Applying ArtificiaI Intelligence in Hospitals: The Origin of TAGNOS AI

Oct 30, 2019 8:00:00 AM / by Jag Padala, Chief Technology Officer

You may know what prompted the start of TAGNOS. However, do you know how TAGNOS began its AI journey for hospitals?

TAGNOS AI - an origin story

The inception of the Artifical Intelligence journey at Tagnos happened during a talk show on NPR about a master control program Amazon had created to run its huge warehouses.

Amazon called it Symphony. The program coordinated millions of orders that needed millions of things to be assembled and shipped. The program determined who needed to pack which material and where they needed to leave it so the next person could pick it up and assemble the order. A symphony running in splendid harmony.

 

Could the Same AI Principles Apply to a Hospital?

We were still a small team at the time and started to brainstorm.

  • How could we apply the same principles to a hospital?
  • Could we predict when equipment would be needed?
  • Could we predict when staff was needed?
  • Could we predict when a rom would need to be cleaned up?
  • Could we predict staff and equipment shortages?

Around this time we did a demo for a hospital in DC and showed them how, by using RTLS, we could show patient delays in real time. In reaction, the extremely sharp chief of peri-operative services asked us this question:

“It is fine that you can tell me where my delays are. Now given that there have been some delays, can you predict how the rest of the day is going to look like?” 

We captured everything we could think of onto a white board and realized we had inputs from EHR, RTLS, Labs, Radiology and other systems. We started with a couple of problems that were easy to identify such as how long would a surgery take and what equipment would be needed for a surgery and began putting together some algorithms that could predict surgery times. 

We simulated some data for the algorithms and the results were too good to be true. Since the simulated data was using some patterns the algorithms were just guessing the patterns.

The only true test would be actual hospital data. We reached out to our then tiny group of customers.

Could the Same AI Principles Apply to a Hospital?

Successful AI requires Lots of Real Hospital Data

Algorithms are data hungry: the more data, the better the results.

An enterprising CIO shared ten years of surgery data with us, a gold mine for engineers. As a result, we set about engineering features that would help us fine tune the estimates. After a few weeks, we got to a pretty decent accuracy in the prediction.

The real test would be then to demo the results to OR directors and see if the features we had come up with made sense. After a couple of demos to two different OR directors, we learned that most of the 25 features we came up with were acceptable to them. Although we discarded a few, most made it.

We now had models that could predict when a surgery ended so we could let the cleaning crew know when to go in, for example. 

To predict further, we needed additional algorithms.

One of the drawbacks we were working with was the remote nature of our team. Luckily a winter break in December combined with a visit to our Hyderabad office made possible a multi-week brain storming session on how to leverage more algorithms.

That, in turn, meant that we needed more hospital data! Luckily another hospital shared their data with us so we could feed those hungry AI algorithms. The results from the second hospital were encouraging.

 

Applying AI in Hospitals: Too Aspirational?

One of the problems of selling an aspirational solution such as AI is the difficulty of explaining that solution in its entirety to end users and also getting all the needed inputs to make the solution work.

There are tons of products that get discarded because they never get the right set of inputs or the right adopter; they're ahead of their time. Then, later, when conditions are right, you'll see reincarnations of those products. We didn't want to wait.

An investor suggested trimming the story down to make it simpler to explain to get a foot in the hospital door.

 

Timely Patient Alerts for the OR and the ER 

We realized that there was lot of value that could be provided to hospitals just by sending out timely alerts when patients go in and out of the OR and ER rooms. These alerts could coordinate cleaning and moving of equipment and patients.

This application of AI needed no manual input, didn't need a lot of EHR data and would work off of already automated RTLS data. Enter OR Orchestration Lite from TAGNOS.

The goal was to use the OR orchestration lite to get some hospitals engaged with the solution and then help them grow into the grander scheme.

When we pitched this, it immediately resonated with a number of our contacts. One of our favorite client sites saw the utility of the first and future phases of the deployment.

We started packaging this as a stand alone solution and were soon in a beta install.

 

Ongoing Hospital Feedback Leads to Continuous AI Improvement

One of the best aspects of working in a startup is the feedback you get during demos and customer meetings.

For example, during one of the demos, a hospital OR executive told us that we were only solving part of the orchestration problem by addressing the OR. The Orchestration needs to really start in the pre-op.

During another demo, there was a suggestion to open up the pre-op dashboard the day prior to surgery to use that as the whiteboard. And then another suggestion to setup a post-op dashboard to coordinate activities after the OR. 

All of these great suggestions feed an ever expanding and exciting roadmap.

Would You Like to Apply ArtificiaI Intelligence in Your Hospital?

Would You Like to Apply ArtificiaI Intelligence in Your Hospital?

So that is where we are right now - a little way away from saying “and the rest is history.” We have come some way from spaghetti on the wall to having a foot in the door in the AI orchestration game.

A lot of people have helped us get to this stage, talk show hosts, doctors, nurses, and directors to name just a few. 

I am sure all of us are looking forward to taking this to more customers and building a true orchestra product that can play a coordinated tune in hospitals. Let us know if you'd like to feed your hospital data to the hungry AI algorithms and orchestrating your OR or ED. 

We invite you to learn more.

Thanks for reading.

Download the Introduction to Tagnos Information Kit  & Learn How to Improve Workflow Efficiencies

 

About TAGNOS

TAGNOS is the future of clinical automation software solutions with Artificial Intelligence. It is the only platform offering predictive analytics utilizing machine learning and RTLS. This groundbreaking platform leverages historical patient data continuously and adjusts operational intelligence to provide sustainable improvement to both the patient experience and metrics.

TAGNOS provides clinical systems integration, customizable reporting, dashboards, alerts, critical communication with staff and family to improve turnaround times, supporting patient flow, workflow orchestration, and asset management. 

In the course of 13 months, hospitals see a 12.7% reduction in its overall cycle time - saving an average of 40 minutes from each case and over $1.6M per year - more than 11x the typical investment.

Click for a Demo