Intelligent Technologies in the World of Health
«Intelligent technologies» is a term applied to artificial intelligence (AI) and other computer-based intelligence systems increasingly making their way into business, engineering, and military operations. For practitioners and others in healthcare, the question is not whether these systems will also begin to reshape contemporary medicine, but when and to what degree.
Such a transformation may appear inevitable now, but that wasn’t always the case, according to Jacqueline Laver, chief technology officer for eHealthAnalytics, a New York–based company comprising world-class specialists in machine learning (a type of AI), data architecture, gerontology, cardiology, and information technology.
«There was an initial overhyping of AI starting in the 1980s, where we made many predictions and failed to deliver on any of that, so people backed off,» said Laver. «But then Amazon later made machine learning practical for marketing, using it to predict what consumers would want. It has been used a lot in banking too, but medicine has been fairly slow to adopt AI technology.»
Laver attributes this delay to medicine’s tendency to want to leave all decisions to physicians, but she sees the economics of healthcare as an area where AI is likely to chip away at that hesitation.
«Using machine learning, we now can provide tools for improving medical devices and medical care and for personalizing medicine, so AI can help the healthcare system draw better results from each dollar.»
AI’s game-changing potential goes far beyond money, however. The sheer volume of data influencing diagnostics and predictions of complications, drug reactions, disease outbreaks, and other events means that AI is becoming a necessary tool to guide health activities ranging from the care of individual patients to the management of global pandemics. Unlike the traditional randomized control study, which requires a hypothesis that is then either disproved or verified, AI can work more like the human brain by adding new ideas based on trends from a multitude of data sources, including social media feeds.
In healthcare, AI’s predictive powers can be harnessed using more traditional data obtained from medical devices and clinical trial outcomes as well as from everyday sources like the popular wearable devices (eg, Fitbit). This can then be used to calculate the chances of valuable outcomes that are usually outside the scope of clinical trials, such as a patient’s likely recovery trajectory or whether they might develop complications or a chronic disease like type 2 diabetes.
«One example is a rheumatology AI system that took patient symptoms and signs, such as inflammation and pain, along with intervention information to predict not just patient recovery, but also behavior of doctors, such as the likelihood that a given doctor would prescribe a particular drug,» says Lorien Pratt, PhD, chief scientist at the California-based Quantellia LLC.
Intelligent Tech’s Many Promising Forms
Although AI is a newcomer in medicine, with still unrealized potential, both Pratt and Laver emphasize that it is not the only intelligent technology that could revolutionize healthcare. «Decision intelligence is also coming on the scene,» Laver points out.
Colloquially, AI simply refers to any computer system that performs tasks that usually require human intelligence. However, decision intelligence (DI) seeks to unify all of the other intelligence technologies—not just AI but also business intelligence, collaborative intelligence (people working in concert over the Internet to solve problems), systems analysis, and decision analysis.
«DI is trying to enforce some conformity and pulling together the other intelligent technologies into a coherent system, or at least methodology/understanding, that solves the essential problem, to understand how decisions, which lead to actions, ultimately lead to outcomes,» Pratt says. «As with other fields, the question in healthcare is how do we use all of these technologies in a coherent way to get better models for how decisions and actions lead to outcomes.»
To be sure, AI and machine learning (which applies AI to improve the function of a computer system without the need for new software) are spectacular at revealing associations and making predictions, illuminating a causal chain that may have otherwise not been apparent. But, in contrast, DI reveals the impact of actions, combining AI’s predictions with other information and analysis to navigate the various outcomes of different decisions. Consequently, DI has been applied in engineering and business to make high-level decisions through a process that involves analyzing their various potential consequences.
DI now beckons in the world of health, where in the most extreme form one might envision the technology making patient management decisions that to this point have been left to physicians, such as choosing medication and dosage, or even choosing between surgical procedures. The healthcare industry might balk at such an application, so it’s possible that DI technology will arrive sooner in researcher settings.
AI in Practice
To illustrate how this might look, consider an issue like the impact of seatbelt laws. It’s a well-trodden topic, with data available from numerous fields ranging from trauma surgery to social work. DI has potential to sort through these data and take a big-picture approach by comparing and contrasting chains of events resulting from a series of decisions. Based on the result of such analysis, a DI system then could offer appropriate policy and treatment recommendations.
Considering an even more complex issue, global pandemics, AI is extremely advanced in that it can unify very different types of information obtained from various sources (eg, vaccine availability, insect control methods), such that it could be used to help predict disease outbreaks. But DI can take things even further.
«Once we have predicted a disease outbreak, there’s probably a huge importance of tradeoff between decisions,» Pratt points out.
Then there are seemingly more straightforward questions, such as how to balance resources for distributing medications versus training doctors. These decisions depend not only on the cost of each option, but also on a plethora of societal and behavioral factors. When it comes to HIV drugs and other medications that must be taken on precise schedules, the optimal use of sources will be different for an urban area of a developed country versus certain rural areas of developing countries where many people do not wear watches.
In other words, DI is a big-picture technology, a problem solver. Only time will tell whether this leads to a paradigm shift on par with the advent of antibiotics, vaccinations, or purification of water supplies.
A Product Manager with expertise in pharma marketing and sales operations