East Coast vs. West Coast: a US perspective on digital health

“…my vision be the clearest” - The Notorious BIG (not a candidate for Luxturna)

I recently attended the Massachusetts Biotechnology Council’s inaugural Digital Health Impact event. Massbio has played a huge role in driving the Life Sciences industry in Boston, and they are now launching a Digital Health track for their successful mentoring programme, MassCONNECT. They covered a range of themes in this space, some of which are summarised below.

Theme One: Digital health can be built from a science-first or tech-first perspective

• There is a difference in the East Coast (science-driven) vs. West Coast (tech-driven) approach to building digital health companies. The assets required to build digital health companies are present in each location to varying degrees (scientists / researchers, providers / hospitals, health data, investors, mentors — more information in this report) however the West Coast approach of ‘failing fast’ was seen as a key challenge (see Theranos), as was the idea of ‘pivoting’ (e.g. from wellbeing to health) given the need for different evidence, advisors and investors
• Some of the big opportunities for the overlap between digital health and life sciences (Luxturna — gene therapy for inherited blindness — is a good example) are 1) Early diagnostics (e.g. diagnosing breast cancer at the early stages, before it has spread); 2) Building drugs around predictive biomarkers (e.g. targeted therapies for HER2 positive cancers); 3) Developing companion diagnostics (e.g. tracking if individuals are HER2 positive before treatment)
• Despite the Dunning Kruger effect for tech companies new to the health space and cultural differences to some other startups, they also bring skills relevant to digital health e.g. drug development based on a cell biology model of disease and could become increasingly relevant in the future. There is significant activity in the big tech players (e.g. Amazon) and the rise of tech’s Chief Medical Officers could additional scientific rigour

Theme Two: Sustainable business models are in digital health can be complex to develop, and often receive less attention than they should

• The key customer groups for digital health are 1) pharma, 2) payers (including self-funded employers), 3) providers and 4) consumers. Targeting two groups is sufficient to diversify your risk, any further will split focus at an early stage. To determine which to start with, consider who is willing to take a clinical and financial risk with their population, how quickly they can move (in order to drive traction and start getting data), and how funding flows through the organisation
• Having evidence of efficacy / cost saving estimates is not enough: you need to prove the value to the payer over time as many solutions end up costing more than they save in real life. RCTs are highly controlled environments (follow-up on missing data, paid patients) which is highly unrealistic for digital health, especially as they often change the patient pathway
• Fundraising / working with VCs can be challenging if they don’t understand the space, for example, Product-market fit can take much longer due to the nature of understanding complex problems. E.g. taking the time to understand mental health comorbidities when designing a product in this space, especially given rates of under-diagnosis. Investors who push for growth too early can burn startups. Additionally, typical VC engagement metric targets may hamper the effectiveness of some digital health interventions, which sit in the background until they are needed (e.g. Lumme Labs uses gesture recognition to deliver just-in-time and preventative cognitive behavioural therapy for smokers)
• Based on the above, digital health solutions tend to be focused on areas with obvious problems & viable revenue models, many of which link to care coordination (e.g. the 4,000 patients that arrive every day to a dermatology clinic in China). These democratisation opportunities can be contrasted with pharma focus on rare diseases (easier to understand conditions, selecting for high willingness to pay in order to cover trial expenses with a limited population) which have driven up average drug prices

Theme Three: Raw health data is unlikely to hold much value, however the regulatory activity around high quality health data / insights will determine the business models of the future

• Key questions around the value from health data are: 1) who owns the data? 2) who owns the insights from the data? 3) how are AI-driven insights regulated?
• Interoperability can take place at a number of levels — there are examples of each, and the status quo may vary by market / use case: Option 1) The patient owns all their health data and acts as a conduit / performs some care coordination. This is the model used by Apple and also in India (albeit patients have paper records they carry around). Option 2) Movement of data between systems (e.g. EMRs). There are a number of industry coalitions on this topic, and progress is being made. Option 3) Longitudinal integration of care records in the cloud. This tends to be use-case driven as it is effort / capital intensive (e.g. Surescripts for prescribing data). Whilst less common, this method is valuable for creating provider and payer insights and improvements — Australia tried to collate patient records in PDF form and had to abandon this effort as it wouldn’t allow for analytics / machine learning
• Panellists thought that a data donation model was more likely to prevail (though there are companies paying for data) given the significant effort required in data aggregation, annotation (e.g. Flatiron annotated records with clinically-meaningful information). Equally, freely available de-identified data sets are valuable in research — 60% of machine learning papers reference the MIMIC dataset). They saw more value in the cloud / applications using the data than in the records themselves