The world is amidst an AI-driven industrial revolution. As a16z cofounder Marc Andreessen recently wrote, AI will save the world. The United Kingdom just staked its claim on the AI frontier, challenging US policymakers to reaffirm America’s role as the powerhouse of tech innovation. Whether we let it improve the health of Americans is up to how we regulate it.
There is no sector where AI can drive more immediate, life-saving impact than in biotechnology and healthcare. In bio, AI will enable our scientists to conduct experiments faster and more efficiently than ever before, resulting in better treatments for patients suffering from debilitating diseases. In health, AI will enable our care providers to tend to more patients with greater precision and less burnout, resulting in better consumer experiences and health outcomes. In our communities, AI will enable people to live healthier, happier lives with their families and loved ones. Patients are poised to reap immense benefits from AI integration into the biotech and healthcare sectors. Broad implementation will democratize access to high quality care.
As policymakers grapple with the integration of AI into life sciences and care delivery, it is essential that they work closely with industry to consider how potential regulations will enable a vibrant and competitive marketplace and maximize patient welfare. At a16z, we’re fortunate to work with many of the most innovative, forward-leaning companies shaping the future of AI, affording us a unique vantage point to anticipate the landscapes of tomorrow.
Like many areas where AI technology is being used, regulatory frameworks already exist, and healthcare is no different. The federal government has been regulating software-based medical products since the 1970s. As policymakers face calls for regulation of artificial intelligence in the biotech and healthcare sectors, they must consider these important themes to prevent stifling innovation and incentivize investment at home.
A tidal wave of investigational new drug (IND) applications is headed for the FDA. AI advancements in drug discovery and development are driving diminishing research timelines and R&D costs while streamlining clinical trial design, recruitment, and enrollment. Without more streamlined IND, BLA, and NDA review processes or a drastic increase in personnel, the FDA could quickly fall behind, which could significantly delay review timelines and patient access to new lifesaving therapeutics.
IND applications have increased significantly in recent years. As the Agency scales in order to meet the growing need, one way to alleviate bottleneck issues is to leverage AI for data review.
The FDA’s 2019 Technology Modernization Action Plan put in place the framework for the FDA to begin working to integrate emerging technology internally to drive efficiency and bring the organization up to speed with the industries it regulates. The FDA needs to double down on this effort and aggressively push to integrate AI as standard practice for data evaluation for medical product review for new applications and post-approval, postmarket review. In order to leverage this tech, the FDA should rapidly recruit new staff that are trained in data and software, AI and ML engineering, data science, and even mechanical engineering.
The Digital Health Center of Excellence also needs to spearhead the implementation of AI and advanced software programs internally to facilitate more aggressive integration and efficient product review across the medical product centers.
At the same time, Congress must work with the FDA to create a novel Software as Medical Device approval pathway capable of reviewing applications that leverage both foundational LLMs and specialty, smaller models. To do this, the Agency will need to require a baseline approval of the foundational model, and a less rigorous review for the smaller, specialty models, which will be subject to postmarket surveillance.
The FDA must be flexible, dynamic, and ready to embrace these new technologies.
From addressing staff shortages to tackling fraudulent claims to helping clinicians practice more effectively, AI can help make our healthcare system more efficient and cost-effective. According to a recent analysis in Health Affairs, administrative spending accounts for between 15 and 30 percent of medical spending. AI-driven efficiency in hospital workflows and documentation, implementation of payment policy, and claims review processes will help save dollars currently spent on manual, clerical work. This can free up resources to provide patient care, which will drive better health outcomes and enhance patient engagement. Policymakers and regulators need to work together to ensure the potential benefits of the AI revolution result in better outcomes that patients and providers can both experience and appreciate.
The Covid-19 pandemic exacerbated staff shortages in the healthcare system, but it also showed that different models of care delivery can be effective and safe, and lead to improved patient outcomes. In addition to prioritizing recruitment and retention of healthcare personnel, Congress should promote AI to enable healthcare workers to do their job more effectively and efficiently, ensuring that their time is focused on care delivery to patients and not unnecessary paperwork.
AI can also benefit Congress’ drive toward price transparency and may allow patients to compare costs and outcomes to become more prudent purchasers of value-based care. AI is already streamlining data procurement, aggregation, and reporting. It is part of a movement to give patients more ownership over their health data and to use their data more effectively to manage their clinical needs and costs. Policymakers and regulators should utilize appropriate enforcement authority that balances the need to avoid harm to patients and the public, while incentivizing the development and implementation of AI-based interventions that can improve outcomes and reduce costs.
Algorithmic bias in AI has received significant attention from the Biden Administration. The White House’s Blueprint for an AI Bill of Rights included a discussion on bias alleviation standards, and HHS included a bias alleviation proposal in a recent proposed rule.
Bias is a knock-on effect of the data used to train the AI model; because the healthcare system has inequities in care, these inequities can be reproduced in data and therefore in AI models built on that data. But this isn’t an inevitability. As Dr. Mark Sendak of the Duke Institute for Health Innovation told NPR (and Dr. Micky Tripathi, the National Coordinator for Health Information Technology at HHS emphasized on Twitter), addressing bias requires “look[ing] in the mirror.” He noted that addressing this bias “requires you to ask hard questions of yourself, of the people you work with, the organizations you’re a part of. Because if you’re actually looking for bias in algorithms, the root cause of a lot of the bias is inequities in care.”
In other words: any type of bias is a legitimate concern. But bias in AI is not insurmountable—it can be addressed with comprehensive data, plus continuous cross-checking of the data for bias and evaluation of the performance of the derivative algorithms. If bias is found in the output of an algorithm, go back and check the data.
Providers, academic medical centers, and industry are working together with regulators through the Coalition for Health AI to promote ways that the federal government can eliminate AI bias. For example, providing a food nutrition-like label with clinical decision support software that could ensure providers are aware of the specific patient populations that would benefit from use of an AI-powered algorithm. Policymakers and regulators should not slow down AI-integration in health care for fear of bias. They must promote bias prevention, detection, and mitigation.
You’ve heard the warnings about AI. Like any new technology, AI offers both opportunities and challenges. The AI doomsday scenario preys upon our natural human fears that unscrupulous individuals will leverage good technology to do bad things. In healthcare, however, legislators have a once-in-a-lifetime opportunity to safely and effectively usher the most significant advancement of our time in life sciences and healthcare. For those who joined government to leave a lasting legacy and positive impact on humanity, AI is your chance. Let’s leverage AI to make Americans healthier.
Colin Rom is the public policy lead on the Bio + Health team, building a16z into a leading voice of health innovation in Washington, D.C.
Jeff Clark was the associate general counsel on the Bio + Health team.
Jay Rughani is an investing partner on the Bio + Health team at Andreessen Horowitz, focused on AI and data products across healthcare and life sciences.
Vijay Pande is the founding general partner of the Bio + Health team at Andreessen Horowitz, focused on the cross-section of biology and computer science.