How Electronic Health Records Are Standing in the Way of AI
At Rock West Solutions in Southern California, a dedicated team of very smart people is working on ways to utilize artificial intelligence (AI) and data analysis for improving healthcare outcomes. They envision a day in which AI and big data are effective tools for predictive analysis, making it possible to provide healthcare services that actually prevent sickness and disease.
How close is Rock West Solutions and its partners? Not as close as they would like to be. In fact, the entire industry is lagging behind its known potential because of one big thing standing in its way: electronic health records (EHRs). An excellent article from Axios contributor Kaveh Waddell illustrates the point well enough.
According to Waddell, AI in the healthcare setting has a big data problem. That problem lies at the feet of EHR systems that are not only failing at what they are designed to do but are also creating a steady stream of errors that are leading to significant medical mistakes.
A Good Idea in Theory
A decade ago, former President Barack Obama signed into law legislation that required the U.S. medical industry to adopt EHRs. The idea was to create a cohesive system that maintains accurate medical records accessible by any facility in the country.
According to a KHN report, the government has already put some $36 billion into developing EHR systems. The IT industry earns billions every year building the network and software solutions intended to satisfy federal mandates. But none of it works as it’s supposed to.
The biggest problem is that EHR systems are essentially a patchwork of individual solutions that don’t play well together. It’s like having 10 doctors trying to agree on a diagnosis when none of them speak the same language. They can jabber at one another all day long and never reach a consensus.
Systems Are Error-Prone
Waddell maintains in his article that even the best EHR systems are error prone. There are lots of reasons for this. First, doctors often use generic insurance codes supplemented with personal notes because the current coding system is so convoluted. This leads to a lack of specificity.
Next, health records make use of a lot of copied data as they progress through the system. According to one 2017 study published by the JAMA Internal Medicine journal, as much is 82% of the data found in the typical patient’s electronic records has been either imported or copied from another record.
Finally, it has been estimated that the useful life of a given medical record is only about four months due to changes in patient health, the evolution of treatment modalities, and so forth. And yet doctors and hospitals are relying on records that could be years old.
A Polluted Stream of Information
According to Waddell, the inability of EHR systems to produce consistent, reliable information has created a polluted stream that cannot be trusted. Once a mistake is made in one health record it populates into the next, and then the next, and so on. This leads to additional errors that are only compounded over time.
The summary of all this is to say that AI holds a lot of potential for predictive analysis. Combined with big data, AI could truly revolutionize healthcare. But that will never happen until the EHR mess gets straightened out. And knowing the history of the healthcare industry’s adoption of technology, it’s quite possible it will take another decade to fix EHRs. In the meantime, how much of the potential AI holds will have been wasted?