Featured in Health IT Analytics.
In an industry as high-stakes and uncertain as healthcare, predictive analytics tools and patient identifiers have emerged as viable solutions for seamless, comprehensive care delivery.
These tactics can help organizations better understand and anticipate patients’ needs, enabling them to deliver necessary services and improve patient outcomes.
At Texas Heath Resources, leaders were able to create patient IDs for each of its seven million patients, as well as develop predictive analytics tools for a more connected care experience.
“We’ve been able to connect more than 20 data sources together and track the care journeys of different patients,” said Michael Parris, vice president of data integration and analytics at Texas Health Resources, told HealthITAnalytics.
“For example, if we have a high-value procedure, we can look at everyone who had that procedure and then track them back from the last ten to 15 touch points and understand where they came in. Did they first visit us through the ED? Were they coming through clinics? Did they come through an urgent care, or were they referred over to us through a specialist? We can see where they journey through.”
The health system is also able to forecast patients’ needs based on the care experiences of similar patients, Parris noted.
“For people that probably would have ended up getting this procedure or may be on the journey to undergoing this procedure, we can figure out how we can make their experience and their journey through our system better. We can improve the process so that they either get healthy and don’t need the procedure, or we do the procedure sooner and shorten their length of stay,” he said.
“We’re able to look at somebody and say, based off your own personal history and other people’s history, we can predict which provider you’re going to want to see in the next three months. So, in May we know who you need to see in July. Because we’ve connected these data sources from different places, we know who’s missing their flu shot or who needs a mammography. All these things come together for an easy, seamless care experience.”
As is typically the case in healthcare, data gaps and inaccuracies are common, but the more data resources available the easier it is to overcome these challenges.
“We’ve found that some of our data sources contain address information that is fairly old – the change of address piece is not kept up in the national databases that are out there. We’ll have somebody we have seen in the last three months, and when you look at their particular address, you sometimes find that the person who lived at that address is a completely different individual,” said Parris.
“And while that does limit your ability to match people across platforms, you can work through it. When you have a lot of joint ventures, or other parts of your organization that are using different systems, it can be hard to connect downstream value to the upstream work. The various data sources allow us to make these connections. We’re able to tie the pieces together.”
The refinement of the tools and solutions developed by Texas Health Resources allowed providers to get ahead of the potential resource strains brought on by the COVID-19 pandemic.
“Before the pandemic, we spent a year and a half setting up a self-service analytics environment, where we allowed supply chain, finance, and HR to have access into the same tools as our business intelligence department. In the first week of March, when we began to realize that this is something we need to start tracking, we called up supply chain, HR, and finance for bed counts and other resources,” said Parris.
“We then brought in our clinical analytics team, and we were able to develop a unified dashboard in about three weeks with information across multiple departments based off their subject matter expertise. By the end of March, we were able to do some modeling based on what had happened in New York and other places, and we put predictive algorithms into our supply chain usage, bed counts, and event usage.”
Using predictive analytics tools, providers were able to see days ahead of time which facility had a high chance of experiencing patient surges. The organization could direct resources to the places that needed them, avoiding any potential staffing or equipment shortages.
“Those are some of the things that we were able to show with the predictive algorithms. And we as a system were able to take action and make the necessary changes so that the stress from COVID didn’t break our system,” said Parris.
For entities looking to implement these tools, Parris stated that it’s critical to ensure that end users understand how to best utilize these solutions.
“You can’t do everything in a centralized model. You have to work on standardizing toolsets and make certain that people know how to use those tools, and create a knowledge base around the data,” said Parris.
Parris also pointed out that comparing different datasets against each other is more effective than using just the data from a single organization.
“Referential matching is also much faster and easier than probabilistic matching between your own data. If you have another data set that you’re matching against, where patients have a name change or an address change, that’s where you start getting the power of it and the speed of being able to add a new source system,” Parris stated.
“We went through and added about 4 million records from a brand-new source system, and it took us about a week to match as many as we could. That is very fast for any system to be set up and turned around.”
Going forward, the health system expects to continue refining predictive tools and patient matching to be able to improve the care experience.
“We’re hoping to be able to know exactly who you are as you enter any one of our channels, whether that be digital or on site. If you’re on the web and you decide to call us, we’re able to pick that conversation up as well as apply some predictive models to be able to make your experience better,” Parris concluded.