By Peyman Zand, Vice President, CereCore
Twitter: @CereCore
Healthcare CIOs face increasing pressure to implement advanced analytics at virtually every turn. According to a recent State of the CIO Survey by IDG, twenty-five percent of IT leaders plan expect advanced analytics, artificial intelligence (AI) and machine learning (ML) to drive their technology investments in 2021. As one of healthcare’s most favorite buzzwords, advanced analytics is a nascent technology term with lofty vendor promises but few real-world case success stories.
A recent CIO cohort sought to demystify the practical progress of advance analytics and related disciplines in a virtual roundtable event. The experts asked ‘if advanced analytics had progressed as much as everyone in healthcare assumes?’ Seven health IT leaders suggest the answer is “no”. From the CIO of Geisinger Health System to the chief data evangelist at Looker, part of Google Cloud, this group of health data experts agree—no one is a healthcare advanced analytics, AI or ML Olympian yet.
This article shares the cohort’s insights and real-world advice for healthcare CIOs facing executive pressure to implement advanced analytics solutions amid scant healthcare implementation success.
Take a pragmatic approach with advanced analytics
Perhaps the most glaring example of healthcare’s unreadiness for advanced analytics is the COVID-19 pandemic. Danielle Mintz, Chief Data Evangelist at Looker, reminded the group that it took a “volunteer, bootstrap effort” to become the de facto standard for county-by-county COVID data in the U.S. Collecting data required the COVID-19 Tracking Project’s volunteers to scrape data from county health departments, but it was the most trusted data source available.
Mintz feels that there is still so much groundwork to lay related to healthcare data in the U.S. before advanced analytics, AI or ML can be truly implemented. One approach suggested by Mintz is to bring data into systems, make sure it is clean, and then let clinicians flag any information that’s incorrect or unusable. These pragmatic steps are essential before giving data to clinicians more broadly and to realize the full value of advanced analytics.
Win physician trust
Robin Clarke, M.D., CEO of Ursa Health, also emphasized the role of clinicians in advancing healthcare analytics. Human usability standards must be established alongside clinician practice standards for analytics models and data to be accepted. Physicians want to understand the individual contribution of data elements in the final model’s outcome before they develop trust in the system.
Knowledgeable data analysts working side-by-side with physicians are a second factor in building physician trust according to Mark Lambrecht, Director of Global Health and Life Sciences at SAS. Hospital and health system CIOs that govern innovation, measure performance, explain, and work together with their physicians will be winners in the move to advanced analytics. Lambrecht also suggested the term “augmented” intelligence may also be used by hospital and health system CIOs as a more palatable term for clinicians, versus “artificial” intelligence.
Valmeek Kudesia, M.D., Vice President of Clinical Informatics and Advanced Analytics at Commonwealth Care Alliance, prefers the term “intelligent augmentation”. In his work with Medicare and Medicaid dually eligible members, physicians rely on analytics to quickly determine the proper intervention for each patient. And often these interventions aren’t medical. They may involve purchasing a refrigerator to store Insulin or delivering an air purifier to the home of an asthma patient to avert emergency visits during allergy season.
Kudesia and his team use analytics to insert expert clinical knowledge into the places it is needed to solve all types of healthcare problems. To support this effort, the organization focuses on progressing their existing data while perpetually expanding the types of data they collect.
Be data omnivorous
One example of Commonwealth’s data omnivorous approach is the incorporation of weather data to notify asthma patients of upcoming concerns and deliver preventative supplies as described in the case example above. Social determinants and spending habits were other data elements mentioned by the group. John Kravitz, CIO of Geisinger Health System, verified the importance of expanding data collection to intercede in patient care.
His organization collects genomic research data to resolve patient issues before they present themselves as healthcare crisis. Kravitz mentioned the ability to identify and surgically stop a growing aneurysm as just one example. Geisinger has sequenced over 200 thousand data points so far. However, Kravitz admits “it is still an evolving practice.”
No one in the CIO cohort—or the healthcare industry—has a magic bullet. However general consensus among the group was that data is the essential first step. While most human time in healthcare IT is spent acquiring data, most of data’s value is derived from building relevant queries, creating models, and interpreting the data to drive predictive practices. Data is the linchpin for moving forward with advanced analytics in healthcare.
Implement best practices for data
All participants provided recommendations and best practices for data extraction, transformation and load (ETL) as precursors to success with advanced analytics, AI or ML in healthcare. Here are a few of their suggestions.
Re-use data assets. Invest in the re-use of a flexible data model for multiple initiatives.
Take advantage of advanced data warehouse technology and horsepower. Warehouses can now process more data and accept different data varieties and shapes. This gives users the ability to pump more raw data into the warehouse and transform it afterwards.
Move from ETL to ELT. Consider cleaning data after it is uploaded to the warehouse. This drives more value from the data, and faster. With ELT, data is extracted, loaded and then transformed or cleaned.
Don’t try to boil the ocean. Pick one problem, solve it and put the solution in front of users to test it.
Consider hiring a Chief Data Officer. The new superhero in healthcare will be the CDO. This position plays an instrumental role in ETL, expansion of data sources, and reducing cycle time for process improvements based on data analytics.
Conclusion
Advanced analytics will be used to inform healthcare process improvements and support greater personalization of healthcare services. The first step must be to gather and use data more efficiently to impact patient care.