. In December 2024, Boomi further strengthened its position by acquiring Rivery, expanding its data management capabilities to include modern data integration for analytics and AI use cases [2].

Boomi's healthcare-specific integration solutions are designed with security and compliance in mind, ensuring that sensitive patient data is handled according to HIPAA regulations and other healthcare standards. By providing a unified platform for application integration, API management, and workflow automation, Boomi creates a seamless flow of information across the entire healthcare ecosystem. This real-time connectivity is essential for building a comprehensive view of patient flow, as it ensures that all relevant data is captured and made available for analysis.

Rivery (Boomi Data Integration): Modern Data Pipelines

Following Boomi's acquisition of Rivery in late 2024, the combined platform now offers powerful capabilities for moving large volumes of data to cloud data warehouses and data lakes. Rivery, now known as Boomi Data Integration, specializes in modern data integration using ELT (Extract, Load, Transform), ETL (Extract, Transform, Load), and Change Data Capture (CDC) processes [2]. This is particularly important in healthcare, where data volumes are massive and growing exponentially.

Rivery's cloud-native, low-code platform makes it easy to build end-to-end data pipelines that can handle both batch and real-time data processing. Its embedded AI engine and growing library of integration templates accelerate the development process, with some customers reporting a 7.5X improvement in time to value for data integration projects [2]. For our predictive patient flow solution, Rivery extracts data from the various hospital systems connected via Boomi and loads it into a centralized cloud data warehouse, creating a single source of truth for analytics. As a Snowflake Premier Technology Partner, Rivery provides seamless integration with Snowflake's Data Cloud, enabling healthcare organizations to leverage the full power of cloud-native data warehousing [3].

Snowflake: The Cloud Data Platform

Snowflake serves as the modern cloud data platform that centralizes and stores all patient flow data in a secure, HIPAA-compliant environment. As a purpose-built cloud data warehouse, Snowflake's unique architecture separates compute from storage, enabling healthcare organizations to scale resources independently and handle massive data volumes without performance degradation [4]. This is particularly crucial in healthcare environments where data grows exponentially and query patterns vary significantly throughout the day.

Snowflake's healthcare-specific capabilities make it an ideal choice for patient flow analytics. The platform provides robust security features including end-to-end encryption, role-based access controls, and comprehensive audit logging to ensure compliance with HIPAA and other healthcare regulations [5]. Its ability to handle semi-structured data alongside traditional relational data means that diverse healthcare data sources- from structured EHR records to unstructured clinical notes- can be unified in a single platform.

Within Snowflake, we maintain multiple layers of data following the medallion architecture: a bronze layer for raw data as it arrives from Rivery, a silver layer for cleaned and validated data, and finally, a gold layer containing analytics-ready datasets prepared by dbt. This layered approach ensures data quality while also providing flexibility for different types of analysis. Snowflake's near-zero maintenance requirements and automatic optimization features allow healthcare IT teams to focus on deriving insights rather than managing infrastructure.

dbt: Analytics Engineering for Data Quality

Data is only useful if it's clean, reliable, and well-documented. This is where dbt (data build tool) comes in. As an analytics engineering tool, dbt allows us to transform the raw data in Snowflake into analytics-ready datasets using SQL-based models. A review of machine learning applications in patient flow emphasizes the importance of having clean, structured data as a foundation for predictive models [6]. dbt excels at this task, and its tight integration with Snowflake enables efficient, scalable transformations [7].

Using dbt, we can build a clear and logical data structure that models the patient flow domain. This includes creating fact tables such as patient_admissions, ed_visits, bed_occupancy, and discharge_events, as well as dimension tables for patients, providers, departments, and time. Each of these models is version-controlled, tested for data quality, and automatically documented, ensuring that the data pipeline is transparent and maintainable.

Crucially, dbt also creates the feature tables needed for machine learning. By pre-aggregating data and calculating derived metrics, dbt prepares the datasets that Amazon SageMaker will use to train predictive models, bridging the gap between raw operational data and AI-ready features.

Amazon SageMaker: AI-Powered Predictions

Amazon SageMaker is the AI and machine learning engine of our solution. Using the clean, feature-rich datasets prepared by dbt in Snowflake, we can leverage SageMaker to build, train, and deploy a variety of predictive models. Research in the field of machine learning for patient flow has identified several key application areas, including prediction of demand on healthcare institutions, forecasting of resource needs for patient transfers, prediction of inpatient resource requirements, and estimation of length-of-stay and discharge timing [6].

For our predictive patient flow solution, we can implement models such as:

ED Admission Forecasting: Time-series models that predict the number of patient arrivals at the emergency department for the next 24-72 hours, based on historical patterns, time of day, day of week, and seasonal trends.

Bed Demand Prediction: Regression models that forecast the number of beds needed by department over the next 7 days, taking into account current occupancy, scheduled admissions, ED queue length, and predicted discharges.

Length-of-Stay Estimation: Models that predict how long each admitted patient will remain in the hospital, based on demographics, diagnosis, comorbidities, and treatment plan.

SageMaker provides a complete environment for the entire machine learning lifecycle, from data exploration and model training to deployment and monitoring. Once trained, models can be deployed as real-time inference endpoints, allowing the hospital to generate predictions on demand and integrate them into operational workflows.

BI Dashboards: Making Insights Actionable

The final piece of the puzzle is to make the insights from our predictive models accessible to the people who need them most- hospital administrators, department managers, and clinical staff. Interactive BI dashboards, built using tools like Amazon QuickSight or Tableau, provide real-time monitoring of patient flow metrics, predictive alerts, and actionable recommendations.

These dashboards can display current ED wait times alongside predicted patient arrivals for the next few hours, allowing staff to proactively adjust resources. They can show bed occupancy by department with forecasts of when capacity constraints are likely to occur, enabling proactive discharge planning. And they can highlight patients who are predicted to have longer-than-average lengths of stay, prompting early intervention from care coordination teams.

Putting It All Together: Key Use Cases

The true power of this architecture lies in its ability to support a wide range of predictive use cases that can have a direct and measurable impact on hospital operations. Let's explore a few examples of how this solution can address real-world challenges.

Use Case 1: Predicting Emergency Department Admissions

Emergency departments are often the entry point for unplanned hospital admissions, and their unpredictable nature makes them particularly challenging to manage. However, by analyzing historical ED visit data along with external factors like time of day, day of the week, local events, weather patterns, and seasonal trends, we can build time-series forecasting models in Amazon SageMaker that predict the number of patient arrivals for the next 24-72 hours with a high degree of accuracy.

Armed with this information, hospital administrators can take proactive steps to optimize operations. They can adjust nursing and physician schedules to match anticipated demand, ensuring that adequate staff are available during predicted surge periods. They can prepare the necessary beds, equipment, and supplies in advance, reducing the scramble that often occurs when patient volumes spike unexpectedly. And by anticipating surges in patient volume, the ED can be better prepared to handle the influx, leading to shorter wait times and a better patient experience.

Use Case 2: Forecasting Inpatient Bed Demand

One of the primary causes of ED access block- when admitted patients cannot leave the ED due to a lack of available inpatient beds- is poor visibility into bed availability across the hospital. To address this, we can build a model that forecasts bed demand across different departments for the coming days. This model takes into account current bed occupancy, scheduled admissions (such as elective surgeries), the number of patients currently in the ED awaiting admission, and the predicted length-of-stay for current inpatients.

The resulting forecast allows the hospital to take several proactive actions. Care teams can identify patients who are medically ready for discharge and prioritize their discharge planning to free up beds. The surgical scheduling team can adjust the schedule of elective admissions to align with predicted bed availability, avoiding situations where planned surgeries must be cancelled due to lack of capacity. And by anticipating bed shortages several days in advance, the hospital can take early action to prevent access block and keep patients flowing smoothly through the system.

Use Case 3: Estimating Patient Length-of-Stay

Predicting how long a patient will stay in the hospital is crucial for effective discharge planning and resource allocation. Using patient demographics, primary diagnosis, comorbidities, treatment plan data, and historical length-of-stay patterns, we can train regression or survival analysis models to predict the LOS for each admitted patient at the time of admission.

This prediction enables the care team to initiate discharge planning early in the patient's stay, rather than waiting until the patient is medically ready to leave. By knowing in advance which patients are likely to require post-acute care services such as rehabilitation or home health, the care coordination team can begin arranging these services well ahead of discharge, ensuring a smooth transition and reducing the risk of discharge delays. Accurate LOS predictions also improve bed turnover efficiency, as the hospital can better anticipate when beds will become available for new patients.

The Nogamy Advantage: Your Partner in Data-Driven Healthcare Transformation

Building a predictive patient flow solution requires more than just technology; it requires deep expertise in data architecture, analytics engineering, and the unique challenges of the healthcare industry. At Nogamy, we bring a wealth of experience in all of these areas. Our team of data architects and engineers is proficient in the entire modern data stack, from enterprise integration with platforms like Boomi to data pipeline development with Rivery (now Boomi Data Integration), cloud data warehousing with Snowflake, data transformation with dbt, and advanced analytics and machine learning on cloud platforms like AWS.

We understand that every hospital is different, with its own unique set of challenges, legacy systems, and data maturity levels. That's why we take a collaborative, vendor-agnostic approach, working closely with our clients to design and implement a solution that is tailored to their specific needs and constraints. Our focus is not just on delivering a technology platform, but on building a sustainable data culture that empowers our clients to make smarter, data-driven decisions long into the future.

Whether you're just beginning your data journey or looking to enhance an existing analytics capability, Nogamy can help you navigate the complexities of modern healthcare data and unlock the full potential of predictive analytics.

Conclusion: The Future of Hospital Operations is Proactive

The challenges facing modern healthcare systems are complex and multifaceted, but they are not insurmountable. By embracing a proactive, data-driven approach to patient flow management, hospitals can not only alleviate the symptoms of overcrowding and inefficiency but also address the root causes of these problems. The modern data stack, combining the integration power of Boomi, the data movement capabilities of Rivery (Boomi Data Integration), the cloud-native data platform of Snowflake, the transformation rigor of dbt, and the AI capabilities of AWS, provides a powerful toolkit for building the predictive analytics solutions that are needed to make this transition.

By integrating data from across the patient journey- from the moment a patient arrives at the emergency department to the time they are discharged- we can create a holistic, real-time view of patient flow. And by applying the power of AI and machine learning to this data, we can turn it into actionable insights that transform hospital operations, improve patient outcomes, reduce staff burnout, and create a better, more sustainable healthcare system for all.

The future of hospital operations is not reactive- it is proactive, predictive, and powered by data. The question is not whether hospitals will adopt these technologies, but when. Those who move quickly to embrace this transformation will be best positioned to meet the challenges of tomorrow's healthcare landscape.

Ready to take the first step towards a more data-driven future for your hospital? Contact us today to learn how Nogamy can help you build your own predictive patient flow solution.

References

[1] Samadbeik, M., Staib, A., Boyle, J., Khanna, S., Bosley, E., Bodnar, D., … & Sullivan, C. (2024). Patient flow in emergency departments: a comprehensive umbrella review of solutions and challenges across the health system. BMC Health Services Research, 24(1), 274. https://pmc.ncbi.nlm.nih.gov/articles/PMC10913567/

[2] Macosky, E. (2024, December 19). Boomi Strengthens Modern Data Movement Capabilities With Rivery Acquisition. Boomi Blog. https://boomi.com/blog/boomi-strengthens-data-movement-rivery/

[3] Rivery Technologies Inc. Snowflake Partners. https://www.snowflake.com/en/why-snowflake/partners/all-partners/rivery/

[4] Mastering Snowflake for Healthcare Data: Real-World Use Cases. Medium. https://medium.com/@anand.sangeeta.0110/mastering-snowflake-for-healthcare-data-real-world-use-cases-from-medicare-to-oncology-d395729d2f01

[5] Snowflake Healthcare Data Cloud. Folio3. https://data.folio3.com/blog/snowflake-healthcare/

[6] El-Bouri, R., Taylor, T., Youssef, A., Zhu, T., & Clifton, D. A. (2021). Machine learning in patient flow: a review. Progress in Biomedical Engineering, 3(2), 022002. https://pmc.ncbi.nlm.nih.gov/articles/PMC8559147/

[7] Accelerate your Snowflake analytics with dbt. dbt Labs. https://www.getdbt.com/data-platforms/snowflake