From the March 2017 issue of HealthCare Business News magazine
By Simon Beaulah
According to the Centers for Disease Control and Prevention, about 5.7 million Americans suffer from congestive heart failure (CHF), which contributes to about 287,000 deaths per year. For individuals over the age of 65, heart failure is the leading cause of hospitalization and accounts for more hospitalizations than all forms of cancer combined. CHF-related costs, including health care services, medications to treat heart failure and missed days of work total an estimated $30.7 billion each year.
CHF touches a large segment of the population and carries a significant financial impact. Health care stakeholders, including health systems, accountable care organizations and payers recognize that the proper identification of CHF patients is critical in order to effectively manage the health of these individuals and to ensure proper provider compensation.
When a CHF diagnosis is overlooked, ongoing follow-up care may fail to address proper treatment regimens, leading to additional complications that require hospitalization. To ensure the accurate identification of all CHF patients, various health care stakeholders are turning to advanced technologies such as natural language processing (NLP) to screen clinical documentation for missed diagnoses and risk factors to improve care and outcomes.
A large physician group participating in a Pioneer ACO program and at-risk commercial contracts was required to track and report clinical quality measures, including ejection fractions, for all heart failure patients. Like many health care providers, much of this detailed clinical data was stored as unstructured free text, rather than within the electronic heath records’ (EHRs’) structured data fields. Because the information was stored as unstructured text, patients having ejection fraction values of less than 40 percent were not always logged as having a CHF diagnosis in the problem lists.
To properly report and store ejection fraction details for ACO reporting — and to ensure that revenue-impacting issues were not overlooked — the provider dedicated resources to manually reviewing patient charts and reconciling patient problem lists. Manual chart reviews were time-consuming, leading the organization to implement an NLP solution to automate the identification of ejection fraction information from within free-text fields.
By replacing manual chart reviews with NLP tools, the provider realized a fivefold increase in efficiency in the chart review process, and clinicians and patients are benefiting from more accurate and complete problem lists. Pulmonary function is similarly extracted for COPD populations delivering a 12-fold improvement in efficiency. Furthermore, the ACO reporting process is simplified and billing is more precise and complete. CHF patients often suffer from comorbidities, such as diabetes, chronic pain, arrhythmia and depression. These conditions can complicate the management of care and, when combined with the impact of social determinant factors, patient ambulatory status and living location, make for a complex set of factors in predicting unscheduled hospital readmissions.
