By STEVEN ZECOLA
On December 19th, the Department of Health and Human Services (“HHS”) issued a Request for Information seeking to harness artificial intelligence (“AI”) to deflate health care costs and make America healthy again.
As described herein, AI can be used in many dimensions to help lower healthcare costs and improve care. However, to achieve significant breakthroughs with AI, HHS will need to completely revamp the regulatory approach to drug discovery and development.
Dimension #1. Incorporation of AI into Drug Discovery
The biggest benefit to the healthcare industry’s performance from AI is achievable from drug discovery. Accounting for the costs of failures, the average FDA drug approval costs society almost $3 billion and takes decades to reach the market from its inception in the lab.
In contrast, AI identifies potential treatments much faster than traditional methods by processing vast amounts of biological data, uncovering hidden causal relationships, and generating new actionable insights.
AI is particularly promising for complex, multifactorial conditions – such as neurodegenerative diseases, autism spectrum disorders, and multiple chronic illnesses – where conventional reductionist approaches have failed.
In the short-run, HHS should direct its grants toward AI-generated basic research, with a particular emphasis on the hard-to-solve illnesses. At the same time, the FDA should be putting into place a new approval system for AI-initiated programs to enable breakthrough treatments in a compressed timetable.
Dimension #2. Incorporation of AI into the Drug Development Process
Simply relying on AI for drug discovery, while subjecting its advances to the current approval process would undermine the use of the technology.
Rather, improvements from AI can already be had in fulfilling the exhaustive regulatory documentation requirements, which today add up to as much as 30% of the cost of compliance.
In the short-run, AI can improve drug development by:
- Automating and validating regulatory documentation
- Enhancing trial design and participant stratification
- Monitoring safety and efficacy in near real-time
- Reducing administrative and compliance costs
For example, in the U.K., the Medicines and Healthcare Products Regulatory Agency reported that clinical trial approval times were twice as fast with AI and associated reforms.
To achieve much greater long-term gains, HHS should collapse all clinical work utilizing AI into one elongated trial rather than discrete Phase I, II and III trials, given that AI can be used to continually update and validate documentation. This change would not require statutory change or agency rulemaking because clinical trial design is not codified in the FDA’s rules.
As participants are added to a trial, safety results can be examined and reported in real time. Once the trial surpasses a certain number such as 1000 participants with proven efficacy and meeting the specified safety protocols, it would be approved for roll-out. The role of the government in such an approach would be as auditor to validate the output of the trial. This function would include experimental validation, mechanistic understanding, and ethical oversight.
With these changes, FDA personnel would shift from episodic gatekeepers to continuous auditors, which would require a fundamental change in organizational culture. While safety concerns would remain important, responsibility and accountability would be more equally shared with the applicants and trial participants. Additionally, the prolonged suffering of existing patients would be factored into the public welfare analysis in reviewing preliminary safety results.
Dimension #3. Enhance Data Collection to Empower AI
Comprehensive, and accurate, data is essential to AI’s success. Yet this another area where the healthcare industry has failed.
The industry has evolved with each provider, or family of providers, encouraging their patients to sign up for a customer portal. The providers typically treat the information on those portals as their own for purposes of research. However, the providers do not own the data. Each patient owns his or her data.
To broaden the scope and applicability of healthcare data, HHS should establish national standards for patient-facing data collection that:
- Use interoperable formats
- Capture both diagnostic outcomes and relevant explanatory variables
- Preserve patient ownership and informed consent
- Enable longitudinal tracking while protecting privacy and security
Once this format is established, HHS should establish a goal of enrolling 100,000 participants within two years.
Dimension #4. Use of AI to Establish Standards of Care and Price Ceilings
There are no national standards of care for diseases or other health maladies in the United States. Patients oftentimes do not understand the nature of their affliction, the options to treat it, or the costs of the various options to remedy it.
On a parallel track, HHS might fund basic research targeted to a particular ailment, the FDA might (or might not) approve it, Medicare might (or might not) cover it, and some insurance companies may cover the treatment and some may not.
Moreover, the costs of various treatments may vary greatly from facility to facility—unbeknownst to the patient.
Layered on top of this market dysfunction, healthcare practitioners have the desire (and the economic incentive) to provide the best (and likely the most expensive) possible service to their patients.
In short, there is a market failure, primarily relating to a lack of actionable information.
In the short-run, AI can help address these failures by aggregating and analyzing how care is delivered across the country and identifying patterns associated with better outcomes and lower costs. These insights could be used to inform evidence-based minimum standards of care and improve transparencies around pricing and performance.
Over the longer term, the outputs of these systems could be used to establish a minimum standard of care for all (or most) ailments. These standards would be mandatorily covered by insurance. Concurrently, the outputs for these standards of care could be supplemented by regional price ceilings for the various practices based on a comprehensive industry analysis.
As experience is gained from these informational AI systems, a future version could be programmed to automatically calculate the prescribed minimum standards of care and the price ceilings to mimic the functioning of demand and supply curves. An algorithm could be constructed using a specified level of subsidy provided by the federal government as the equilibrium. As the federal subsidy exceeds certain pre-set limits, AI would be used to address the disequilibrium by providing to law makers various options that would lower the price ceiling for certain conditions and/or lower the minimum standard of care.
In scenarios where the stipulated federal subsidy was exceeded, some classes of patients would be denied receiving payment for the best available treatment (unless they had supplemental insurance) and/or some healthcare providers would suffer a diminution of profits.
Such an approach would require Congressional approval, but such tradeoffs are occurring now—without informed choices. In this dimension, AI could be used to address the industry’s massive information failure and tackle the ever-increasing subsidies.
Dimension #5. Incorporation of AI into HHS’s Internal Processes
AI can also improve the efficiency and effectiveness of HHS’s internal operations. While the potential percentage gains would be smaller than that for the discovery and development dimensions, even modest improvements can yield meaningful savings given the scale of federal healthcare spending.
Conclusion
AI offers the opportunity for significant improvements in healthcare outcomes and efficiencies—but only if it is integrated into a regulatory and governance framework designed for its capabilities. Shoehorning AI into existing structures will blunt its impact and increase the risk of implementation.
Each dimension described above requires a separate dedicated, multidisciplinary team reporting to the Office of the Deputy Secretary. After the strategic direction for each dimension is established, these teams should be tasked with:
- Developing detailed implementation plans, including budgetary requirements
- Identifying any statutory or regulatory barriers
- Establishing timelines, milestones, and evaluation criteria
- Addressing ethical and equity considerations
Drug discovery and drug development represent the highest-impact dimensions for AI implementation. HHS should make use of external expertise in fashioning the details of an appropriate regulatory framework for these dimensions.
The detailed plans for implementing AI should be approved and finalized before the end of 2026. As described herein, HHS should take a proactive, forward-looking role in harnessing AI to constrain healthcare costs and improve care.
Steve Zecola sold his web application and hosting business when he was diagnosed with Parkinson’s disease twenty three years ago. Since then, he has run a consulting practice, taught in graduate business school, and exercised extensively
