By Abboud Chaballout, co-founder, Diagnoss.
Over the past several years, we’ve heard a lot of predictions about new and innovative ways artificial intelligence (AI) will dramatically impact healthcare. For example: robot doctors, drug discovery and clinical diagnosis. While there is – and should be – excitement around what innovation may come, we need to look at what technologies exist today, and more importantly, what providers need to deliver higher quality and more coordinated care at a lower cost.
In 2020, we’ve watched a pandemic upend the healthcare industry as ICUs became overwhelmed, clinics had to close their doors and patients avoided care. Physician burnout that was 42% in 2018 is soaring as COVID-19 cases surge and disruptions continue.
With this backdrop, it’s time for us all to agree that the biggest need in healthcare right now is to help providers do their job more effectively and remove burdens that both stand in the way of the delivery of care and lead to that unacceptable rate of burnout.
In 2019, the U.S. is predicted to have spent more than $3 trillion on healthcare, over 95% of which was dispersed through an insurance company, where every dollar must be codified. The current international standard for those codes is ICD-10, which contains more than 70,000 codes for diagnoses. With that many codes, it comes as no shock that estimates for annual medical coding errors are around 30%, with billing errors reaching as high as 80% in many cases.
With such a complex coding system, we are adding to a provider’s administrative burden which, for many, is already at a breaking point. Today, a physician spends an average of 16 minutes on administration, which adds up to several hours every single day.
To date, AI has not been used to meaningfully improve the coding process; the focus has been on merely automating it in areas where human intervention can be eliminated. We need to leverage technology to help doctors and coders on the front lines of the coding process.
Imagine an expert medical coder with more than 30 years of experience standing over the shoulder of a doctor or a junior coder, reading patient notes and whispering the correct codes into their ears, so they don’t have to expend any more energy or time. This AI assistant would not only help reduce the number of clicks required throughout the code selection process, but it would also improve the efficacy and accuracy of the code selections. The medical coding process should be that simple for every single provider, coder, and biller.
In a study we recently conducted on more than 39,000 de-identified EHR charts, we found that when compared to human coders, the machine reduced error rates by 50% for commonly used codes.
Population health management
Today, population health – defined by the CDC as the approach that allows health departments to connect practice to policy for change to happen locally – mostly remains elusive. While some clinics, systems and communities are further along than others, our healthcare system is still far from where it needs to be when it comes to connecting all of these important pieces.
According to TechTalks founder Ben Dickson, the complex and dynamic nature of population health management makes it especially convenient to handle with machine learning, the branch of AI that specializes in uncovering hidden patterns and predictive trends in large and disparate sources of data.
AI algorithms can ingest data from sources such as health records, financial data, patient-generated data, and IoT devices to automatically discover groups of patients that share unique combinations of characteristics. Patterns gleaned from this data can help health institutions engage in preventive and predictive care, which can result in great savings in managing and treating diseases that become costly and complex when they’re discovered at later stages.
Today, providers can find themselves stuck down time-sucking rabbit holes as they look for information that connects the dots between all of these elements that make up a patient’s health and health journey.
We have already seen improvements in some areas of patient care, such as screening efficiency and accuracy (AI has made mammograms faster and more accurate). In addition, as Healthcare Weekly points out, AI and analytics programs can work together to digitally represent human physiology, showing the projected evolution of a disease, allowing doctors to tailor treatment plans based on actual projected outcomes.
Imagine how much more time a physician could spend with a patient, discussing a diagnosis or how they’re actually feeling if some of these things were automated. Unfortunately, most physicians are spending more time on the admin – medical code entries, information research – than on patient care, which is one of the biggest reasons we’re seeing such high levels of burnout. We have the technology today in AI to begin addressing some of these burdens. It’s important that we start with the things that will make a real difference.