Hey Reader,

AI is only as intelligent as the data that is used to train it. In the rush toward AI in healthcare, organizations often overlook the fundamental importance of quality data. While many healthcare systems are playing defense with their AI strategy, success requires active participation and preparation. The key lies not in waiting for the perfect application, but in building a strong foundation with reliable location data that can evolve with our technological capabilities.

TLDR:

Healthcare organizations can prepare for AI integration by focusing on quality location data collection through strategic RTLS deployment and agentic design principles, turning current operational data into future AI gold.

Key Concepts:

Let's break down some essential terms that will help us understand this journey:

Care Traffic Control - An approach to optimizing mobile workflows in healthcare settings, encompassing everything from patient flow to asset management.

Planning Horizons - The time required for an AI agent's sense-think-act cycle. This can range from real-time monitoring to long-term pattern analysis.

Agentic Design - A modular approach to workflows are structured as agents that run through cycles of sensing-thinking-acting. Agentic design is human-in-the-loop meaning components can be performed by humans or machines.

Digital Twin vs. Digital Shadow - While a digital twin provides real-time system representation, a digital shadow offers a lighter-weight alternative that captures essential data without the complexity of real-time processing.

Detailed Explanation:

The Path to Better Data

Healthcare organizations often struggle with data quality issues. Consider the example of patient transport systems where staff might use only 3-4 delay codes out of 24 available options, simply because it's easier. This simplification leads to incomplete data and potentially misguided problem-solving efforts. Utilization rates for medical equipment are driven by request information and stocking rates instead of the time the device was actually being used. This goes on and on.

The Power of Location in Healthcare

Location data allows us to accurately monitor the nominal flow in the process and detect delays. Location-based systems also provide a unique opportunity to detect anomalies at scale. Through the Theory of Constraints, we can identify and instrument key bottlenecks to implement sustainable solutions. This approach aligns perfectly with organizational performance improvement plans.

A New Approach to RTLS Implementation

The digital shadow is a stepping stone to a higher fidelity digital twin. Synchronizing with the real world can be done with RTLS but incremental approaches to implementation should be considered. Instead of traditional wall-to-wall RTLS installation, consider a "Lean Infrastructure" approach:

1) Start with Strategic Sensing

Focus RTLS deployment on specific locations where state changes occur in your workflows. For example, when tracking medical devices, identify key points in their usage cycle: storage, patient rooms, cleaning areas, and central storage. Pick the locations that are the most likely to show the bottleneck. Perhaps a PAR stocking cycle is struggling because devices are not being put in the soiled utility for retrieval.

2) Event-Based Tracking

Transform simple location data into meaningful events by focusing on state changes. When an IV pump moves from a patient room to a soiled utility closet, that's not just movement - it's a workflow state change that provides valuable insight.

3) Long-Term Pattern Analysis

Analyze data over extended periods (like monthly segments) to identify both normal patterns and anomalies. This approach helps pinpoint specific process breakdowns by department, time, and frequency. It also does not require expensive software.

4) Correct the Bottleneck

Find the root cause of the bottleneck and fix it. This could be a process change or some technology. The key at this stage is having the sensing in place to be able to see the impact of an corrective actions with immediate feedback. This immediate feedback translates to sustained improvement, over the longer planning horizon.

5) Move on to the Next Bottleneck

Continue the analysis to determine where additional bottlenecks are and if possible measure the delay so that the priorities can be assigned to them. If there is enough delay to target for improvement the process can stop.

Example:

The above process is used to address specific performance measures. For instance, if the performance measure is "rate of arrival for dirty IV pumps" into the soiled utility location.

1. Put sensors only in the soiled utility closets (lean infrastructure)
2. Capture the arrival events for the pumps into the closets
3. Collect the data for one month and find the patterns where there is a low rate of arrival. Present the data to the departments that are struggling with the process.
4. Address the specific locations where the pumps have a low rate of arrival. Continue the process of collecting the data and addressing the low arrival rates until the arrival rates are satisfactory.
5. Move on the next bottleneck. For instance, an overly long duration for the IV pump retrieval rounds.

Continuing with the above example: Let's say the next iteration the retrieval process shows a long duration because the transporters are visiting empty soiled utility closets. This is wasted effort and delay. Now the sensors can see that there are pumps in the closet and the transporter can avoid the empty closets. The cycle repeats from there.

Agentic Design

We get started by thinking of a workflow as an agent and people are actors in that workflow. Agentic design uses a modeling technique called PEAS.

Performance Measures - Measures that align with the goals and outcomes for the agent.

Environment - People, Property and Places that are part of the workflow (agent).

Actuators - The actions or process that is performed that targets the improvement we want in the performance measures.

Sensors - The sensing that is necessary for the control loops we want to install to sustain the improvement.

Example:

• Performance Measure - Rate of arrival
• Environment - IV Pumps, Nurses, Transporters, soiled utility closets
• Actuators - Nurses put IV Pumps in the soiled utility closet
• Sensors - RTLS sensors in the Soiled Utility closets

This is rather straightforward with one performance measure, but a single process might have many performance measures. When expanded across an entire process PEAS becomes an organization tool that is critical to continued improvement and automation. Processes are also interdependent so handoffs are included in the horizontal nature of operations.

There is a vertical nature to performance measures when they are hierarchical. The pump retrieval process is subordinate to pump utilization rate so all of the performance measure that we can gather to build up to utilization rate allow the improvement cycles to methodical and organized. When can represent this diagrammatically in a tree structure.

Examining the Data

In each iteration we are looking are carefully examining data that is used in the performance measures as we use PEAS to model the process. This will expose issues with the data and intelligent decisions can be made as to the value of correcting them with sensors. Because these are tools that are used in automation, good data is is necessary for any control loops that are put in place.

Future-Proofing with AI in Mind

As we these agentic design tools, we're not just solving today's problems - we're preparing for an AI-enabled future. These cycles are building quality location data that becomes the foundation for:

• Training AI agents with reliable, contextual information
• Building more sophisticated digital twins
• Enabling predictive and prescriptive analytics
• Supporting human-in-the-loop systems that can gradually incorporate more AI capabilities

Key Takeaways:

• Focus on collecting quality data at key workflow points rather than attempting comprehensive coverage
• Use state changes and events to transform raw location data into actionable insights
• Implement a modular approach that can evolve from human-centered to AI-augmented processes
• Build confidence in AI systems through demonstrably reliable underlying data

Conclusion:

The journey to AI-enabled healthcare doesn't start with artificial intelligence - it starts with intelligent data collection. By focusing on quality location data and meaningful events, healthcare organizations can build a foundation that delivers immediate operational benefits while preparing for an AI-enabled future. Visit Why Where Matters regularly for more insights on transforming healthcare through location intelligence.

Want to learn more about implementing these strategies in your organization? Schedule a consultation or connect with us on LinkedIn to continue the conversation.

Paul E Zieske ​Location Based Services Consulting ​​​