Why data integrity is key to achieving value in healthcare
Ankur Teredesai says the entire premise of value in healthcare is based on the ability to measure performance metrics.
Value is extremely hard to define when it comes to healthcare, according to University of Washington computer science professor Ankur Teredesai, co-founder and chief technology officer of Seattle-based KenSci, a company that provides a risk prediction platform powered by artificial intelligence and machine learning.
Teredesai, who co-chaired this year’s Association for Computing Machinery's KDD Conference, held in August in Anchorage, says the entire premise of value in healthcare is based on the ability to measure performance metrics while simultaneously establishing baselines for reducing unwarranted variation.
“Data is central to cost prediction and estimating unwanted variation,” he adds, noting that eventually, providers will use data and AI-driven decision-making for optimizing schedules and assessing patient risk.
At the KDD Conference, Teredesai encouraged healthcare communities to think beyond the size of data and focus on the complexity and integrity of data sources. “Advances along these lines will help bend the utilization curve much before we see doctors being replaced by AI,” he says.
How are some payers taking advantage of data and how could some do better?
Teredesai: Payers care about various use cases that ensure they have appropriate tools for keeping members healthy, managing campaigns against rising risks, such as prevalence of diabetes and keeping acute care utilization in check. Data helps them identify cohorts that may be high risk or rising risk members. Data even helps facilitate risk adjustment and improve payments to reflect the state of the population under the various plans offered by payers.
There is much payers can do to improve operations with the value of their data assets. First and foremost, most large payers need to focus on Identifying high-risk utilizers who may be at risk of increased complications, or other cost-related use cases.
Where does machine learning (ML) come in, when it comes to population health management. How important is it?
Teredesai: ML versus actuarial sciences is going to become a heated space of disruption in the next few years. With increasing ability to create dynamic cohorts and predict risk, ML is very important.
What are the most important trends in machine learning that could benefit health plans in their quest to lower costs, yet improve the quality of care?
Teredesai: It would be good for the Centers for Medicare & Medicaid Services (CMS) and private health plans to collaborate with leading healthcare machine learning players. At Kensci, we’ve been working with quite a few payers and it’s been fascinating to explore advanced ML techniques like explainable AI algorithms at both population and individual beneficiary levels to understand factors that influence utilization. Another important trend is to understand the pros and cons of data imputation. While everyone talks about deep learning as a trend, I think time- sensitive neural learning approaches have promise in this space for care pathways.
How do you see machine learning and data analytics changing the face of healthcare over the next five or 10 years?
Teredesai: While most progress will be incremental and we are not close to automated plan recommendations or interventions by robots anytime soon, I place much hope in the power of what I call “assistive intelligence.”
Teredesai, who co-chaired this year’s Association for Computing Machinery's KDD Conference, held in August in Anchorage, says the entire premise of value in healthcare is based on the ability to measure performance metrics while simultaneously establishing baselines for reducing unwarranted variation.
“Data is central to cost prediction and estimating unwanted variation,” he adds, noting that eventually, providers will use data and AI-driven decision-making for optimizing schedules and assessing patient risk.
At the KDD Conference, Teredesai encouraged healthcare communities to think beyond the size of data and focus on the complexity and integrity of data sources. “Advances along these lines will help bend the utilization curve much before we see doctors being replaced by AI,” he says.
How are some payers taking advantage of data and how could some do better?
Teredesai: Payers care about various use cases that ensure they have appropriate tools for keeping members healthy, managing campaigns against rising risks, such as prevalence of diabetes and keeping acute care utilization in check. Data helps them identify cohorts that may be high risk or rising risk members. Data even helps facilitate risk adjustment and improve payments to reflect the state of the population under the various plans offered by payers.
There is much payers can do to improve operations with the value of their data assets. First and foremost, most large payers need to focus on Identifying high-risk utilizers who may be at risk of increased complications, or other cost-related use cases.
Where does machine learning (ML) come in, when it comes to population health management. How important is it?
Teredesai: ML versus actuarial sciences is going to become a heated space of disruption in the next few years. With increasing ability to create dynamic cohorts and predict risk, ML is very important.
What are the most important trends in machine learning that could benefit health plans in their quest to lower costs, yet improve the quality of care?
Teredesai: It would be good for the Centers for Medicare & Medicaid Services (CMS) and private health plans to collaborate with leading healthcare machine learning players. At Kensci, we’ve been working with quite a few payers and it’s been fascinating to explore advanced ML techniques like explainable AI algorithms at both population and individual beneficiary levels to understand factors that influence utilization. Another important trend is to understand the pros and cons of data imputation. While everyone talks about deep learning as a trend, I think time- sensitive neural learning approaches have promise in this space for care pathways.
How do you see machine learning and data analytics changing the face of healthcare over the next five or 10 years?
Teredesai: While most progress will be incremental and we are not close to automated plan recommendations or interventions by robots anytime soon, I place much hope in the power of what I call “assistive intelligence.”
More for you
Loading data for hdm_tax_topic #better-outcomes...