CEO of Think Tank Innovations, a communications solutions company, known for ShareSmart for virtual health appointments and consults.
Never in my life did I imagine Tesla’s self-driving cars, speaking to customer care bots or robots helping to assemble my Amazon order, but all of this has been made possible thanks to artificial intelligence (AI).
As AI improves, more industries are considering how to best implement it to improve efficiency and outcomes, including healthcare. The health sector stands to benefit greatly, as the increasing demands of an aging global population promises to continue to strain already tired systems. Medical AI creating automated processes in healthcare could significantly increase efficiency, reduce spending and enhance patient care and health outcomes. In fact, several AI models that can outperform physicians in radiology diagnostic accuracy already exist, ultimately reducing human error.
In medicine, AI is being employed in predictive analytics to explore patient datasets and forecast the likelihood of certain diseases/disorders. Emerging studies have shown that AI can detect traditionally difficult to identify or diagnose conditions, including rare hereditary and neurodegenerative diseases.
As AI evolves, external factors such as social determinants and lifestyle choices may be incorporated into these models. An individual’s health outcomes are estimated to roughly be influenced 40% to 50% by behavior and 20% to their physical and social environment, while genetics account for only 30%. Meaning, a model that evaluates an individual’s combined genetic and behavior/social data would greatly improve a physician’s ability to choose the best treatment path/medication for each individual. Furthermore, the interaction of complex social determinants that play a role in creating health disparities between ethnic groups and different socioeconomic groups could be captured and potentially rectified like never before.
Thought-provoking conversations arise when considering which diseases should be prioritized when building models. Medical conditions that result in different patient prognosis depending on when they are identified along the disease progression timeline are an optimal target for AI integration. This is especially critical for conditions where providers are missing key insights that allow them to diagnose with confidence. For example, developmental dysplasia of the hip has a significantly better prognosis when caught early; however, there are often little to no symptoms during early stages. Therefore, a predictive model that is able to consider relevant data and correctly decide an otherwise difficult diagnosis could be life-altering for many individuals.
With ever increasing pressures on healthcare systems, it’s time to ask the question: Where should priorities lie when directing predictive model efforts? There are various considerations to be made when posing this question, including, disease prevalence, life expectancy, cost to the system and patient triaging impacts, to name a few.
An interesting hurdle related to implementing predictive analytics in medicine is choosing when a model is ready. AI algorithms can continuously improve through machine learning, which is a subset of AI that allows computers to learn and improve without the need for human intervention. Learning is possible when the algorithms are exposed to more data, since it adjusts in accordance to the data it has been exposed to, much like operant conditioning. However, at what point do we trust an AI model to be the deciding body in a patient’s healthcare journey? What error margin are we willing to allow an AI model to have? Would it be at a similar rate to human errors made by healthcare professionals or should it be much higher for AI?
Several experts agree that the sky’s the limit with AI, since no boundary is imaginable. When considering the potential of AI in healthcare, it is pretty amazing that it could transform a highly reactive system into a proactive one. This could lead to improved patient outcomes, less human error, improved efficiency and lower healthcare spending. However, stringent boundaries must be decided upon to ensure careful guidance of AI’s rapid growth, especially in the healthcare industry. Conversations regarding the applications and implementations of Medical AI must take place in order to ensure this new technology is adopted properly into ethically responsible medical systems that are patient-centered.