Healthcare is flatlining, but GPT-4 could revive it

We’re facing a rough outlook for the single biggest sector of the U.S. economy: healthcare. Hospitals make up 6% of the U.S. gross domestic product (GDP), but half saw negative margins in 2022. Three of the top U.S. health systems collectively lost $6 billion in operations in the past six months alone. 

Meanwhile, incumbent payers are also facing huge headwinds. Healthcare labor shortages and the limited availability of risk-bearing primary care providers (PCPs) that they can acquire are slowing payers’ ability to vertically integrate into “payviders”—a core driver of their recent growth.

But a new opportunity is emerging for providers and payers—they’re sitting on some of the largest and most valuable data sets in the world, and AI is finally starting to unlock their value.

SignalFire hosted a private event at HLTH to bring together a group of leading healthcare executives, our portfolio company founders, and industry advisors to examine how technologies like GPT-4 can make sense of the oceans of healthcare data. We were joined by Eric Larsen, president of the advisory board and venture partner at SignalFire. With 25 years of experience partnering with healthcare executives and spearheading UnitedHealth Group’s market strategy, Eric is a leading national healthcare strategist and advisor to CEOs and boards of directors for healthcare companies globally.

Here’s a look at the roadmap that Eric and the executives laid out in terms of what’s needed from healthcare startups to help incumbents reach financial stability.

Unprecedented challenges for healthcare providers and payers

Healthcare is a $4.3 trillion industry, making up more than 18% of the U.S. GDP, and it’s expected to grow to $7.1 trillion by 2030. This massive industry is dominated by incumbents: six thousand hospitals representing $1.4 trillion, 950,000 physicians, and self-insured employers.

But recent financial growth of this industry has depended on a strategy of vertical payer integration. Essentially, health systems acquire PCPs where profits are uncapped, and patients can be referred to their owners’ hospitals, pharmacies, or insurance—including elders signing with private insurers under the Medicare Advantage program. Profits can be shared between doctors and insurers if the PCPs take on the financial risk of providing value-based care. 

The 2010s was the decade of vertical payer integration. Those who owned the float on insurance premiums invested it in building ambulatory networks. This effectively created hospital-less integrated delivery networks (IDNs). Pioneered by Optum and Steve Hemsley, this model brings provider groups in-house and transitions their incentives from small business owners / employees into corporate employees. We’ve seen Amazon acquire One Medical for nearly $4 billion, CVS Health (which owns Aetna) acquired Oak Street Health for around $11 billion, and Walgreens Boots Alliance bought a majority stake in VillageMD for $5 billion.

But there are fewer independent PCPs and upstart clinic networks left to acquire, which could stall verticalization into “payviders.” Today, 80% of physicians are corporate employees. That’s up from 60% just five years ago. The aging U.S. population is also poised to spike demand for care, exacerbating a rampant healthcare labor shortage. Some underpaid, overworked PCPs and nurses, burned out from COVID, are leaving to start medspas, while grim conditions and a lack of schooling options are pushing replacements away from the industry.

The big story of payers bringing ambulatory networks in-house has largely played out, and the question is: can they retain them amidst the current labor shortage? The strength of noncompetes are being threatened at the federal level, and physicians in high-contribution margin specialties like musculoskeletal are aware of their outsized value to the finances of these networks. These folks in particular make for very attractive poaching targets for private equity firms and other roll-ups.

So what powers the next phase of growth for healthcare giants without jeopardizing care?

The rise of “GPT” and opportunities for generative AI

The upstarts in the healthcare ecosystem include 13,000 digital health companies that were spawned during the pandemic. Big Tech is expected to play a much more prominent role in an AI era despite many assuming this class of companies was mostly focused on low-probability moonshots.

This class of companies is being supercharged by technologies like OpenAI’s GPT-4. This fourth version of the generative pre-trained transformer initially just completed phrases with the next most likely word. But with fine-tuning, products built on large language models (LLMs) like GPT-4 can deduce billing codes from doctors’ notes (like CodaMetrix), price personal injury settlements based on medical records (like EvenUp), and build personal payback plans for patients (like PayZen). (Check out Nvidia’s health expert Renee Yao’s post on some of the leading generative AI startups that also joined us at HLTH this year.)

Interestingly, “GPT” also stands for general purpose technologies, which are advancements that affect entire economies at the national or global scale. Historians claim that there have been 24 general purpose technologies that altered human civilization (e.g., steam power, electricity), and GPT-4 may in fact be GPT 25.

One of the clearest paths to that depth of impact is in how AI can be used to mine critical information from the stunning amount of data the healthcare industry now generates. 

The explosion of untapped healthcare data in the zettabyte era

“Today approximately 30% of the world’s data volume is being generated by the healthcare industry” according to RBC Capital Markets. “By 2025, the CAGR of data for healthcare will reach 36%. That’s 6% faster than manufacturing, 10% faster than financial services, and 11% faster than media and entertainment.” Electronic medical record adoption, wearables, and new regulatory requirements are all driving this data proliferation.‎

This gives new AI technologies the fuel they need to improve all aspects of how we develop, deliver, and pay for healthcare. Tapping healthcare data for a wide range of purposes is bringing new light and heat to the industry among all the doom and gloom that incumbent providers and payers are facing. Given the incredible strides made in digitization and interoperability of healthcare data through the HITECH Act and 21st Century Cures Act this past decade, incumbent providers and payers are sitting at the center of this newly spun web of interconnected healthcare companies. AI could also help reduce the $570B in wasteful administrative spending in healthcare, augment understaffed healthcare workers, and meaningfully accelerate drug discovery and development.

But for any of these opportunities to come to fruition, all the data needs to be kept secure.

Data privacy to protect the future of healthcare 

The Cambrian explosion of LLM-powered applications in healthcare surfaces a critical question: how can owners of these powerful but sensitive, regulated data sets safely participate in the AI revolution?

Some of the core concerns include how to avoid data exfiltration and ensure patient anonymity. At the intersection of these two needs is a suite of privacy-enhancing technologies (PETs) that have existed for years or even decades but have yet to be productized as a platform for healthcare applications. Historically, the role of PETs has been confined to project-based deployments, or a mish-mash of point solutions for single-technique enablement. These include:

• Homomorphic encryption
• Differential privacy
• Secure multiparty compute
• Zero-knowledge proofs
• Synthetic data
• Hardware-based solutions like trusted execution environments

We believe one of the next generational companies, not just in healthcare but across many markets, will be one that successfully synthesizes and productizes the multitude of PETs we have available to us today in order to allow for a secure, private, dynamic data-sharing workflow that truly unlocks the potential of untapped healthcare data assets.

If you’re interested in joining this AI revolution in healthcare, please reach out to us and register here for our Gen AI in Pharma Showcase on Nov. 8 in Boston!‎