After setting out to examine digital healthcare from the inside by launching its own women’s health clinic as an app last year, French startup Nabla is executing the next step in a planned pivot to b2b – announcing today that it’s opened its machine learning tech stack to other digital health businesses and healthcare providers so they can offer what it bills as “personalized medicine”.
Nabla’s AI-powered patient communications and engagement / retention platform is designed to support clinicians to deliver a more continuous, data-driven service, whether the client is offering real-time telehealth consultations or delivering a service to patients via asynchronous, text-based messaging .
Nabla’s messaging and teleconsultation communication modules sit as a layer atop the customer healthcare service, ingesting and structuring patient data – with its machine learning software supporting clinicians with real-time prompts and visualizations, as well as offering ongoing patient outreach features to extend service provision.
The startup argues its approach can improve medical outcomes by supporting healthcare professionals to be able to ask relevant questions during a consultation, based on the AI’s ability to aggregate patient activity and surface contextually relevant data – and afterwards, with features like automated transcription and by suggesting updates a clinician could make to a patient’s medical file.
It likens the platform’s capabilities to having a really attentive family doctor who knows their patient’s full medical history and situation – and has a fault-less memory for all that detail. But the tech can go beyond what even a great doctor can offer as it enables healthcare providers to supplement in person consultations with ongoing, asynchronous outreach to provide a layer of continuous care – such as via follow on scheduled messaging (eg to offer treatment reminders or ask patients about their progress etc). And, of course, even the best human doctor isn’t going to be able to provide patients with that level of check-in and attention in between visits.
Nabla’s premise is therefore that blending digitally delivered, synchronous (human) care with data-driven (AI-powered) support and asynchronous follow ups can offer a win-win: For patients, who get more ongoing (and potentially holistic) care than they could expect from traditional healthcare service delivery; and for digital health businesses which get to drive customer engagement and retention thanks to the smart, personalized assistance and outreach enabled by its platform.
Customer retention has become a pressing problem for digital healthcare providers, Nabla argues – pointing out that after the flood of interest in the space during the pandemic many of these businesses are likely coming back down to Earth with a bump as patient attention disperses, and as the wider global downturn complicates the task of scaling by raising funding.
“Health tech of course is affected a lot by the economic downturn,” says co-founder and CEO Alexandre Lebrun. “Around us we see lots of health tech startups that… owing to the COVID-19 crisis they automatically had lots of patients… It was very easy for them to get lots of patients and engagement. And now that COVID-19 is over – and plus cash is not free anymore – they discovered they have the same problems as ecommerce companies – I have to take care of my customers, I have to work on retention, I have to make them happy . It’s not just automatic. ”
Nabla is not (currently) in the business of automating healthcare; rather its platform offers real-time clinician support and clinician-approved outreach to patients – which means that, crucially, a qualified human doctor remains in the loop and in charge of patient care decision-making at all times. So its product is not itself a medical device – although Lebrun can envisage taking further steps in that direction down the line.
“Our long term goal is to use this data not just for the benefit of one patient but learn and aggregate all this data and, for instance, try to predict what will happen next with the patient or to do faster diagnostics,” he tells TechCrunch . “Of course the data we have is super valuable for research because we have very, very detailed information about the patient and not just the typical hospital records.” [We have data on] what they eat, how they live, their social environment, family environment – we know it’s very important for health but this information is nowhere to be found in existing medical records. But we have part of it. And so this is incredibly valuable for future academic research – and when we ask our users would you agree to share this data for medical research… most say yes of course, if they understand the scope of what we share. ”
Lebrun cut his teeth in tech working on chatbots – and clearly has a strong appreciation of the limitations of the technology. After selling a prior AI startup (Wit.ai) to Facebook he stayed at the tech giant to work on developing its hybrid general purpose AI concierge service (aka “M“) – which Facebook ultimately decided did not scale for its user base. But Lebrun had seen the potential of combining human-plus-AI for decision-making support, and decided to return to startup land to apply a similar hybrid approach in the narrower domain of healthcare where utility looked easier to hone.
Setting up and running a women’s health clinic was how Nabla’s founders subsequently decided to get to know the needs of the industry they wanted to supply and support with machine learning software – launching their clinic as an app in April 2021. The app, which Nabla says it will continue operating for the moment (although it’s no longer their main focus or product), has amassed some 25,000 patients to date.
This approach means Nabla’s tech is in the relatively novel position – certainly compared to general health products historically – of having been informed during development, primarily, by women’s experience. And its co-founders argue that resulted in a product which is both more attentive vs alternatives and more useful as a healthcare tool regardless of the sex of the user. So another win-win, as they tell it.
Having direct access to patients and doctors through the clinic provided Nabla with a link to core users, data and expertise it needed to develop the machine learning health stack product it’s now seeking to monetize. Although it emphasizes that patient data confidentiality requirements has meant always working with strict limits on data access – such as its engineers not being able to directly access users’ medical information (including during the development of the AI-powered tech stack).
“I think the consequence of choosing women’s health is that we focused a lot on empathetic care. On the continuous and pluri-disciplinary aspect of care – and that is completely forgotten in the existing healthcare systems, ”suggests Lebrun. “It was a hard decision to make for me to say okay. I’m opening a women’s health clinic. I started to spend all my day learning lots with gynecologists… If my co-founder was not a woman I wouldn’t have had the confidence to start a women’s health clinic. So it was great. We didn’t plan it when we started Nabla together – but it was good, ”he adds.
“It enabled us to focus a lot on these things. Empathy, pluri-disciplinarity on the provider side, and trying to have a whole person view of the patient is more important for women than for men. I’m a man, I have a problem with my arm, I go to the doctor, ten minutes later I know what to do. That’s solved – but this is not what women need. And this is not what the existing system provides. So we learned quickly to provide this kind of care with a mix of asynchronous and synchronous care. And what’s interesting is we realized that now, today, that this kind of care is actually better for everyone. Even for men. ”
“What we really want to build is a patient engagement stack,” adds Delphine Groll, who is co-founder and COO. “And when we did some research we did a lot of beta version of the app and we found that women were the more engaged population regarding remote care. And as our focus was to drive engagement thanks to our ML models it was – I think – the best choice to have this kind of population so we could be in a position where we could understand a lot the insights from them and then put the best stack we could regarding engagement, retention… which is the main challenge healthcare companies have at the moment. So I think it was not the only reason – but one of the reasons we choose also to focus on women’s health. ”
Nabla’s communication modules, which are connected to its machine learning-powered physician console, have been available to third parties, via APIs and SDKs, for about the past three months – and it says it’s signed up around 10 customers so far – but it’s announcing the formal opening today.
Early customers – which span a range of markets including the US, the UK and France in Europe, and Africa – include digital health startups such as Resilience, Cardiologs, Aura Fertility, Omena, Umana, Jeen, and Tchak; and established healthcare organizations such as AP-HP, which it notes is the largest hospital group in Europe. The idea is for the b2b business to be international from the get-go, per Groll.
Discussing the competitive landscape, Lebrun names as its closest rivals the US firms Canvas Medical, Seqster and Zus Health – and he confirms Nabla has strong designs on the US market, given how much digital healthcare action it accounts for.
“The closest companies are all in the US where, I think, they understood quite quickly that there will be a new tech stack of healthcare – like what happened in ecommerce 20 years ago – where every piece of ecommerce is managed by one of a few companies and then you assemble these bricks. The same thing is starting to happen in healthcare, ”he argues, also likening Nabla’s approach to clinician support as akin to GitHub’s ‘AI pair programmer’ software, Copilot.
“There is no question we’ll compete soon. Of course the needs are slightly different in the US. These competitors – I think we have the same philosophy of enabling healthcare providers to build the experience they want and make the life of the healthcare providers, of the doctors easier. But I think we have more focus on asynchronous care – how building asynchronous with synchronous care is interesting.
“Machine learning can of course help do that and with our 20 years and three companies before in the machine learning [space] I think we have great real world experience… We know what’s possible, what’s not possible and what’s desirable or not and we are trying to apply this. ”
Groll suggests Nabla’s edge vs rivals is its focus on not just extracting what patients are telling their doctors but structuring that information so that it can be put to work in wider service of improving healthcare provision for them by surfacing suggestions for personalized follow-ups – and also, potentially, for research purposes, if patients agree. (Patient consent for their data to be processed is required for use of the AI-powered service; and, separately, for any wider sharing of aggregated and anonymized data for research purposes, the startup confirms.)
“When we are talking to potential clients what they really like at the end of the day is we are extracting the data from all the communication – especially messaging and teleconsultation – and we are not only extracting it… we are also structuring it and normalizing it so it can make a strong asset for them, ”she says, adding:“ So I believe our differentiation is… we are enabling them to have a communication module but also to have a way to leverage the data they have inside those communications. ”
On the research side, Lebrun suggests the approach – as / if Nabla scales usage – will likely entail collaborations between its healthcare provider partners and public research institutions that would carry out studies in specific areas of interest, relying on aggregated, anonymized data from the service. provider / s.
“Research is better conducted by public institutions or academia,” he argues. “So I think it would be a three parties collaboration where Nabla’s providers agree to contribute their data. They of course they ask the consent from their patients to agree to share their data. And then the academic partner or public institution does the actual research but you could see Nabla as a network of healthcare providers who can have an easy way – because it’s already structured – to contribute data for research. ”