Artificial Intelligence (AI) and Machine Learning (ML) continue to demonstrate impressive results in improving quality and operational efficiency in Healthcare and Life Sciences.
But generating real, lasting value requires more than just the ‘best’ AI/ML models. And with the somewhat unchecked proliferation of this technology, it all starts with responsible development and focusing on solving the right problems. Only then will your models shine. So yes, wish responsibly!
As we wrap up another successful year for Equilibrium Point and for our customers, our team, as always, has spent time researching key technology and market trends that will continue to push innovation across the industry.
This is the time to finalize and vet goals for 2023, with accompanying plans to implement and optimize over the next 3-6 months. To that end, we identified three areas of consensus across our customers in the Healthcare and Life Science space, that should be key strategies for 2023:
🧞♂️ Build an end-to-end data strategy leveraging the variety of data available
Across Healthcare and Life Sciences, from large health plans to small research organizations, the challenge is similar: gleaning insights to take real-time actions from multiple data sources (usually from different vendors, locations, legacy systems and new data lakes on the cloud) and with a variety of access points from APIs to dashboards and mobile applications.
Today, we have more and more data sources at our disposal with increasing volumes, as never before:
EHR solutions
Time series data from many IoT health monitor devices
Video/Images from diagnostic imaging tests
Scanned documents and other paper-based health records
Application Logs
Patient self-reported data
3rd Party Data (consumer data, demographics, geospatial)
But with more data comes more challenges. Data are typically not curated, are poorly cataloged and often stored in silos, and thus not ready to produce real-time insights, much less act upon them with high confidence and urgency. It’s the old, harsh (but true) saying: garbage in, garbage out. Only now, its automated.
How then to improve this all-too-common situation? ⬇️
There are so many vendors, tools, and even many data operations platforms with literally thousands of services. So, where to start? Enlist and recognize the people in your organization that manage your different data assets and that have the desire to address the problem or opportunity at hand (hence, the importance of focus). We should recognize them as data stewards within each team (research, product, development, operations, IT teams).
Then, find the right partner with a proven track record for technology innovation in the industry and join these two groups in collaboration, with an eye towards upskilling your internal teams.
The importance of an end-to-end strategy and platform cannot be overstated, as this is the only way to monitor and improve ML performance. Otherwise, bias, gaps, performance “drift” and other issues will automate a problem, not a solution. 🧞♂️ Standardization of infrastructure and tools
The reach of AI/ML continues to grow. Global spending on AI/ML technologies will reach $204 billion by 2025. And, 57% of executives say AI/ML would transform every part of their organization within the next three years.
More healthcare and life science organizations are creating AI ecosystems such as AI in clinical trials for intelligent patient enrollment improving enrollment and randomization rates and reducing screen failures and dropouts.
For example, Tufts Medicine has built a digital health ecosystem that dramatically enhance the value of the EHR patient information for better patient care.
How to get there? ⬇️
Start with one use case for ML and experience your first win. As you develop your first use case, start building a first iteration of your end-to-end data strategy. If you have already successfully delivered on a few use cases, congrats! Then move on to our third key strategy:
🧞♂️ Automating use cases with embedded ML models
Using ML for making decisions at scale is not a one-person task. It requires diversity of skills on your team and a process to work effectively together. It involves applied data science/ML, however, it also involves a deep understanding of your business domain including technical, operational, social, managerial and financial drivers.
What are the challenges? ⬇️
First challenge: we work frequently with our customers and prospects on how to avoid using ML to solve the wrong problem. We circle back to wish responsibly! This points to the field of Decision Intelligence (applied ML, statistics and analytics to solve business problems). Decision Intelligence (DI) provides a framework for best practices in decision-making and processes for applying machine learning at scale.
Second challenge: designing your business performance metrics upfront. Once you have defined the problem(s) to be solved, develop hard-core performance metrics that have specific data sources and exact formulas that have been vetted by the analytics team. Forget trying to incorporate trendy, ambiguous KPI buzzwords. Always evaluate performance based on your business metrics.
We are very excited for what the future holds. With the rapid changes in technology, industries, and societal patterns and processes, due to expanding global interconnectivity and smart automation, Equilibrium Point is well positioned to leverage our people and technology to solve increasingly complex business problems, and contribute to the success and well-being of both our and customers and team.
So, tell us, what are your focus areas for 2023? We would love to hear from you and include some of your insights into our roadmap. Until then, thank you for your continued support and partnership.
Wishing you and yours a very happy, healthy, and prosperous New Year!
The Equilibrium Point Team