Veradigm, formerly known as Allscripts, sought to navigate a series of crises with a rebrand. Now, interim CEO Yin Ho is using the $140 million acquisition of AI startup Science IO to give it the only major competitive advantage over giants like Epic. We have a bold plan to utilize this data.
by katie jenningsForbes staff
WThe hat looks like a small wine refrigerator, but it's hidden discreetly in a corner of Veradime's 12th-floor office in Manhattan's Flatiron District. But instead of burgundy bottles, the server rack is visible through the sides of tinted glass. Surrounded by fans whirring at high speed are his Nvidia H100s, his three most powerful GPU chips on the market. These three chips, along with two more in a hallway closet, power ScienceIO's AI model. ScienceIO is a small startup that was the cornerstone of a Hail Mary strategy for a struggling electronic health records company, formerly known as Allscripts.
Veradime, 38, is listed on the Nasdaq after reaching a $145 million settlement with the Justice Department over taking kickbacks from opioid manufacturers, selling loss-making hospital businesses and failing to file financial statements over a year and a half It has been abolished. During February. The $820 million market cap company is still awaiting the results of an independent audit into a $20 million revenue misstatement, which it says was caused by a glitch in compliance software from an outside vendor. did.
Veradime's $140 million acquisition of five-year-old Science IO in March is central to interim CEO Ying Ho's answer to the company's woes. The idea is to use Science IO's generative AI models to breathe new life into Veradyme's treasure trove of medical data. “We acted quickly because we knew the day would come when generative AI would be all around us,” says Ho, 54. forbes. “And we lose control over how those tools are used.”
ScienceIO's model reduces the labor-intensive process of organizing and building hastily written, abbreviation-filled physician records from months to days. Veradime plans to apply these to 30 years' worth of data: his 200 million patient records and his 5 billion notes, mostly generated by independent clinics. Unlike other datasets provided by large academic medical centers located in major cities, Veradyme's data includes a significant portion of rural America, and Ho said it is “a large part of America's footprint.” “Mostly,” he said. Veradyme sells millions of anonymized patient health records to drug companies looking to understand how people respond to drugs outside of carefully calibrated clinical trials. It's planned.
ScienceIO was co-founded in 2019 by Thiel Fellow Will Manidis, a former researcher at cancer genomics company Foundation Medicine. The 25-year-old CEO had raised $25 million from Quiet Capital's Section 32 and Lachy Groom by touting his idea before the ChatGPT frenzy: The large-scale language model that powers the AI chatbot could track all clinical records. Records made by doctors and nurses that can also be used to understand.
ScienceIO's language models currently range from 1 billion to 13 billion parameters and aim to transform that chaotic information into usable structured data. Unlike OpenAI's models, which are trained on the entire open internet, ScienceIO's models were all trained on healthcare-specific datasets. “We quickly realized that the only thing that mattered to the quality of the model was the quality of the data,” Manidis said. forbes. “LLM is just overwhelmingly reflective of the data it was trained on.”
Combining ScienceIO's model with Verradigm's largest and largely untapped asset, medical record data rights, could be a powerful combination. Most medical records companies do not own the patient data flowing through their software. Their health system customers do. Anonymized patient data can often be used for research, but one major challenge is how to link patient records between different hospitals without using names or other identifying information. Electronic medical records giant Epic has admitted that it does not own any of its customers' medical record data. Cerner, which was acquired by Oracle for $28 billion, declined to comment.
Veradigm's setup is different because it owns the rights to the data, Manidis said. forbes. These rights allow the company to connect patients across different clinics over time. By linking these records, after that We anonymize information to protect people's privacy. “Our ability to link over 25 years of longitudinal medical data is unique,” he said. Veradyme does not currently sell personally identifiable data to drug companies, but may become involved in recruiting for clinical trials in the future, he said. forbes.
said Jeff Smith, Vice President and Life Sciences Industry Research Analyst at Gartner. forbes While there are “benefits” to Veradim's pivot attempt, he was quick to warn that “owning the data is only 60% of the challenge.” It also requires a combination of technology and healthcare expertise, which is rare and companies that can achieve this, he says, are “like discovering a leprechaun.”
This data is potentially extremely valuable because it can help pharmaceutical companies understand how patients respond to drugs over time in the real world. Recently, the FDA began allowing the use of real-world data from patients outside of clinical trials to speed up the drug approval process, which also has the potential to significantly reduce costs. The market for this type of data is expected to reach $7.5 billion this year, and is growing at a rate of more than 20%, according to investment advisory firm Beecher & Company. According to a presentation at the Barclays Healthcare Investor Conference, Ho hopes that ScienceIO's AI models will be able to structure, anonymize and package Verradigm's data for sale to pharmaceutical companies starting in 2025. The company expects its revenue to increase by 10% by 2026. The plan will initially target three areas: heart disease, metabolic disease and central nervous system disease.
Unlike many delisted companies, Veradime is highly profitable and has cash reserves. Unaudited 2024 guidance calls for sales of $620 million to $635 million and post-acquisition cash of $140 million. However, some analysts are concerned about the feasibility of Ho's plan. “While we are encouraged by management's positive outlook, some may be skeptical about the achievability of these growth targets given recent performance,” TD Cowen analysts wrote in a March note. writing. Veradime's stock price has been flat as of Wednesday's close since news of the deal, and Ho and Manidis have been on roadshows trying to convince bankers about the company's AI-powered future. . Mr Manidis said at the Barclays Health Conference in March that Veradime was “in a period of reinvention”.
“Our ability to link over 25 years of longitudinal medical data is unique.”
TThis isn't the first time Veradigm has tried to reinvent itself. For nearly 40 years, the company has been the Frankenstein of M&A activity. Founded in 1986 as Allscripts, the original business was selling prepackaged medications for prescriptions at doctors' offices. Ten years later, Allscripts expected him to lose $3 million on his $75 million in sales. That's when Glenn Talman, now a serial healthcare entrepreneur of Livongo fame (sold to Teladoc for $18.5 billion) and an investor at 7wire Ventures, took over.
Thalman sold profitable parts of Allscript's business, such as pharmacy benefits managers, and doubled down on software for e-prescriptions and related data. He believed there was great potential value in reselling patient data related to diagnosis, treatment, and outcomes. “Companies that can collect this data will have the 'holy grail' of medical information for resale.” [to insurers]” Talman recalled in a 2003 Harvard Business School case study.
In 1999, Allscripts went public, and Thalman continued to purchase and build out additional capabilities, growing it into the electronic medical records company we know today. But his last big deal, a $1.3 billion merger with hospital records company Eclipsys in 2010, created a rift on the board. “The board started to break down because we wanted to take EHR to the next level,” Talman said. forbes. “If you're a doctor and you see someone with a particular condition, your EHR should make you smarter. It shouldn't just be a record.”
Thalman was optimistic about the company's new direction. “Berdime's [current leadership] “We're smart enough to say, 'Let's pivot and use all this information,'” he said. “That's what I wanted to do 15 years ago.”
Manidis is excited to train ScienceIO's models on Veradigm's data. This is because this data includes patients from small physician practices in rural and low-income areas that are not typically captured. He gives the theoretical example of a pharmaceutical company producing a new weight loss and diabetes drug known as GLP-1. The drug company wants to learn how patients from different ethnic backgrounds in low-resource settings respond better to different drugs. “You can't buy that dataset. It just doesn't exist,” Manidis said. But Veradigm makes it happen right away.
The decision to acquire ScienceIO rather than use it as a vendor was a way to reduce risk, Ho said. forbes. “Because we brought it in-house, we don't allow it to learn from data that we don't control,” Ho said. In addition to Veradigm data, we also incorporate data from the disease registry collaboration with the American College of Cardiology and the American Diabetes Association. The acquisition consists of $96 million upfront and $44 million in installments over three years as an incentive to keep Manidis and ScienceIO co-founder Gaurav Kaushik at Veradime for at least that amount of time. was.
America's fragmented health care system undoubtedly requires more robust data. Primary care physicians may not use the same medical records software as hospitals and specialists. This is also an issue when using real-world data for FDA submissions. For example, if a pharmaceutical company wants to know if a patient has taken a particular drug, but the dataset doesn't include important parts of the patient's medical history, there's nothing they can do. “It's very easy to do it poorly,” says Shirley Wang, an associate professor at Harvard Medical School who studies the behavior. Without all the relevant information, the data is “completely unfit for purpose,” Wang says.
Privacy experts have expressed concerns about how AI will impact medical privacy. Once the data in these massive datasets is anonymized, it is no longer covered by the federal health privacy law HIPAA, said Deven McGraw, former deputy director for health information privacy at the federal Department of Health. Masu. AI models could relink sensitive health data with people's identities, at which point he said HIPAA would not protect people's privacy. Manidis said Veradime does not train its models on “re-identifiable data.”
For now, Manidis and Ho said they have begun pre-selling the data product to pharmaceutical customers, with pilots expected to begin by the end of the year. “We will start to see real revenue and margin expansion in 2025,” Ho said. forbes. What about 2024? “This is a year of investing and reinvesting in our company.”