Why Developing Decision Intelligence And Support Is The Crucial Next Step For AI In Healthcare
When it comes to innovation in healthcare, pretty much everyone is all-in on AI. Functions ranging from diagnostics and pharmaceutical development to appointment scheduling and doctor-patient interaction are all being machine-learning-enhanced and algorithmically influenced in myriad ways across the sector.
To be clear, much of what is commonly dubbed “artificial intelligence” today doesn’t really live up to the definition. In healthcare, particularly, plenty of purportedly AI-powered solutions merely represent long-overdue IT automation implementations. But even that embrace of automation is an important first step on a clear course to a truly AI-enabled future.
Make no mistake: AI will play a critical role in the evolution of modern healthcare.
The success of that transformation is going to require understanding what we need our technology to do — and changing the way we do things to ensure we enable it.
Where’s the real value?
To pave the way for AI success in healthcare, we need to grasp why it is necessary. What value does it bring?
The answer is that it solves a problem that is already battering the industry on a host of fronts — from doctors waylaid with data entry and managing apps instead of practicing medicine to institutions debilitated by brittle and/or fragmented computational infrastructure — we are increasingly digitally connected, device-dependent and drowning in data.
Essentially, AI is required to manage our rapidly changing relationship with information. As noted in the Wall Street Journal recently, the digital age has introduced a host of new issues to the sector, “chief among them is managing the deluge of data already threatening to overwhelm the healthcare system. AI and machine learning advances will be necessary to find important signals in the sea of noise that streams in from digital devices.”
What we need is help to master all that data. What we need is decision support. Therein lies the value of AI.
Decision Intelligence And Support
At both the individual or organizational level, decisions can be influenced by a multitude of experiences and biases, but are generally derived from a process of vetting information. Decision-making researcher and Princeton University psychology and computer science professor Tom Griffiths explains that “humans are constantly weighing up probabilities based on incoming information.” But our ability to do that and make logical decisions gets mucked up when there are too many competing probabilities for the human brain to compute.
Consider a problem where there may be 50 solutions to get to the desired endpoint. Given some time and access to a good stream of information on each solution, a team of people can reasonably determine probabilities and figure out which of the 50 is the best bet.
Now consider a problem where there are 50,000 options and 50,000 streams of information and, oh yeah, you must find the best bet in a matter of minutes.
That volume of complexity is beyond human vetting. But it is just the sort of standard situation the average healthcare organization faces in our big-data-driven world. Consuming flowing streams of massive amounts of information and finding patterns and ranking relevance and sorting variables and determinants for all those options is what our AI models can do very well for us. They can surface what is pertinent to support wise human decision-making.
Decision support is the immediate need driving AI development in healthcare today. Integrating data, analytics and AI allows the creation of decision intelligence platforms to support, augment, automate and speed decisions. Millions of dollars and literal lives are at stake. Near-term development focus in this effort will need to include:
• Data Fabric Design: Data is AI’s fuel, but it often remains ineffectively siloed within applications and it is also generally not organized to account for the importance of relationships. To unlock its potential, healthcare IT must embrace the data fabric design concept. Gartner defines data fabric as “an integrated layer (fabric) of data and connecting processes. A data fabric utilizes continuous analytics over existing, discoverable and inferences metadata assets to support the design, deployment and utilization of integrated and reusable data across all environments.” The real value of data exists not in simply having it, but in how it’s used for AI models, analytics and insight. A data fabric makes data available everywhere it’s needed.
• Cybersecurity Architecture Meshing: Digital business assets are distributed across cloud and physical data centers, devices and endpoints, and the healthcare industry’s digital assets remain a prime target for cyberattacks and ransomware. Traditional, fragmented security approaches focus on enterprise perimeters that no longer apply. A cybersecurity mesh architecture provides a composable approach to security based on identity across endpoints to create a scalable and interoperable service. The common integrated structure secures all assets, regardless of location, to enable a security approach that extends across the foundation of IT services to protect sensitive data and its legitimate use for healthcare.
• Privacy-Enhancing Computation: Privacy-enhancing computation (PEC) approaches allow data to be shared across ecosystems, creating value while preserving privacy (which is paramount in healthcare). Approaches vary but can include encrypting, splitting or preprocessing sensitive data to allow it to be handled and utilized without compromising confidentiality.
In a world of rapid change, and rapidly changing relationships with information, the healthcare sector must make better decisions, faster. What AI promises is decision intelligence that functionalizes our ever-growing streams of data — and improves organizational decision making by modeling through a sound and adaptable framework.