Unlocking insights and value within health care data through NLP

I recently called natural language processing (NLP) artificial intelligence (AI) with an ROI — and yes, it’s a catchy phrase, but it’s also true. NLP is an essential asset in today’s health care landscape.

Let me tell you a story. Bill and Ted work at the same company – Excellent Adventures, Ltd. – but they live on opposite sides of the city. After feeling ill, they both seek care at hospitals close to their homes – and each discovers he has a chronic condition that requires a minor procedure and a prescription. Their clinical teams take great care of them, and each of them goes home after a short post-op stay. Here’s where their stories split in different directions.

  • A couple of weeks after his procedure, Bill receives a letter from his health plan – his claim is denied. He spends hours on the phone to sort out the problem, adding anxiety about his family’s finances to his recuperation. After nearly two months, a coding error is discovered and the denial is reversed. When he’s asked to respond to a post-care survey by his local hospital, he expresses his frustration by answering with low patient satisfaction scores.
  • A couple of weeks after his procedure, Ted receives a letter from his health plan – his claim is approved. He focuses on recovering and learning about how to prevent his disease from progressing, and when he receives that survey, his scores are positive across the board.

Bill and Ted received the same diagnosis, the same care, and had the same coverage from the same health plan. What was the difference? You guessed it -Ted’s hospital used software with NLP to help assign the right codes to his care. Bill’s hospital didn’t.

Up to 80 percent of the data in health records is unstructured – so if you’re not using NLP, you’re depending on staff to read pages of documentation and then select the appropriate codes from over 155,000 ICD-10 options. Tack on complications and comorbidities, and mistakes are bound to happen. NLP can automate that process, augmenting the skills of highly trained people by scanning health records, suggesting codes and validating coding accuracy.

In Bill’s case, that means his experience could have been more like Ted’s. Bill would have avoided all of that frustration, giving him and his family more peace of mind. The hospital would have been reimbursed more quickly, and would have received higher patient satisfaction scores. The health plan wouldn’t have spent so much time and money researching Bill’s complaint.

This story is just one example of the powerful and meaningful ways AI and NLP are changing how health care is delivered, researched and reimbursed. The ability to interpret complex clinical narratives and rapidly process the results is opening doors that benefit patients and all players in the health system.

When AI is paired with health care intelligence, it can make a world of difference. You can learn more here.

About the Author

Steve Griffiths HeadshotSteve Griffiths, PhD, MS
SVP and Chief Operating Officer, Optum Enterprise

Steve Griffiths has more than 20 years’ experience in health analytics management, and currently heads up the Optum Enterprise Analytics organization. His main focus is to drive growth and innovation through Optum products and services. Steve has a master’s degree in biostatistics from the University of Washington, and a PhD in health services research, policy and administration from the University of Minnesota.

2 thoughts on “Unlocking insights and value within health care data through NLP

  1. With the AI Tsunami coming – mostly led by the new “healthcare” companies such as Google, Amazon, and Apple – I think Optum is one of the few healthcare organizations which can successfully compete with them, and I think that is based on your comfort with data, perhaps due to your cultural roots in insurance through United Health.

    Regarding NLP, although the technical challenges likely center around the data wrangling from the various EHRs, I suspect the much much greater challenge will be ethical – (Google DeepMind’s interaction 2 years ago accessing 5 years of data of 1.5 million patents records without patient consent through the British National Health Service comes to mind).

    These ethical challenges have deep implications for patient’s clinical management, however for such things as billing and sentiment analysis, I believe we are there.

    Here’s one of my favorite applications of NLP sentiment analysis studies coming out of the NIH 2 years ago about obesity, one of our greatest healthcare problems “Obesity in social media: a mixed methods analysis” by Dr Wen-ying Sylvia Chou. It’s a super read – if you like that type of thing! . (As an aside, whenever I think of Optum’s culture I wonder if they are one company best positioned to address the obesity crisis within the country).

    Anyway, my prior organization of 20 years , The Everett Clinic – a supurb multi-speciality clinic of +600 providers, will complement Optum very well in the future ( I am recently retired from clinical practice and getting back to my engineering/software roots – Python here I come! )

    Ref article: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4167901/

Leave a Reply