(US and Canada) Christer Johnson, Chief Analytics Officer at Healthfirst, speaks with Jake Dreier, Director, Strategy & Growth, HiLabs, about the impact of AI and ML in health care, managing directory data, and the issue of having a single source of truth for data. 

Johnson states other industries use AI and ML to drive sales and understand purchasing behavior compared to demographic profiles. However, the health care industry uses AI and ML to change members’ and patients’ behaviors to improve their health.

AI and ML pinpoint the critical insights needed to improve the population’s health care at a lower cost. They optimize timing, content, and delivery channel to influence the information patients receive, promoting positive behavior change. Johnson states, for example, that numerous steps are taken to analyze health care measures for the diabetic population. Without AI and ML, it’s impossible to carry out the analysis at scale.

Johnson notes that Healthfirst and many other companies are shifting their attention toward Natural Language Processing because it involves unstructured data. It’s hard to access unstructured data while building analytical models. Hence, the companies focus on leveraging NLP to drive insights out of clinical charts in a scaled way.

AI and ML are equally fundamental to the video analytics domain, says Johnson. He says that AI and ML successfully identify certain conditions on scans in the radiology space, thus deriving hidden insights.

Healthfirst is a nonprofit health care organization serving 1.8 million members. Johnson states that 70% of the business is founded on value-based care arrangements. This increases the company’s need to responsibly leverage the data of a million people while constantly analyzing member interactions and providing insights to generate better health care.

Johnson highlights the time sequence/customer journey concept, explaining that all data is merged to form a sequence and define patterns. These include claims, call center, and click stream data from mobile interactions. Johnson again affirms that analysis of this scale is impossible without implementing AI and ML. 

Regarding directory data, John suggests spending time on audits to understand and improve provider data management quality and learn how the data can be accessed. Johnson maintains that they are expanding directory data by including the notion of health care access. He states that health equity is critical to Healthfirst because they primarily serve Medicare members.

Johnson further adds that blockchain is the apt technological solution for the problem of provider data management. He voices the need for a provider data management/blockchain platform that is secure and which all the health care stakeholders can visit for the correct information.

He explains how doctors, payers, and members could use blockchain to fill in the information at once. The problem lies in creating the blockchain, says Johnson. He believes that the industry should come together for overall industrial benefits.

In the context of technologies such as blockchain, AI and ML, Johnson discusses the notion of a single source of truth for data. Philosophically, he believes that any data is collected through a source system.

Johnson explains that a single source of truth is where the data is collected. The farther from the source system, the less data can be trusted. He notes that while collecting data from the source system, there can be issues with the completeness of data. He considers this a data collection and process problem rather than a data management problem. 

Johnson believes that improving data collection is one element of a single source of truth, whereas moving away from source systems will lead to issues. He states that results may differ based on the various databases used within an environment. 

He appreciates the use of data fabric and how companies like Snowflake have made it easier by concentrating everything in a place. Creating such data environments by lowering degrees of separation from the source system used for analysis helps derive better organizational values, states Johnson.

Johnson concludes that organizations need to start quantifying data and explicitly eliminate redundancy across data infrastructure. He reflects that having all the data analysts and scientists pointing at the source system will create “slowness problems” and other issues. Hence, thinking through to reduce degrees of separation from the source system in a systematic way is crucial.