CEO SUMMARY: In recent years, Labcorp invested significant sums to use artificial intelligence and machine learning technologies—often integrated with robotic systems—to improve work processes and gain real-time insights from vast amounts of data. In this exclusive interview, Lance Berberian, Labcorp’s Chief Information and Technology Officer, discussed several of the successes the lab company is having with these solutions, which involve lab operations, customer service, and more.
FOR LABCORP, ARTIFICIAL INTELLIGENCE (AI) AND MACHINE EARNING ARE NOT COMING—they are here—and being applied to support the company’s services to hospitals, physicians, and patients in myriad ways.
This fact should catch the attention of lab administrators and pathologists in hospital and health system labs who want to be fully competitive in all aspects of lab operations and delivery of clinical services. The current generation of AI and machine learning technologies are robust and can make a significant contribution in improving the performance of different aspects of a lab’s workflow and service mix.
Labcorp is one of the world’s largest clinical laboratory networks with 36 primary laboratories in the U.S. and $14 billion in revenue in 2020. The company has applied AI and machine learning across drug development and diagnostics including clinical laboratory workflow and operations.
“We use AI for a ton of different things. AI has had a tremendous impact on our operational capabilities,” said Lance Berberian, Executive Vice President and Chief Information and Technology Officer at Burlington, N.C.-based Labcorp in an exclusive interview with The Dark Report. “This includes right test at right location, maximizing our throughput, and minimizing turnaround time. Those are core foundational operational areas.
AI and Patient Care
“Improving the care of patients is getting a tremendous amount of effort that involves use of AI,” he continued. “AI is providing a better consumer and patient experience. Similarly, applying artificial intelligence to our clinical capabilities is another aspect.”
In its white paper, “How Artificial Intelligence Will Change the Clinical Lab,” Siemens Healthineers defined AI as “sophisticated software systems that enable computers to augment, or even emulate, human intelligence and decision making.”
Within AI, Siemens says, there is machine learning which “uses algorithms to parse and learn from data and then apply this learning to provide insight and make informed recommendations.”
Berberian noted, “When it comes to AI, the most important thing to have is a database that is true.”
Labcorp has developed its own machine learning models to manage supply distribution and identify employees needed across the company’s labs.
“If you don’t have supplies and labor, you don’t have turnaround time,” Berberian said. “If you think of AI as quality data upon which to build, you can model staff you need at a location.”
Furthermore, leveraging software, mechanical, and electrical engineering expertise, Labcorp has designed and built robots that take up 7,000 square feet. This work involved Protedyne Corporation of Windsor, Conn., a Labcorp subsidiary that is integrating robotic hardware and software infrastructure with data management and process tracking. One such system, the Protedyne Propel Plate Accelerator (PPA), is installed in Labcorp’s Burlington and Phoenix labs.
“We have designed our own robots to work in laboratories. Most labs do not do that,” Berberian said. “In eight hours, our robot system can process 750,000 test tubes with absolute precision.”
Vision-based AI enables the PPA to recognize loaded test tubes and appropriate positioning of them. While boosting lab efficiency and conserving costs, the robot also helps address the challenge of insufficient specimen quantity, which leads to the dreaded TNP (test not performed) designation, Berberian said.
PPA also records data on remaining amount of specimen. That helps Labcorp serve physicians who want to add a test after getting Labcorp’s report.
“In the old days, we didn’t know how much specimen remained. We didn’t count it or have it. No lab did,” Berberian explained. “But what the robot does—after the first order—is count the amount of liquid, so when the call from the doctor comes in we can say, ‘We have enough’ (to do the additional testing).”
Because of the automation of several work processes handled by this robotic system, lab workers previously tasked with tediously scanning barcodes and putting test tubes on racks, have been given more challenging assignments, according to Berberian.
AI for Better Performing Lab
Berberian next described how, in its services to hospitals and healthcare systems, Labcorp created a Performance Insights AI model to address a health system lab’s workflow hiccups. The challenges often include inconsistent test names and test codes, as well as routing specimens across a wide network.
“Because of mergers and acquisitions, in the lab division of a multihospital system, the same test isn’t the same test code at every site,” Berberian noted. “We have a model that overlays all the labs in a health system. It ties all testing sites together with the correct tests and provides guidance to optimize workflow within the health system laboratory.
“It (the model) gives them a dashboard that is backed by artificial intelligence and by algorithms that helps them understand the performance of their laboratory,” he added. “This is an overlay on top of LISs (laboratory information systems) and provides operational improvement for all lab sites. Our lab clients using Performance Insights can determine need—in advance—for reagents and other consumables at various locations.”
Enhancing PSC Experience
Labcorp has deployed AI and machine learning across the nearly 2,000 patient service centers (PSC) it operates throughout the United States, specifically to enhance consumers’ experience. For example, patients can use the Labcorp Pre-Check online process to make a test appointment and share insurance information.
Upon arrival at PSCs, consumers may place driver’s licenses and insurance cards on a tray without manually entering information. From there, a complex neural network developed by Labcorp works with cloud computing and optical character recognition to read card data. “Even though there are thousands of formats, we recognize the card type and the member ID,” stated Berberian.
Berberian heads a team of data scientists and bioinformatics professionals who develop Labcorp’s machine learning models for use in its laboratories and with customers. The company calls on vendors for some projects such as one with Ciox Health, an Alpharetta, Ga.-based health information management company, and another with PathAI, a Boston-based provider of AI research tools for pathology. (See sidebar below.)
Investment in AI
A deep dive into AI also requires a deep investment. The Siemens white paper says 69% of clinical laboratory leaders responding to an “Artificial Intelligence in the Diagnostic Lab” survey believe AI will be implemented in the lab by 2022. But 54% do not know where to begin.
“Your lab has to be willing to invest in these technologies,” Berberian advised. “People who report to me have PhDs.They are credentialed. This is not an inexpensive venture. Even with talented people, your lab team may not get to the end goal. We made the necessary investment, and it is hard to adapt these technologies on a small scale.
“I believe we are a better laboratory because of our applications of artificial intelligence and machine learning,” Berberian concluded. “AI supports our patients and our customers and I believe that it will be a pivotal technology going forward for our company.”
Lab administrators and pathologists may want to reassess the current state of artificial intelligence and how it might be valuable when used in their labs.
Labcorp Works with Ciox Health, PathAI
Labcorp and Ciox Health are collaborating to use artificial intelligence (AI) as part of a comprehensive patient data registry built from de-identified information on patients who were tested for COVID-19. It is aimed at helping researchers better understand COVID-19 diagnoses and treatments.
The data set is expected to offer up a “complete view of clinical paths and outcomes,” as it is supplemented with additional longitudinal medical record data, according to Labcorp and Ciox.
“Ciox retrieved medical records and returned them to us. We then had to turn that into structured data,” said Lance Berberian, Labcorp EVP and CIO, in an exclusive interview with The Dark Report. “We applied natural language processing that included an artifical intelligence model we use. That allowed us to extract data and load it into a dataset. Once we generate that information for the medical records, we merge it with Labcorp’s historical information for those patients.”
Labcorp is also collaborating with Boston-based PathAI to expand computational pathology applications in oncology. This involves the analysis of digital pathology images.
The project entails deployment of PathAI’s algorithms in prospective clinical trials of cancer and other diseases managed by Labcorp Drug Development, according to a PathAI statement.
The need for AI-powered pathology, according to Berberian, stems from two parallel trends: an aging population and fewer pathologists.
“The question becomes: How to optimize the time of pathologists so they can have higher output without sacrificing quality,” he added.
Contact Lance Berberian at 336-436-8263.