AI & Machine Learning
“The health-bots will see you now.”
A world with AI assisted healthcare is not science-fiction. Today, machine learning and artificial intelligence are making progress on:
- Automatically processing images, clinical notes and patient reported outcomes
- Creating precision diagnoses, treatment plans and interventions
- Computing risk scores based on hundreds of variables
When we think about a healthcare AI, most people imagine a singular artificial entity like Data from Star Trek or HAL from 2001: A Space Odyssey. The fourth industrial revolution for healthcare is more likely to meet you in small interactions at the grocery store, at the gym, through your smart phone, or in behind-the-scenes ways that you never even see, rather than through a bedside android or in a doctor’s office.
At the moment, the practical applications of machine learning and artificial intelligence are best suited to niches and pockets of decision-making that professionals encounter over and over again throughout their career. Successful strategies for making an impact concentrate on:
- Small areas of opportunity with a highly focused objective
- Applications that are well-suited to the strengths of the technology
- Achievable goals
Think Big to Start Small
Incremental changes and quick wins are still the name of the game, but the capabilities that the latest industrial revolution enables mean those small changes can have a much bigger impact. Instead of automating a static risk-scoring formula to rank ER patients upon arrival, use a learning algorithm that improves with feedback and provides benefit to an increasingly greater number of patients rather than the same “top 10%” of admissions. The benefits will accrue exponentially over time as your scoring algorithm improves with new experiences.
In order to find those quick wins, you should be thinking about what your biggest problems are in terms of scale. Where do you have the greatest number of patients or the greatest number of procedures, or the greatest variation in outcomes? Where the magnitude of the challenges is large, there will be great opportunities to automate a piece of the decision-making process and reap large scale benefits. These places with high volumes of interaction and large amounts data are where machine learning and artificial intelligence thrive, but you have to give the algorithms a focused outcome or objective to reap the biggest benefits.
Practical Priorities
Where machine learning shines today are in applications such as natural language processing, comparative classification or scoring of individuals, and identifying subtle patterns that we may see as “something I can’t quite put my finger on.” You may have other big challenges to address, but starting with something that machine learning and artificial intelligence are known to do well will have the greatest likelihood for success.
Begin by decomposing your big challenges into the individual activities and thought processes that either comprise or support the larger whole. Within those components, look for those that are data intensive, time consuming, internally repetitive or high-volume. These will be places where machine learning and artificial intelligence have the best chance of making an impact.
Knowing the Limits
Don’t set your zoom-level too far out on your hopes and dreams for machine learning and artificial intelligence. Neither of these technologies are magic wands that can instantly create actionable insights or automate expert clinical judgement in a snap. Decompose bigger challenges into their components and use proven techniques to address targeted needs.
AI with Amitech
As artificial intelligence and machine learning technology continue to unfold across the healthcare landscape, the potential applications and impact will broaden. But if you’re looking to capitalize on today’s possibilities, here’s a few practical examples and areas of opportunity to mull:
- A virtual health coach that sets activity goals on a daily basis, learning from your past behavior, your calendar, and verbal feedback you provide in response to prompts
- Diagnostic support for physicians and specialists reviewing radiology reports or lab tests, helping them see things they might otherwise miss and reduce the time they spend reviewing results and notes
- Scaling our healthcare delivery model to touch more lives, by automating routine tasks like chart summarization and abstraction to eliminate mundane and repetitive tasks
- Deploying instant care to previously inaccessible care settings through telemedicine and smart adaptive devices
- Supporting follow-through on treatment recommendations or interventions through continuous remote monitoring and virtual nudges, both human and automated
Could AI and machine learning fit into your current strategy? Call us and find out.