It was about five years ago when executive leadership at the Connecticut-based Stamford Health came to the realization that their reporting across the organization was quite poor and needed a major improvement. What was most perplexing, according to Divya Malhotra, M.D., director of analytics and innovation at Stamford Health, was that there wasn’t much insight into why the reporting was failing, just that data was not being received in a timely enough fashion.
As such, Stamford Health’s CEO created a business intelligence (BI) department in 2012, which Malhotra was tapped to lead. One of the first things she noticed was that even though the data infrastructure at the 305-bed hospital with 125 physicians was sound, meaning getting the data from all of the organization’s applications on a daily basis, there was no way to get those insights in a real-time fashion. “That was the biggest gap I identified as I stepped in,” Malhotra recalled.
Indeed, as a single hospital system without a $5 million budget to invest in a big BI and analytics infrastructure, Malhotra’s team piloted software from Seattle-based data visualization vendor Tableau to a few departments and key personnel—such as physician leaders, the chief quality officer, and the chief medical officer. “We focused on the biggest impact areas at first,” says Malhotra. “So we took out the operational reports that were coming out of the external vendor software and converted that in-house, and we slowly started to convert our back-end live database, which was literally just a replica of our Meditech application, into a structured data warehouse, which then hooked up to Tableau on a nightly basis,” she says.
One of the very first people to pilot the technology was Stamford Health’s orthopedic service line leader, who was previously getting all of the necessary data from an external software company, which would then turn it around, analyze it, and give the department graphs and charts. But that became repetitive with Tableau in-house, so all of that data then was converted into a Tableau scorecard. Since Stamford Health is running some 100 data sources in all, having a system that was able to connect to those sources and manipulate the data quickly became mission critical, notes Malhotra.
In fact, siloed reporting was one of the biggest pain points at Stamford Health for years, Malhotra says. “People were running reports out of applications. So just like in any healthcare system, there are multiple applications for different departments, such as financial, inpatient, and outpatient, and even siloed electronic health records (EHRs) that were running out of their own applications,” she says. “The result was that nothing was syncing up.”
An example of this, Malhotra explains, is having the finance department pulling up length-of-stay data one way, but clinicians pulling that data some other way from a different database which they submit information to. This would lead to the numbers from the two departments not matching up when they should be. “So we would get into this constant challenge of who’s right and who’s wrong,” says Malhotra. Is the [length of stay] definition right? What about the numbers are different?”
But with Tableau all of the data sources are connected, and there was also a desire to work on the definitions. “That’s a critical component. Data governance is critical for any kind of analytics program, as everyone needs to be talking the same language. If people are measuring things even slightly differently, it’s important to outline that and have that disclaimer anytime you publish a report, so everyone is on the same page, and we know when we’re comparing apples to apples and when we are not,” Malhotra says.
Another benefit, notes Malhotra, is that because they are now able to take data from different systems, what lacks in overall system interoperability is somewhat offset by the ability of these tools and technologies to connect data. For instance, she explains, Stamford Health’s eClinicalWorks outpatient EHR is not technically integrated with Meditech, its totally separate hospital-based EHR system. “We have used Tableau, data warehousing, and SQL queries, and we have connected patient level information and have actually matched patient identities,” says Malhotra. So If I am a patient in eClinicalWorks and end up getting admitted to the ED, we have reports running that would tell the primary care doctor that I ended up in the ED last night. That’s where the power lies in having a good analytics system that’s nimble and fast, and that’s where you can impact patient care and change the paradigm,” she says.
What’s more, in regards to financial, cost, and clinical variation that are certainly top of mind these days, Stamford Health has connected its OR module data with supply chain information. What this means is that for each surgeon and for each procedure, they can drill down to supplies used per case, and identify what the supplies are and what the pricing of that supply is, Malhotra says.
Taking a simple procedure like an appendectomy as an example, Stamford Health now has the ability to put all the surgeons who perform that procedure into one chart and look at variable costs. “We mapped our cost accounting data into the surgical data and then bumped it against our supply chain data. So I can say I am the cheapest but I might be taking three times as long [for surgery], for example. We can now get a comprehensive picture of clinical variation by surgeon, and our surgical chairs have shared that information with the surgeons,” says Malhotra.
Although many organizations struggle with proving the value of certain analytics efforts to end-user clinicians, that was not a problem at all at Stamford Health. “We were really hungry for the data, across the board, and challenges often come when you don’t like the data that’s being put in front of you,” Malhotra says. “You need data integrity and you need to be able to cleanse and normalize the data—especially for physicians, who I say are scientists and are driven by data,” she says.