AI In Healthcare And Other Scary Stories

No greater buzzword exists than Artificial Intelligence (AI) to make a story headline flash. In healthcare, stories about AI being used to discover new drugs and faster, more accurate medical diagnoses are guaranteed clickbait. These goals may be the stuff of dreams, but make no mistake, AI is already here and it’s a moneymaker for the healthcare industry, and we need to understand how.

What Is AI?

First, it is important to define AI. Though a seemingly new addition to the lexicon, AI is not new. The discipline of combining computer science with robust datasets to facilitate problem solving began three quarters of a century ago. The earliest successful AI program was written in 1951 by Christopher Strachey, later director of the Programming Research Group at the University of Oxford. Today’s key to AI strength and success lies in access to robust datasets.

Data Robustness is the overall degree to which a given dataset can tolerate variations in its collection and integration procedures without suffering a loss of information content, statistical validity and/or scientific meaning. Having volumes of data points is not equivalent to robust data. Instead, robust data refers to the quality of data collected and to whether it is weak or strong. Contrary to common opinion, few in the financial markets have access to truly robust databases save perhaps the top Quantitative Trading Hedge Funds and Private Equity (PE) firms with algorithmic analytical modeling.

Why is Healthcare so compelling for AI? With its long history of insurance archives, marketing data, intellectual property precedents, and regulatory drug approval guidance, no other endeavor has as many data points to consider. Think of the recent large pharma Inflation Reduction Act lawsuits and why preemptive legal actions from multiple companies should be as swift? Unless there was no favorable stochasticity to suggest other corporate actions, the data die was cast and to believe other than pecuniary forces matter isn’t rational.

How Does AI Work?

AI produces its results in the form of machine learning algorithms. The algorithms identify entities, relationships, and clusters. Algorithms based on historical data have been around for a relatively long time, however, the use of AI in the development of stochastic algorithms is a new, stimulating, and fraught undertaking. In this new arena, AI analyzes randomness and probability to optimize results when random occurrences can affect the outcome.

Healthcare and AI

According to Built In, an online community for startups and tech companies, the potential beneficial applications of AI in health care look to include improving medical diagnosis, speeding drug discovery, transforming patient experience, managing healthcare data and performing robotic surgery. Despite these laudable goals, the current uses of AI are not so altruistic.

Without equivocation, today Big Pharma has focused AI primarily on making marketing decisions. It helps companies decide on the best marketing strategy for each of their drugs based on analysis of market size and the highest price point tolerance. For established drugs, AI can use contemporaneous data, e.g., the prevalence and demographics of a particular disease or condition targeted by the drug, and compare that with past marketing campaigns to identify the best approach under current conditions. For new drugs, the process is similar with an emphasis on the demographic most likely to use them and, most importantly, to identify the payers for the new therapy.

Hospitals (more and more for-profit institutions), like other corporations, seek to maximize profits in the name of efficiency. That is a coded way of saying maximize profits to satisfy investors (public or private). The public expects hospitals to be more than efficient money-making businesses. Yet in all too many cases, the drive for profitability supersedes the delivery of quality, patient-centered health care.

As hospitals consolidate to form conglomerate behemoths, they rely heavily on AI’s profit/efficiency business model. The most lucrative of these amalgams use AI to optimize staffing levels, predict patient revenue maximization, achieve economies of scale, market power, risk diversification and financial synergy. It is not important that investors and hospital administrators have little or no medical expertise because AI can produce algorithms that adapt, in real-time, to ongoing changes in patient access and reimbursement protocols. As this process continues, patients will become little more than revenue generating cogs in the giant wheels of an AI fueled powerhouse.

Tempering AI

AI is neither going away nor is it going to self-regulate. The reason for these realities is that the investors, managers, endowments, pensions and municipal funds pouring money into these health care ventures look for one thing, and one thing only, double digit returns on their investments. While some may pay lip service to investing in improvement of health care, their real goal is not healthier patients but rather a healthier bottom line.

Unwittingly, aiding and abetting this drive for profit in the health care sector is the United States Government (USG) with its seemingly unlimited dispensing of cash through Medicare and Medicaid. No individual or individual organization can tame AI in the health care arena. Only the largest health care payer can do that. The USG can and must step up to ensure that AI is used to benefit patients not just health care investors.

Given that the USG has the power of the reimbursement purse, one would reasonably expect that the government is actively involved in AI use and regulation. Nothing could be further from reality. There is little or no communication, much less seamless communication, between governmental health care agencies.

The Scary Part

Given the COVID pandemic, every American is aware of how the Centers for Disease Control and Prevention (CDC) is charged with protecting America from health, safety and security threats both foreign and domestic. The CDC is “organized” into 12 Centers and Institutes, 11 Offices and 7 Director Divisions. Instead, it provides a glaring example of governmental disarray. For a bureaucracy of this size to function with any degree of efficiency requires, at a minimum, real-time communication between each unit if one hopes to limit duplication of effort, waste and in-coordination. As an example of the chasm between the USG and AI powered efficiency, a significant means of data transmission from individual states to the CDC is via fax. The transfer of information from fax requires manual entry for digitization. This process isn’t just inefficient, time-consuming, and expensive; it is subject to significant error.

Think of it this way, the CDC has grown up to be the World’s health data aggregator, but it’s hard to aggregate information by facsimile technology and worse, it’s impossible to make it work real-time. Is it any wonder that there have been so many pandemic response missteps when the agencies charged with providing the public with best practices cannot even collect accurate data? Forget about AI.

The inability of the USG to incorporate AI has far reaching consequences. And as proof of the impact of these primordial shortcomings, look no further than how the USG tolerates crucial chemotherapy drug shortages.

Despite the usual explanations for drug shortages – supply chain disruptions, pandemic effects, increased demand – the reality is that many cancer drugs have very low profit margins because they are relatively easy to produce. Hearken back to the AI profit/efficiency model and it is easy to understand that such medications are not money makers. What better way to improve a profit margin than by not providing or even eliminating low return medications? Unless the USG intervenes, low-return interventions such as chemotherapy will continue to be eliminated in the name of a better bottom line. Patients will bear the burden of such governmental inaction.

What it comes down to is AI has remarkable potential to improve health care for patients and investors but that will not happen unless the USG steps in. The administration must upgrade its own systems and incorporate AI in the process. Next it must promote the beneficial aspects of AI use in health care while disincentivizing the profit-at-all-costs schemes currently growing in number by the day because it is possible to improve both health care and bottom lines.

As AI turns the page of history, its development is inevitable, but it can and must be tempered.