AI in Health Care

Let’s distinguish the difference between health technology and artificial intelligence in health care, because they are not the same thing.

Health technology

This refers to any device that is programmed by a human, through coding, to perform a specific task. It can record, like a Holter monitor that tracks heartbeats. Health tech can react to predictable events, like a pacemaker that responds to the needs of a heart.

Although the systems may appear to have intelligence, they were simply designed and constructed by intelligent people to follow a human-designed path. If there were a failure in the technology, it comes down to human error in the programming or a defect in the mechanics.

AI is different

AI refers to the capability of a machine or software to simulate intelligent human behaviour—including instantaneous calculations, problem-solving, and evaluation of new data based on previously assessed data—and to perform tasks in real time.

A subset of the general heading AI, machine learning refers to technologies that allow systems to identify patterns and improve themselves through experience and data. In other words, machine learning allows the program to evolve without further human input.

Current uses

Health care services use vast amounts of data and increasingly rely on information technology and analytics. At minimum, AI tools can help save time for record-keeping and research, and potentially improve patient outcomes and reduce costs.

AI is being used to analyze data from electronic health records to provide clinical decision support, and it’s particularly useful in image analysis. For example, a Stanford University study tested an AI algorithm against dermatologists to detect skin cancers, and the program performed at the level of the humans.

AI has been used to analyze data from pre-op medical records to guide a surgeon’s instrument during surgery, which may lead to more precise and less invasive surgeries—and could reduce both healing time and costs.

New natural language (voice) understanding programs can capture and record provider-patient interactions during telehealth appointments to reduce record-keeping time. Other trials suggest that natural language programs can sort through vocabulary and tone of voice in 911 calls to determine the appropriate emergency response.

Limitations of AI

Who hasn’t consulted Dr. Google with a symptom checklist, only to arrive at a variety of life-threatening potential outcomes? Most of the time, thankfully, those diagnoses aren’t right. Your home computer and the data at its disposal can’t take into account all the subtleties of your physiology, biology, behaviour, environmental influences, or mental outlook.

Your own internet health care exploration is a simplistic example of the limitations of AI to generate personalized health care information and advice: computer programs simply miss the nuances of being human.

Lessons from personal experience

When I was about seven months pregnant with my second child, I went to my obstetrician because I was feeling some discomfort that didn’t make sense to me. We sat face to face as he listened to my concerns, and when I was finished talking, he said, simply: “I don’t like the way you look. Let’s get you to the hospital for some tests.” A computer can’t look at a patient and make that kind of intuitive assessment.

To err is human

The same subtleties that make us different as patients and practitioners can also influence programming of the codes that run these AI systems. Bias might unconsciously shape a programmer’s viewpoint in a way that excludes important data.

A recently published analysis of a health care algorithm showed that black patients were less likely to be identified by the algorithm as candidates for potentially beneficial care programs than were white patients with the same number of chronic illnesses.

Researchers concluded that category labels used in creating the algorithm were to blame for the error and suggested that changing the labels would solve the problem. At the same time, however, researchers also noted that swapping out the label may resolve one type of bias only to introduce another. In other words, the algorithms can’t be perfect.

Ethical considerations

Ethicists argue that before integrating AI within the health care system, developers, practitioners, and specialists should consider its impacts on medical ethics and humanity and how the negative aspects weigh against the positives.

For example, while AI might improve speed of access to information, it may also exacerbate social inequalities around the world or trigger economic losses right in our own backyards as workers, including bookkeepers and managers, lose employment or suffer wage cuts.

As we potentially advance into more automated health services, we must consider that quality of care, compliance, and healing may also be negatively affected if patients lose the empathetic and compassionate touch of human practitioners in exchange for the efficiencies of robotic physicians and nurses. Some patients—including mental health patients, obstetrics patients, gerontology patients, and children—may adversely react to robotic care.

It’s also important to be aware of and protect against potential abuses. The use of AI in health care requires the collection and analysis of vast amounts of sensitive patient data. This raises significant concerns about data privacy, security breaches, and the potential misuse of information.

Where is the line?

Individually, we must also advocate for our own vision of the future. AI could be the slippery slope that ushers in a transhuman agenda. Transhumanism is a movement that advocates using emerging technologies not simply to eradicate aging and disease but instead to speed evolution so we become an enhanced species that transcends humanity.

As some ethicists have argued, transhumanism smacks of eugenics and brings about complex questions on nature, including economic (who benefits?), sociological (will invasive control of body and mind arise?), and ethical points of view (if augmented capacity becomes the norm, will others feel coerced to make similar changes?). Will we have autonomy?

Critical thinking is key

Watch for politicizing of AI and technology in health care, as governments and other stakeholders continue to opportunistically ride the fallout of recent lockdowns, the fragile state of our health care systems, and both sides of the climate change narrative to potentially rush through introductions of AI into our health systems.

It is up to each of us to perform our own due diligence about the future of AI and health tech in the human experience, and not simply to let the developers and algorithms run amok.

Real-time input for surgeons

A machine learning algorithm can analyze 3-D scans up to 1,000 times faster and can provide almost real-time input for surgeons who are operating.

Promise for diagnosing cancer

Results of early tests on a deep learning algorithm indicate it can outperform humans when identifying breast cancer metastasis.

Virtual reality (VR) in medicine

VR headsets allow doctors and surgeons to become familiar with the inner workings of the human body without risk to patients or a steady supply of medical cadavers.

VR is also a useful therapeutic tool in treating chronic pain and anxiety, as well as teaching coping skills to children with autism.

Origin of the word “robot”

Karel Čapek coined the term “robot” from the Czech word robota, or “forced labour.”

AI can’t change human behaviour

Those anticipating AI-driven health care transformation are likely to be disappointed, because it won’t address the biggest health problem: human health behaviours.

This article was originally published in the January 2024 issue of alive magazine.