Babylon Health’s Parsa founds new AI medical assistant venture, Quadrivia, one year after Babylon Health’s failure

November 13, 2024 | by

Ali Parsa back in the news, just over one year after Babylon Health’s implosion.  Babylon’s CEO/founder Ali Parsa has a new and stealthy AI-related venture called Quadrivia. It was announced, unusually, by a personal post on LinkedIn yesterday (12 Nov). As one might surmise, it’s AI-related (this year’s flavor) and provides an AI assistant to clinicians. The company is incorporated in Jersey (Channel Islands) and is UK-based. It is seed funded by Norrsken, a Swedish VC. The amount is undisclosed. (More on this below from Sifted-FT and JFSC registry research)

Information on Quadrivia’s capabilities is limited to Parsa’s posting, their website, and a ‘first-person AI assistant’ narrated demo video embedded in the post and the home page of the website. Qu, the personal medical assistant in beta, is designed to support clinicians in multiple tasks as diverse as type 2 diabetes check-in, daily care coordination, menopause hormone therapy, flu vaccination education, and chronic kidney disease (CKD) follow-up. Quadrivia promises that its “AI agent for healthcare” capabilities will extend across the entire healthcare ecosystem for the clinician and patient, including hypotheses around diagnosis, investigation, clinician selection, treatment plans, monitoring, and more. According to Quadrivia’s information in TechFundingNews, the platform has a dual structure based on types of cognitive reasoning processes: “System 1 includes tasks that rely on quick decision-making, such as answering direct questions or following standard procedures. System 2 involves more complex, analytical tasks, like assessing patient symptoms and considering multiple possible diagnoses.”

Quadrivia attributes Qu’s capabilities to its clinical knowledge base, the patient’s medical records in the EHR, natural language (but not real time) text/audio conversation, and proactive/reactive care. Beyond providers, Qu can be used by payers, pharma for research, and startups, much as Babylon Health was originally positioned.

Clinicians are invited to test drive Qu using a signup form featured on the website. Seven clinicians listed on the website constitute Quadrivia’s Global Clinical Advisory Council and are investigating various use cases. Per the website, the assistant will not be released until it meets regulatory and safety requirements. Though the website states that “Qu is tested rigorously to ensure clinical accuracy and safety in every action it takes,” it does not state whether eventually it will be submitted to the FDA, Health Canada, or for EU CE Marking.

In the US, there is growing concern about the rigor of FDA testing for AI assistants and interpretative/diagnostic models, with many falling into what is lately being seen as the looseness of 2016’s 21st Century Cures Act ‘breakthrough device’ category. 

Returning to Quadrivia financials, Sifted’s research indicates that the Jersey filing for Quadrivia Limited is for the parent company of a UK-based entity. Parsa is termed as a “person with significant control”. The VC Norrsken holds 638k shares in the company.

This Editor then took a dive into the Jersey Financial Services Commission (JFSC) registry for Quadrivia. Both listings filed in 2023 establish the company, the first for a “reserved name company” on 10 October 2023 (essentially empty) and then on 31 October 2023 for a “registered company private”. The incorporation documents are singularly unrevealing and procedural save for the Special Resolution document (PDF) of 24 May 2024 that determines that the company now has 9,440,000 A shares of no par value and the more revealing Entity Profile that confirms Norrsken holds 638,297 shares between NVC Fund 2 (D) plus (E) and ALP Partners Limited holds 9,444,000 shares. A check of ALP Partners Limited’s LEI (Legal Entity Identifier code) identifies its registration in St. Helier, Jersey. Its other entity name is Babylon Partners (Jersey) Limited.

Returning to Sifted’s article, Parsa was listed at HLTH as CEO of the company and was scheduled to speak before canceling. Based on Companies House (UK) filings, Parsa is a resident of Spain and holds more than 75% of the company, which corresponds to the 9.4 million share owned by ALP Partners and not owned by Norrsken. Two former Babylon research scientists, Damir Juric and Adam Baker, are early hires and the company is hiring others. Of note: Swedish investors have a long relationship with Ali Parsa. Major Babylon backers that owned–and lost–over 42% of the company included Kinnevik, VNV Global, pension fund AMF, and Swedbank Robur.

This Editor’s opinion–and to be clear, it is only my opinion: The speed of development–less than a year–and the scope of the AI assistant service does leave this Editor in a state of mystified wonder. For starters, accurate, responsive AI and vast stores of data are extremely difficult to build up to a clinically testable form in a year, much less clinical validity. The company is still at seed/stealth, yet inviting clinicians to test it.

Let’s look back at the Babylon Health parallel. It started to gain notice as early as 2014 at various UK conferences and won over GPs with its chatbot app. By 2017, GP At Hand debuted in parts of London allied to select GPs and was promoted in a flurry of tube adverts and billboards. It received huzzahs from none other than UK Health Secretary Matt Hancock in 2018. It also claimed a great deal in automated diagnosis of routine illnesses, not all of which were valid and problems surfaced fairly quickly. By 2020, with the pandemic barely on, Babylon’s accuracy was debunked at multiple stages by Dr. David Watkins, a consultant oncologist, better known as @DrMurphy11, cardiac activists on both sides of the Atlantic, and Hugh Harvey, Babylon’s former regulatory affairs head from 2016 to 2017 [recap and links TTA 14 Sept 2023]. Babylon moved into the US market, never gained FDA approval, IPO’d via a SPAC that later cracked, bought up a supermarket blood pressure tester (Higi), a few practices, worked with the government of Rwanda to bring care to rural residents, cut deals with major insurers…and then by August-September 2023, went thoroughly, completely belly up in US and UK. It left 2.4 million Rwandans in the lurch. GP At Hand survives with an NHS practice group in London and the rest acquired by eMed.

Based on the dates, it took no time–less than one month–for Ali Parsa to step out of the Babylon wreckage and create a new AI-based diagnostic and care assistant entity–and find another Swedish backer within that next year. One year to a new AI-powered concept. The resilience is…amazing. And leads to more than a few questions.

Robots, robots, everywhere…even when they’re NHS 111 online algorithms

January 23, 2018 | by | 1 Comment
[grow_thumb image=”https://telecareaware.com/wp-content/uploads/2014/01/Overrun-by-Robots1-183×108.jpg” thumb_width=”150″ /]The NHS continues to grope its way towards technology adoption, gets slammed–but is it justified? The Daily Telegraph (paywalled–see The Sun) revealed a draft December NHS report that recommended that the NHS 111 urgent non-emergency care line’s “enquiries will be handled by robots within two years.” Moreover, “The evaluation by NHS England says smartphones could become “the primary method of accessing health services,” with almost 16 million inquiries dealt with by algorithms, rather than over the telephone, by 2020.” (That is one-third of demand, with one-quarter by 2019.)

Let’s unpack these reported statements.

  • An algorithm is not a ‘robot’. This is a robot.[grow_thumb image=”https://telecareaware.com/wp-content/uploads/2016/06/robottoy-1.jpg” thumb_width=”100″ /]
  • What is so surprising about using algorithmically based questions for quick screening? Zipnosis in the US has been using this method for years as a pre-screener in major health systems. They call it an ‘online adaptive interview’ guiding the patient through branching logic of relevant questions; a provider can review the provided clinical note and make a diagnosis and treatment recommendation in 2 minutes. It also captures significant data before moving to an in-person or telemedicine visit if needed. Babylon Health uses a similar methodology in its chatbot-AI assisted service [TTA 26 Apr 17].
  • Smartphones as a primary means of accessing health services? How is this surprising when the Office of National Statistics says that 73 percent of adults use the internet from their mobiles? 51 percent go online for health information.
  • Based on the above, 66 percent would still be using telephonic 111 services.

It seems like when the NHS tries to move forward technologically, it’s criticized heavily, which is hardly an incentive. Over New Year’s, NHS 111 had a 20 percent unanswered call rate on its busiest day when the flu epidemic raged (Sun). Would an online 111 be more effective? Based on the four-location six-month test, for those under 35, absolutely. Yes, older people are far less likely to use it, as undoubtedly (but unreported) the disabled, sight-impaired, the internet-less, and those who don’t communicate in English well–but the NHS estimates that the majority of 111 users would still use the phone. This also assumes that the online site doesn’t crash with demand, and that the algorithms are constructed well.

Not that the present service has been long-term satisfactory. David Doherty at mHealth Insight/3G Doctor takes a 4G scalpel to its performance and offers up some alternatives, starting with scrapping 111.