Analytics reframes decisions from bench to C-suite

Anne Paxton

July 2021—From takeout margaritas to the embrace of remote work, the pandemic upended convention, leaving behind permanent changes that were nowhere on the radar in 2019. In the world of pathology informatics, the new online COVID-19 data dashboards at the Cleveland Clinic illustrate how much the pandemic has raised the profile of data analytics in managing the laboratory.

Supplying rapid summaries of COVID testing metrics like test orders, positivity rates, and turnaround times through live data from the Sunquest laboratory information system, updated every 30 minutes, the clinic’s dashboards have filled an urgent need. But through analytics and dazzling, interactive visualization tools mostly new to the laboratory, the dashboards have also taken on a new role, giving clinicians and Cleveland Clinic executives novel, compelling windows on laboratory data and insights into the laboratory’s value to the enterprise.

“We wanted to do something more rigorous and intentional than just bolt dashboards onto the laboratory information system,” says Walter Henricks, MD, laboratory director at the Cleveland Clinic and vice chair of its Pathology and Laboratory Medicine Institute. Interactive tools from visualization software—in the institute’s case, from Tableau Software, which focuses on business intelligence—helped make that happen.

The Pathology and Laboratory Medicine Institute had formed its data analytics program the year before the pandemic, and business analysts and information technology staff were on board, but when the COVID-19 pandemic hit, there was still not a fully functional, cohesive analytics team. “We were in the very beginning stages of putting the analytics team together,” says Ashleigh Muenzenmeyer, MSLA, SC(ASCP), the institute’s director of analytics.

Analytics is an area within informatics that supports laboratory operations, quality, and decision-making, Dr. Henricks notes. He draws a distinction between transactional information processing and analytical information processing. “A typical LIS and electronic health record are geared toward supporting processes and doing things right: accessioning specimens, interfacing with instruments, reporting results. Analytics, on the other hand, looks at the information to analyze it, draw conclusions, and represent it to provide a basis for decision-making.”

To perform the full spectrum of analytics work, the institute looked beyond the initial team, adding more laboratory professionals and analysts with expertise in visualization software who could quickly devise interactive dashboards. At Cleveland Clinic, it happened that the enterprise finance team had been using Tableau Software for a few years. “So we also tapped those people to assist us in designing and implementing” Tableau for the laboratory, Muenzenmeyer says.

This ramping up of the analytics team amid COVID-19 gave the program its chance to shine.

The institute conducts a process called ETL, which stands for Extract, Transform, Load. “So we extract the data out of the Sunquest LIS, then we load it into a data warehouse, and then we create different objects in that data warehouse that we ultimately use to connect up to Tableau and show these visualizations,” Muenzenmeyer explains.

Tableau defines a dashboard as “a collection of different views, allowing you to compare a variety of data simultaneously.” But that description doesn’t fully convey the sleekness and agility of the visualizations. A representative institute dashboard, showing test volume and positivity, demonstrates how a high-impact visual representation of regularly updated data relating to COVID testing can be produced by varying the prism of “flags” that filter the data (see dashboard).

In this example, two line graphs track the number of total COVID-19 tests and the positive tests by day. The viewer can interact with the page by clicking on “Caregiver Flag,” to make the graph show whether the test code is for Cleveland Clinic caregivers (yes or no); or on “Symptomatic Flag,” to show whether the patient reports symptoms; or on “PreOp Flag,” to show whether the tests were performed prior to surgery/procedure. There are several other options.

Another dashboard features rapid updates of data on COVID testing volumes, turnaround times, positivity rates, and more, with the viewer able to filter the figures by provider, patient type, rotation, ordering location, or ordering physician. Where were tests being ordered from across the health system? What areas were sending in COVID tests to the main campus? Those are sample questions the dashboards can answer.

“We came up with a suite of products that went beyond LIS reporting capabilities and out-of-the-box tools,” Dr. Henricks says. “One of the visualizations we have on our volume dashboards is a sort of comprehensive dashboard showing an overall health system level that allows you to put various filters on it to look at different populations. We also have dashboards that show our reference lab services broken down by service line, so you could look at the extended care population’s positivity rate and drill down to get even more granular on a client basis.”

The turnaround time dashboard gives instant feedback on variables that could potentially be affecting whether the laboratory is meeting targets it has set, Muenzenmeyer says. “We can start asking questions about where we have differences in the performance between different patient classifications—our inpatients versus outside clients.” In this dashboard, “we took not just the overall TAT but we broke it down into histograms for all parts of the process. So we have ‘collect to receive,’ ‘receive to result,’ and ‘collect to result.’”

“All of these different filters can be used in the histogram bins to help analyze if there’s a certain process here that they could drill into for improvements,” she says. Other visualizations show failure trending. “So you could see if failures are coming from a certain location on the hospital floor, the idea being that we can allow people to focus on areas to improve without having to sift through every data point.”