When artificial intelligence (AI) first came to the scene, many individuals were hesitant to jump on board with the movement. And rightly so. After all, this invention seemed quite scary from the perspective of privacy and data retention. Not to mention the number of individuals who were terrified by the idea of machines with the ability to “think” for themselves.
As time went by, however, more and more people started realizing that AI indeed has many incredible benefits that can be used in almost every market. According to Philosophy professor and consultant William Jaworski, the healthcare industry is poised to start making use of these benefits to enhance patient care and potentially cut healthcare costs. How exactly is AI reshaping the healthcare market and what can one expect in the future?
When somebody goes to see their doctor because they have a problem, they will usually start with a list of symptoms. Then, the doctor will ask a plethora of follow-up questions that help rule out some major conditions and diseases. Ultimately, the doctor may order a few tests to help confirm the diagnosis by obtaining direct evidence of it. Although the process is much more extensive in reality, it is a great example of something that can be automated.
Instead of having a doctor or a nurse ask patients questions and jotting down their answers, an AI machine could handle the initial screening. It would include a natural language processing program to convert the patient’s responses into usable data which could then be analyzed by a process known as machine learning (ML). ML can operate in a number of different ways; for example, by finding similarities among a set of data points and clustering similar cases together, or by using data to construct models that predict patient outcomes. Based on the results of its analysis, the AI could order tests or give suggestions for a diagnosis. And because the process would be automated, it would be accomplished much faster than it would have been using conventional methods.
William Jaworski also points out that a diagnostic machine would not be influenced by the cognitive biases that affect human physicians. Cognitive biases are psychological tendencies that lead people to make systematic errors in judgment. Examples include the Anchoring Bias, the tendency to weight more heavily the information that one first hears about a case, and the Availability Bias, the tendency to think that examples that come readily to mind are more relevant than other, lesser known examples. Cognitive biases have been shown to lead to errors in medical diagnosis and treatment. Because an AI-based diagnostic machine would be free of cognitive biases, it would decrease or eliminate altogether the likelihood of medical errors due to bias.
There is already evidence that AI makes better use of data when interpreting radiological images than humans do. AI has also been shown to be effective at making diagnoses in oncology, neurology, and cardiology. An AI using a type of machine learning called “deep learning” was shown to be 90% accurate in diagnosing skin cancer from clinical images—an accuracy rate that rivals that of experienced human physicians. Similar results were achieved by a natural language processing system that was trained to diagnose cerebral aneurysm by identifying disease-related keywords in clinical notes. The future will likely see medical professionals making increased use of AI “consults” to make diagnoses and identify patients who are at risk for life-threatening conditions such as heart attack or stroke.
During surgeries, doctors often have to make a series of precision movements where any minor mistake could expose the patient to grave danger. AI has been tested as a way of assisting surgeons during actual procedures. The surgeon’s movements are translated into commands that are carried out by a robotic system which uses AI to compensate for hand tremors and other variations in the surgeon’s performance. Those variations can have an enormous impact on patient outcomes, and there is an increasing amount of data that AI-assisted procedures can improve those outcomes.
To date, the most successful uses of AI in healthcare have involved machines that enhance human performance rather than replacing it; it is nevertheless possible that the future could see AI-driven robots provide an alternative to human surgeons performing procedures that require a great deal of complexity and motor dexterity. Unlike a human, an AI-driven robot does not experience fatigue or distraction, its performance is not affected by a poor night’s sleep, or by emotional reactions to the patient or its own circumstances, and it has a practically non-existent margin of error.
Electronic Health Records
AI also has the potential to allow healthcare providers to spend more time doing their job, namely caring for patients instead of keeping records about them.
In the past, almost all patient files were stored in enormous warehouses for preservation purposes. During the 1990s and early 2000s, healthcare providers had begun converting patient information into electronic health records (EHRs). In 2004, approximately 21% of office-based physicians reported using EHRs. That number grew over the next decade, and by the end of 2015, 87% of office-based physicians reported using some form of EHR. (Source)
EHRs have many advantages over paper records. Among other things they make it much easier for healthcare providers in different practices to access a patient’s medical history and other information. But the adoption of EHRs has come at a cost: inputting patient information is very time-consuming. Providers can often spend more time updating patient records than caring for the patients themselves.
Enter AI! Instead of having healthcare providers manually input every observation and statistic about a patient, it is possible to have an AI-based system do it. For example, a machine equipped with a microphone and video camera could accompany a physician to record an initial examination. AI could then be used to extract from the audio and visual record all the information needed to update the patient’s EHR automatically.
Better Recognition and Analysis
As far as the future goes, it is not unrealistic to think that in the foreseeable future many doctors will be utilizing AI in their daily practice. This includes everything from online machines that handle low-skilled labor like data input to machines that perform tasks that have traditionally required high levels of medical training such as diagnosis and surgical decision-making.
Jaworski says that AI has the potential to help especially disadvantaged communities in which healthcare providers or specialists in a medical subfield are in short supply. He nevertheless adds a cautionary note. “AI can accomplish a wide range of tasks in a timeframe that humans cannot rival, but there are things that AI is still not good at, and that it may never be good at. Those include some of the most important factors in healthcare: making judgments about what is in people’s best interests. It’s one thing for a machine to calculate the probability that a certain treatment will be successful, and another to weigh that probability against the risks and potential impact on one’s family and quality of life.”