AI, in conjunction with predictive analytics and high-performance computing, is disrupting practically every industry you can imagine and healthcare is no exception, with advances in diagnostics already underway.
While android medics and lasers that seal wounds without scarring are still a long way off, in the next couple of decades, we can look forward to a wide range of innovations in patient care that will ultimately save lives and perhaps extend human lifespans.
The healthcare industry is well-known for resisting new technology or at the opposite end of the scale, blindly becoming early adopters of unproven solutions.
Of course, before AI can realize its full potential in healthcare, much needs to change. Can you speculate on what’s needed to ensure healthcare professionals are prepared for AI advances? What will the future hold in terms of clinical care? What will actually change?
Preparing For AI Advances In Healthcare
Futurists and healthcare experts believe AI will transform the healthcare landscape in the next 20-30 years but it will not be a quick change, given that technology adoption and integration never occurs overnight. Some preparation is needed.
AI will change the healthcare system and affect physicians and hospitals, predicted Alan Cudney, a principal healthcare consultant in SAS’ Health and Life Sciences global practice.
Before this happens, infrastructure upgrades are required.
“Just as electronic medical records required major focus and development of infrastructure, AI will require similar prioritization and growth. This will continue to exacerbate the consolidation of healthcare institutions as significant infrastructure will be needed to make the most of the technological resources inherent in AI,” said Cudney.
Staffing is another consideration.
“AI must rest upon a mature analytic platform that is accompanied by powerful software, skilled analysts, clinicians and data scientists,” said Cudney.
Training in the use of new technology is always essential.
“Physicians, clinical leaders, administrators and other healthcare professionals will have to become "data savvy" to make the best use of the new technologies,” said Cudney, adding that, “The importance of the role of data scientists will continue to grow, amid limited resources.”
An important point, as even now, skilled data scientists are in high demand and can command high salaries.
If the groundwork is in place, “AI will enable the delivery of personalized, proactive health care that is more efficient, more effective and less expensive,” said Cudney.
Future Applications of AI In Healthcare
According to Cudney, the use of AI in Healthcare (he also provided practical examples) will include but is not limited to:
1. Improved Treatment Decisions
“Assuming centralized data collection is the norm, clinical decisions will be based on a patient’s prior medical history and historical data on others with a similar genetic makeup. This allows clinicians to select the treatment option with the best chance of success,” said Cudney.
For example, Mary’s doctor evaluates treatment options to better manage her diabetes. He can instantly see her complete medical history, including a related hospitalization in a different city, her prescription refill history, and her completion of diabetes management education. As he begins to place an order for a new type of insulin, the clinical system responds that, due to Mary's clinical history and genetic map, she would be only 39 percent likely to successfully manage her diabetes with this choice of drug.
2. Predictive Analytics
“Predictive analytics will allow generation of an evolving treatment model, one that changes as new data is added. Forecasts will become increasingly accurate as data volume grows. Patients will receive more personalized, proactive care interventions,” said Cudney.
For example, using data sources that are aggregated and integrated across the continuum of care, an accountable care organization can utilize the predictive model to effectively and efficiently intervene, helping groups of patients get back on track with their “LifeCare” plans. Machine learning is used to update the predictive models as new data are received. The optimized models then re-run the cohort selection, risk stratification, and subsequent intervention assignment.
3. Behavioral Indicators
“By analyzing behavioral indicators on social media and by monitoring shopping habits and aggregating health-related data, clinicians can confirm that patients maintain a healthy lifestyle, thereby reducing risk. By proactively sending patient alerts, healthcare professionals can send medication reminders and point out habits that aggravate their conditions. These reminders are customized to the type and method of messaging that has been effective for the patient previously,” said Cudney.
For example, Ron hates to fill out surveys and has missed two of his last five doctor appointments. He doesn't use voice mail but responds to text messages and automated reminders from his cell phone to take his medication. Artificial intelligence is used to continuously optimize a predictive model, which accurately predicts the best way to communicate with Ron and influence healthy behaviors, based on his previous activities.
4. Inventory Management
“In the same way that Amazon predicts future orders, medical supply companies will utilize predictive analytics to stock their warehouses and eliminate shipping delays to their customers. They will effectively predict where specialized drugs and equipment will be needed so that care is not delayed and costs of carrying inventory are minimized,” said Cudney
For example, demand for home oxygen is predicted to peak in a six-county area of Michigan, according to public health data and hospital data. Artificial intelligence quickly understands and predicts the future need, suggesting redirection of home oxygen canisters and delivery personnel to that geographic area. The needs are anticipated and response is initiated in the most rapid manner possible.
5. Drug Prescription Analysis
“The same predictive techniques will be used by pharmaceutical companies to customize prescription drugs, to match unique patient characteristics,” said Cudney.
For example, Marla is taking a new medication to control her hypertension. The molecular structure of the medication as well as the dosage have been slightly modified to account for her genetic predispositions and previous response to other anti-hypertensive medications.
6. Image Analysis And Diagnostics
“Academic medical centers and other care delivery organizations will use automated image analysis, supported by deep learning of radiological images. Combining insights from these types of analyses with more traditional analytics will help providers standardize diagnosis and assist radiologists, increase efficiency, and reduce costs. Physicians as academic medical centers will use automated image analysis to help with detection of unusual lesions and to support medical training,” said Cudney.
For example, hereditary hemorrhagic telangiectasia (HHT) is often missed on imaging studies. The University of SAS Medical Center, a leading HHT treatment center, has stored images of arteriovenous malformations (AVMs) in the lungs of hundreds of HHT patients. These support an AI model that advises radiologists when an AVM is in the early stages of development.
Given all the potential advances outlined by Cudney, it’s clear that the benefits cover many aspects of clinical care. The question is, are healthcare institutions and their staff ready for change?
7. Planning For New Tech
Digital transformation has taken place in healthcare in many countries, leaving paper-based medical record a memory. The same approach will be needed to transition from recording information digitally to embracing AI and associated technology.
“Hospitals and care delivery organizations need to develop the analytic maturity necessary to take advantage of the coming analytic advancements. The continued implementation of value-based care will make it imperative to optimize every clinical decision and patient interaction, so as to drive the most value,” said Cudney.
This will involve implementing the necessary software, purchasing access to use software, or simply outsourcing the analytic capability to realize this level of maturity, he added.
“It will also involve training all clinicians, leaders and operational staff in basic skills of working with data, as well as setting organizational policy to incorporate analytic models and processes into daily operations,” said Cudney, adding that, “Finally, these organizations will need to have consistent data governance and data management practices to ensure the viability and quality of models that arise through AI.”
The last is key, for AI to provide the previously mentioned benefits in the future, robust data privacy and compliance requirements are necessary, to ensure personally identifiable information (PII) is removed where applicable. It’s all well and good to harvest medical data but it must be performed in an ethical manner. Nothing enforces ethics better than heavy penalties for breaking compliance regulations.
In conclusion, in an era where robot surgeons are firmly fixed in science fiction (yet in development), AI is the means of bringing all of us one step closer and offers improved healthcare to all as the technology evolves. In the next 20 or 30 years, or perhaps sooner, all predictions will prove true and if I don’t have to visit an iDoc or iSurgeon for medical advice, I’ll be okay with the progress made in this area.