AI could save lives and cut health care costs

Generative AI could save hundreds of thousands of lives, make health care affordable for every American, and let clinicians spend more time with their patients. But this won't happen unless our nation embraces the opportunities this technology makes possible.

The need for swift and bold action has never been greater. With average medical costs now topping $14,000 per person, U.S. health care is both unaffordable and unsustainable. Employer health-insurance premiums are expected to rise 9% this year, approaching $30,000 for a family of four.

Yet despite spending far more than other wealthy nations, the United States ranks last in life expectancy, maternal mortality, and childhood deaths.

Generative AI can help reverse these trends.

Here are five ways this technology could make care safer, smarter and more affordable:

1. Closing medicine's knowledge gap

Every 26 seconds, a new medical study is published, more than 5,000 a day. No clinician, no matter how diligent, can keep up. Generative AI can.

By scanning and synthesizing the entire body of medical literature in seconds, it can surface the latest evidence and help clinicians make faster, more accurate decisions. Already, three in five health care professionals use GenAI regularly, and among medical students and residents, adoption is nearly universal.

Soon, patients won't worry when their clinicians consult generative AI for guidance. They'll expect it.

2. Empowering patients everywhere

American medicine is built on a 20th-century model that assumes the best place for care is a doctor's office or hospital. That worked when most illnesses were short-term and acute. But today, 60% of Americans live with at least one chronic disease such as diabetes, hypertension, heart failure -- and 42% have two or more.

These lifelong conditions can't be effectively managed through brief office visits every few months, as is the current standard.

When connected to wearable devices, GenAI would analyze blood pressure, glucose, and oxygen levels in real time, alerting patients and clinicians to early signs of trouble. For those with heart failure, it would identify fluid buildup days before an emergency, allowing for timely treatment.

3. Preventing medical errors

Even when clinicians know the right steps to take, the pace and pressure of modern medicine make it hard to follow every safety protocol every time. The result is predictable and devastating: medication errors, wrong-site surgeries, and preventable hospital-acquired infections that kill an estimated 371,000 Americans each year.

Generative AI can close the deadly "knowing-doing" gap. Using multimodal capabilities -- vision, speech, and text analysis -- it could observe care in real time and compare clinician actions to nationally accepted safety protocols. When a critical step is missed, the system would alert the clinician before the patient is harmed.

4. Delivering smarter, more personalized care

In medicine, "treating everyone the same" sounds fair, but it often produces worse outcomes for everyone. Patients who need close attention get overlooked while those recovering well are awakened in the middle of the night for unnecessary checks. Rather than discovering a patient's status is deteriorating after a bedside monitor signals a life-threatening crisis, this technological approach would notify them hours earlier.

Using bedside data (heart rate and rhythm, blood pressure, oxygen levels), GenAI would continuously calculate the rate of change for each. The information would allow nurses to focus on those patients who need immediate help and let others rest undisturbed. The result: earlier interventions, faster recoveries, and better sleep.

The same principle applies outside the hospital. By training GenAI systems on real (and anonymized) patient-clinician conversations, families everywhere could receive accurate and personalized medical guidance whenever symptoms arise.

5. Unlocking medicine's hidden data

American medicine is drowning in data yet starving for insight. Every bedside monitor, surgical procedure, and clinical interaction generates more information than any human could possibly retain or understand.

In a typical hospital, bedside monitors alone produce about a terabyte of data each year (that's a one followed by 12 zeros). An estimated 97% goes unused.

It's a vast, untapped gold mine that could accelerate discovery and improve patient care. Generative AI could investigate all that data to identify which diagnostic and treatment approaches deliver the best results.

It could also examine technical details from surgical robots and cardiology labs to determine how top-performing clinicians remove tumors or reopen arteries most effectively. In time, GenAI could even replicate their precision.

For decades, America has tried to fix its broken health care system through small, incremental changes. It hasn't worked. During that time, costs have tripled, clinical outcomes have stagnated, and access has eroded.

Generative AI offers a way forward. If fully embraced in clinical practice, it would help clinicians deliver safer, smarter, more affordable care for all Americans.

Robert Pearl, the author of "ChatGPT, MD," teaches at both the Stanford University School of Medicine and the Stanford Graduate School of Business.