Newsbytes

How a project team streamlined an immunology lab workflow

April 2022—Michelle Stoffel, MD, PhD, supports the use of Excel spreadsheets in some areas of laboratory medicine, but not necessarily as a laboratory workflow tool. It’s a realization she came to when, as a clinical informatics fellow at the University of Washington School of Medicine, she led the charge to revamp the workflow for the immunology laboratory’s Merkel cell antibody panel.

“I think in any lab you’ll find at least one of these very work-intensive, low-technology processes going on for which a full-automation solution might be unattainable in the near future,” she says.

Fully automating UW Medicine’s highly specialized, multi-step Merkel cell antibody panel, which tests for antibodies to the Merkel cell carcinoma oncoprotein, would have been a lengthy project and require extensive information technology resources, Dr. Stoffel explains. Therefore, she and her colleagues took a different approach, targeting smaller, semi-manual Excel spreadsheet-based processes within the workflow that were cumbersome to perform. “Cumulative inefficiencies within your workflow can really add up,” she says.

In late 2019, immunology laboratory decision-makers, medical laboratory scientists, and members of the pathology and lab medicine informatics team met to kick off the workflow project, says Dr. Stoffel, who gave a presentation on the project at the 2021 Pathology Informatics Summit. The core informatics team that worked on the project included Patrick Mathias, MD, PhD, associate medical director, informatics division, Department of Laboratory Medicine and Pathology, who helped guide technology decisions; data scientist Nathan Breit, who handled the programming for the project; and Dr. Stoffel, who spearheaded the project and was the primary liaison with the immunology laboratory, where the tests are performed.

The bulk of the time spent on the project was dedicated to the analysis phase, during which Dr. Stoffel made site visits to the immunology lab to map the assay workflow by sitting with medical laboratory scientists and observing each step they performed. Process mapping, says Dr. Mathias, was critical to helping the informatics team understand the pain points the lab was experiencing.

“It’s really important to have informaticists play that role of translator between the two sides,” he explains. “Being able to understand what the lab needs and help develop the requirements for the solution and then go back and forth [between the informatics and lab teams] is a very valuable experience.”

Ironically, when Dr. Stoffel shifted to observing the panel’s workflow through Zoom meetings during the pandemic, the process mapping became more efficient. During the Zoom calls, she recorded the dozens of steps the medical laboratory scientists performed. Afterward she played back the recordings slowly to make sure she captured all of the information.

“I could ask questions in a more targeted way because I had a little time to think about it,” Dr. Stoffel says, adding that the in-person visits were still valuable for getting to know medical laboratory scientists and assessing unusual cases.

The informatics team identified two processes within the Merkel cell antibody panel workflow that could benefit from automation—patient lookup and curve fitting. The patient look-up process was an Excel spreadsheet-based system the lab had created to search for patients’ past tests. Excel spreadsheets are widely used in clinical labs because they are easy to create and manipulate, Dr. Stoffel says, but these same characteristics make it easy to inadvertently alter the data, compromising the integrity of the document. Excel spreadsheets can also have reporting and formatting limitations, she adds.

The immunology lab had created the Excel spreadsheet-based patient lookup because the pathology department’s lab information system uses a sample-centric rather than a patient-centric identification system, Dr. Mathias says. This is because the pathology lab performs reference testing on hundreds of thousands of samples submitted from outside the health system. Tracking specimens using a specimen-based code rather than a patient code saves the time of creating a patient record, he notes.

The immunology lab can process samples more efficiently if it doesn’t have to consider previous results, Dr. Mathias explains. “You can bring in a sample and test it without taking extra steps to register the patient.” But the immunology lab needed to track past samples to give ordering providers not just the current assay results but a historical perspective on which direction the patient’s results were trending, he says.

Laboratorians had been cutting and pasting testing data from a weekly LIS report into an Excel spreadsheet and using the spreadsheet as a database of test samples. The immunology lab also had created a type of unique patient identifier for every Merkel cell antibody panel test sample and was using that identifier to search the Excel file for records of patients’ past tests. The process was time-consuming and required extra steps to minimize the risk of cutting and pasting errors, Dr. Stoffel says.

What the lab needed instead was an application for searching LIS data in a customized way. And it turned out the laboratory informatics department had already built a tool with this functionality for another internal specialty lab, which the immunology lab could deploy with minor modifications. The existing search application—developed in the R programming language, with a graphical user interface written in R Markdown—is one of many applications the lab informatics department had built to obtain information from the lab’s data warehouse, Dr. Mathias explains.

Because the existing application was so similar to what the immunology department needed, it was easy to deploy. Once the programmer understood the immunology lab’s data requirements, it took only about 10 to 15 minutes to set it up, Dr. Mathias says.

The search application operates like a dashboard, Dr. Stoffel adds. It refreshes daily with information from the data warehouse, eliminating the need for the immunology lab to copy and paste data from weekly LIS reports into Excel spreadsheets.

The informatics team demonstrated the tool to the lab via Zoom, but Dr. Stoffel found that an in-person site visit was more effective for facilitating the rollout. “I had followed up and saw they hadn’t gotten a chance to use it, so that’s when I used a hybrid model of following up the Zoom call with a lab visit,” she says.

The second process the informatics team targeted for automation—curve fitting—also involved manually cutting and pasting data into Excel spreadsheets and invited similar opportunities for pasting errors, Dr. Stoffel says. The immunology lab had been taking output data from lab instruments and entering it into Excel spreadsheets to format it for use with commercial graphing software.

The informatics team found a readily available solution for automating this process as well, but this time the solution was online. “There were open-source widely available Python libraries that you could use for that curve-fitting function,” Dr. Mathias says.

The tool that the team constructed looks like a Web page to the user, but it is a Web application that formats the data output from the analyzers and delivers it to the graphing program. Users need only drag and drop a data file into a field on their screen to obtain curve fit graph results, he explains.