4 AI-Enabled Healthcare Projects
January 22, 2024
AI innovation in specialty healthcare is ramping up every day, in Canada and elsewhere. The four projects described here give a taste of what’s happening and what’s on the horizon.
#1: IQVIA uses AI to improve clinical trial recruitment
Global investment in AI for clinical trials is growing, with pharmaceutical companies taking the lead. AstraZeneca has announced a new health tech unit called Evinova, dedicated to bringing AI to clinical trials,28 while Roche’s Genentech unit has partnered with computing giant Nvidia to fold AI into the drug discovery and development process that culminates in human trials.29
Clinical trial recruitment poses a particular challenge for diseases that affect a small number of affected patients. To recruit enough subjects in a reasonable timeframe, rare disease trials need to include patients with undiagnosed disease. IQVIA, a world leader in the use of data to drive healthcare, has just made the process easier with an AI algorithm that combs through medical databases and identifies diagnostic patterns suggestive of an undiagnosed rare disease.30 If genetic evaluation confirms the disease, interested patients can move onto a clinical trial.
As an example, an estimated 150,000 people in the US have or are at risk of spinocerebellar ataxia (SCA), a disease characterized by progressive loss of motor control. IQVIA used its AI algorithm to analyze the characteristics of SCA patients and, on that basis, create a shortlist of SCA predictors. In essence, they used AI as a screening tool, simplifying and accelerating the search for SCA patients suitable for clinical trials by finding potential undiagnosed patients.
#2: Pentavere’s AI engine identifies diseases from physicians’ clinical notes
Real-world data (RWD) can identify disease patterns and treatment outcomes that clinical trials lack the power to uncover. One problem: the information needed to understand and treat a patient often lies buried in a large, unstructured dataset within electronic health record systems.
Take plaque psoriasis, for example – a common skin problem that can have a profound impact on quality of life. To meet patients’ needs, clinicians need fast and reliable ways to identify disease that may require more intensive treatment. Pentavere, a Canadian-based healthcare AI company, has given the job to their AI engine called DARWEN™, which can quickly and economically extract RWE “trapped” in unstructured data repositories and transform it into reports that clinicians can easily interpret to improve patient care.31
In a feasibility study, DARWEN analyzed the electronic medical records (unstructured data) of over 10,000 patients to identify those with plaque psoriasis and describe the disease characteristics in each case. Of the 663 psoriasis patients flagged through this process, 259 had features indicating severe disease, while 135 fell into the moderate category.31 According to the study investigators, such a “seamless and comprehensive assessment approach could allow the dermatologist to self-audit their management of patients with moderate to severe [psoriasis] and determine if adjustments are required to reach treatment goals.”
#3: Auxita ‘s machine learning tool digitizes patient enrolment
Automating administrative tasks may not be AI’s sexiest application, but – when you consider that AI could free up 1.8 billion healthcare hours per year, which healthcare providers can devote to their patients – it ranks among the most important.13
The administrative burden is especially heavy for the patient support programs (PSPs) that help patients navigate treatment with specialty medications. Indeed, some clinics have to hire extra staff just to manage the paperwork – hardly the best use of healthcare resources, when the time could be spent with patients. Mindful of this, Ontario-based software company Auxita, as part of their agnostic digital PSP platform, has leveraged their AI tool to digitize PSP enrolment forms, with a focus on improving efficiency. The system can interface with fax machines – an antiquated technology that the healthcare sector has shown a curious reluctance to give up – recognizing the form type and converting handwritten and typed faxes into data that populates a digital enrollment webservice fully available to the PSP’s CRM.32
Illegible handwriting? Not to worry: the AI tool feeds it into a large handwriting database to makes sense of the scrawl, which it then converts to structured data. Not only does Auxita’s system leave PSP staff with more hours to allot to patient care, but it requires no procedural change from physicians: they can continue to scribble notes and send faxes, and the system takes care of the rest. It’s no surprise that Auxita cites “care over data entry” as one of its foundational pillars.33
#4: AI and synthetic control arms for patient populations in need
Synthetic control arms work like this: instead of recruiting a control group for a clinical trial, investigators can model a control arm by collecting real-world data from existing sources, such as electronic health records and disease registries. In brief, the control group lives in the data.
Along with saving a lot of time and money, synthetic control arms make it possible to run comparison trials when ethical concerns (for example, offering a placebo instead of a highly promising treatment) preclude recruitment of an actual control group.34 In a recent Canadian example, a trial with a synthetic control arm created from real-world Alberta databases determined that the novel drug lurbinectedin offers better survival odds to patients with small cell lung cancer than the historical standard of care.35
This is where AI comes into its own: sifting through vast datasets, using analytic techniques such as machine leaning and natural language processing to extract relevant information. While the use of AI to create synthetic control arms is still in its early days, you can expect to see a lot of more of it in the years to come.
References
13. The socioeconomic impact of AI in healthcare. MedTech Europe. October 2020. https://www.medtecheurope.org/wp-content/uploads/2020/10/mte-ai_impact-in-healthcare_oct2020_report.pdf
28. Robson K. AstraZeneca builds health tech unit for AI based clinical trials. Verdict. Nov. 20, 2023. https://www.verdict.co.uk/astrazeneca-builds-health-tech-unit-for-ai-based-clinical-trials
29. Wu G. Roche’s Genentech partners with Nvidia in AI drug deal. BioPharmaDive. Nov. 21, 2023. https://www.biopharmadive.com/news/genentech-nvidia-biotech-ai-drug-research/700221/
30. Recruiting rare disease patients just got easier. IQVIA. Nov. 7, 2022. https://www.iqvia.com/library/white-papers/recruiting-rare-disease-patients-just-got-easier
31. Vender R, Lynde C. AI-powered patient identification to optimize care. CPD Network. https://pentavere.ai/wp-content/uploads/2023/03/CDA-Poster_AI-Powered-Patient-Identification-to-Optimize-Care.pdf
32. 20Sense original research.
33. LinkedIn. Auxita. https://www.linkedin.com/posts/auxita_genai-healthtech-healthtechevolution-activity-7138558539337252864-Ooyk
34. Goldsack J. Synthetic control arms can save time and money in clinical trials. STAT. Feb. 5, 2019. https://www.statnews.com/2019/02/05/synthetic-control-arms-clinical-trials/
35. Boyne DJ et al. Comparative effectiveness of lurbinectedin for the treatment of relapsed small cell lung cancer in the post-platinum setting: a real-world Canadian synthetic control arm analysis. Targ Oncol 2023;18:697.