Trend Topic: Mining Electronic Health Records

By Lillian L. Dittrick, FSA, MAAA

Hospitals and health systems are increasingly designed to excel at the Triple Aim of providing patients with quality care and optimal experience at a lower cost.  The challenge is achieving financial viability in a quickly changing environment that is shifting from fee- for-service to value-based care.

Providers must realize the value of the services they deliver. Part of this realization comes from accurate coding and documentation to provide appropriate care management and to collect reimbursement for services rendered. Providers can only get paid for what they correctly code.

A loss of revenue due to missed codes, coding errors and procedural issues is a major concern for every health care environment. One solution that can aid in capturing clinical data input and translating it into useful outputs is natural language processing (NLP).  NLP can also enable the many functions of a hospital or health system to improve the collective knowledge of health care providers based on their shared insights.

Most U.S. hospitals have adopted Electronic Health Record (EHR) systems, according to a Federal government report released in late May 2016. While that is welcome news, the increased use of EHRs also means millions of unstructured records that are rich in medical information may sit untapped.

The corporate analytics department at UnityPoint Health® is taking that massive amount of data and will be using it to improve patient outcomes.

UnityPoint Health estimates that up to 40 percent of its diagnosis codes are not recorded in structured fields.  This under coding impacts both timely disease identification and billing. UnityPoint is using NLP to turn this unstructured data into structured data.

The NLP can scan doctors’ notes for valuable information, such as family history and ailments to help predict patients’ medical needs.  It can also help identify chronic conditions that have not been recorded in structured fields in electronic medical records.

This enriched information can be used to complete patient risk stratification models, including risk scores and to analyze missed coding opportunities.

The NLP annotators have ‘cognitive’ abilities similar to human coders. UnityPoint is using the annotators to analyze unstructured data for diabetes and chronic obstructive pulmonary disease. The next analysis will be around social determinants of health, such as living arrangements and employment status, to help with care management initiatives.

Some of the other UnityPoint Health modeling using EHR data includes a clinic appointment no-show model and a staffing model to forecast patient demand and guide staffing levels.

While nothing in health care can be predicted with certainty, actuaries use predictive analytics to identify new models of care that achieve the Triple Aim of improving health care quality, costs and outcomes for our patients.

Sharing is caring.
  • Subscribe to our feed
  • Tweet about this post
  • Share this post on Facebook
  • Share this post on Google
  • Share this post on LinkedIn


No responses to "Trend Topic: Mining Electronic Health Records"

There are no comments yet, add one below.

Leave a Comment