Leveraging Digital Healthcare: Part I

Luba Greenwood , Seyla Azoz, Colleen Carroll, Lily Wolfenzon

Now more than ever, stakeholders across our global ecosystem understand the significant value of prediction and scalability that digital platforms and AI/ML bring. The first in a two-part series on digital health, read insights from Keystone’s digital healthcare practice and one of our experts Luba Greenwood.

At this time in our history, understanding and leveraging digital health tools is imperative. Be it to identify new diseases, adapting wearables to monitor a patient’s health, devices to streamline physician workflows, or data mining tools to make R&D more effective, digital platforms are disrupting the healthcare industry. Stakeholders across our global ecosystem understand the significant value of prediction and scalability that digital platforms and AI/ML bring.

The first in a two-part series on digital health, this article features insights from Keystone’s digital healthcare practice and one of our experts Luba Greenwood. Our practice leaders, Seyla Azoz (SA) and Colleen Carroll (CC), sat down with expert Luba (LG) – a healthcare and tech investor, ex-Google and Pharma exec, and now a Professor at Harvard – to hear her insights and perspectives.

SA: Given your experience at both Roche and Google, how do you think tech companies are approaching healthcare differently than more traditional healthcare companies? How is that manifesting in the products and services they offer? Especially as it relates to AI/ML and data.

LG: With Roche’s acquisition of Flatiron, it became evident for pharma players that utilizing data for discovery and clinical trials was key. Tech companies, however, saw the emergence of this trend years earlier, and as experts in data, they had anticipated the main hurdles pharma would face, namely the ability to gain access to and combine, store, cleanse, annotate and properly analyze structured and unstructured data. Pharma companies first rushed to Tech companies in the hopes of combining AI “magic” with R&D, only to discover that the “magic” was not in AI but in the data itself. They were not only sitting on a gold mine of data but also had access to additional data such as patient and genomic data through many of the partnerships they had already built with leading academic institutions and other partners.

The challenge for pharma companies remains in the ability to apply deep learning algorithms and other methods to their data sets correctly. Seeing an opportunity to play a bigger role, tech companies have stepped up AI/ML-driven offerings to pharma in two primary areas:

  • Research: Pharma companies have been hiring computational biologists, gathering, and analyzing genomic and other data for years. As tools for mining that data have been lacking, tech companies began providing these capabilities to pharma, in a form of services offered.  In the last two years, we have seen a proliferation of tech-enabled discovery companies that offer products that help pharma companies understand the biology of diseases, mapping out entire biological systems and discovering new targets.
  • Development (clinical trials): Pharma companies have been conducting clinical trials in an antiquated way for years, from trial design and cohort selection to data collection. Tech companies big and small have entirely transformed the way clinical trials are conducted.

Products that tech companies have introduced include sensors to collect real world data and accelerate trials. Tech companies quickly understood that virtualizing clinical trials would enable much faster enrollment of patients, better adherence, and faster tracking.

For example, Apple enrolled hundreds of thousands of patients in its Heart Study in less than a year, a previously unheard of number for a trial in that time frame. Tech companies have since introduced a number of tools and services in the space, including those to enable faster enrollment and better trial matching.

In both of these categories, many pharma companies have become savvy and recognize the importance of these tools, but they struggle with whether or not to build out those capabilities internally, partner with other big tech companies, or partner or buy smaller life science, tech-enabled companies. Although tech companies can provide these AI/ML tools and analytics expertise, the wrinkle is that providers (i.e. large academic hospitals) are truly the experts when it comes to properly annotating data and understanding how to match the right problem to the right deep learning model. As such, we now see tech companies partnering with providers directly to offer them ways to capitalize on their data and clinical expertise.

CC: Who brings the biggest value in deals between pharma/biotech or providers and tech, and has this affected how deals are structured?

LG: There are two key components of every deal between pharma/biotech and any tech players: (1) data use, ownership, and sharing, and (2) business model.

While the main assets of life science collaboration agreements had been patents, in deals with tech, they are now data and background IP. Partners often struggle with properly outlining data access and background IP terms and more importantly with agreeing on the use of insights gained from that data. Another challenge is the proper structuring of such agreements. While life sciences players are used to co-development, collaboration, and licensing agreements in deals where they are collaborating with a partner to discover a new therapy or build a new tool, tech players are used to Master Services and similar agreements where they get wide access to data and gain exclusive IP rights to newly formed products and services. When it comes to using patient data to build out different products, the privacy, safety, and IP concerns vary depending on what product you are collaborating and the ultimate business model.

The business model in pharma collaborations used  to be straightforward: you knew who provided what value because it was indicated in the license, you had a work plan, a joint committee with expertise in drug development, and once the drug or device was approved, you split up the royalties. However, in tech collaborations, partners often miscalculate the value that each side is providing. Moreover, the product or service resulting from the agreement follows different reimbursement and payment models, where often the value add that tech provides is minimally compensated. From what I have seen, this is a challenge many haven’t yet figured out and are doing incorrectly, in many cases paying out milestones that were incorrectly set given the overall value of the deal.

CC: What are the common pitfalls you are seeing with this challenge of valuing the offerings between healthcare providers and big tech?

LG: Life science companies and providers often overestimate the value of analytic tools, such as  AI and deep learning, which are now becoming a commodity. This is where providers, payers, and pharma companies could capture more value given expertise they provide. Ability to access, cleanse, store, and aggregate data, as well as to apply deep learning analysis to that data, all services tech provides, is valuable but the higher value comes from the ability to match a problem to the right kind of deep learning model, train the model, and label the data, all expertise that healthcare players provide. Moreover, clinical, medical workflow, uptake, scale, install base, and market expertise and knowledge, among many other factors, are highly valued and often proprietary information provided by the healthcare player and often more valuable than the analytic value provided by the tech company.

On the compliance side, negotiations between providers, tech and pharma had historically limited involvement of HIPAA experts and underestimated what could and could not be shared under HIPAA.

SA: What are some other areas in the intersection between tech and pharma that are seeing challenges?

LG: Another area that is emerging is mining data to inform discovery. There are a number of small tech startups that are trying to mine any data, including through analysis of thousands of cells and tissue images. By applying AI to that data, they hope to inform better discovery or determine why a particular discovery has failed. Many of these companies are partnering with biotech and pharma companies to assess trials that have failed using the power of the data or help with patient recruitment.

Pharma and biotech companies are becoming increasingly bombarded by these solutions, often distracting them from their internal and external R&D projects. Pharma and biotech companies often lack internal resources to pick out start-ups with the most promising approaches. More often than not, these start-ups lack resources and ability to collect the right type and amount of data to properly train their AI/ML models. Many often lack the right analytics to enable proper generation of insights and predictions that would be useful for drug discovery and development.

SA: The digital healthcare field is flooded with so many digital solutions right now. Should pharma build out tools internally or partner?

LG: Almost every pharma, device, and diagnostics company has a digital group. However, not all are successful. In my analysis of the most successful and effective groups, the key factors have been:

1. Focus – prioritization of projects and keen understanding of which functions to build out internally and which to partner-up externally and how.

2. Alignment – full alignment with R&D, BD, Marketing and Commercialization functions.

3. Talent – Formation of vision and culture that is separate from the rest of the organization. This approach has resulted in a better ability to attract and retain talent with expertise in data, analytics, and market access.

Success in innovation doesn’t come from adding a new tool to your tool box, but rather from the focused application of technology to your business, and the agility to adapt your business to a new business model.