Aligned with the complex language of healthcare, Clinical Natural Language Processing (cNLP) enables healthcare organizations to unlock the value of their data.
An important goal of cNLP is to provide structure to highly unstructured data sources. In this second installment of our four-part series on cNLP, we uncover the opportunities and challenges providers and payers face when extracting data from free text for quality measures reporting. If you missed our first blog related to risk adjustment, we invite you to read it here.
Whether a provider, payer, or patient, all benefit from clinical records that are thorough, accurate, understandable, and interoperable. Having quality data available is foundational for providing quality care to patients and ultimately improving outcomes. However, the vast amount of data paired with the varied types of data available in electronic health records creates challenges when attempting to review information meaningfully, extract it from the records, and share it effectively.
An essential component of quality clinical documentation is accurately depicting the patient story. To accomplish this, clinicians often combine the art of storytelling with the science of medicine to create their notes. Many different types of documents can be found in patient records and the content within varies by document type. To construct clinical notes, a variety of documentation methodologies are used to input data such as point and click entry of data elements from templates, narrative free-text type, speech recognition, and dictation. Often, more than one of these techniques are utilized. Notes are typically then augmented by incorporating free form data entry that provides detail to the record that may not be available in the EHR templates, or not easily documented using a templated format. Common areas of the chart where unstructured data can be found are the history of present illness, assessment and plan, progress notes, procedure notes, and entries from comment fields scattered throughout the electronic health record.
Data review and extraction
The information contained in patient records is complex. In healthcare, we have our own unique vocabulary. There is a large lexicon of medical terms and there are many ways we can describe a medical concept. While training providers to use speech recognition software to augment notes, I quickly realized the importance of using a program that included a medical dictionary. While there are many benefits to having multiple ways to input data, including speech to text and typing, we are often left with a myriad of medical terms, synonyms, acronyms, approved or unapproved abbreviations, and misspelled words present in the clinical notes. Patient records include a mix of structured and unstructured data which leads to a more complete patient story, but at the same time presents us challenges in the review of records, and the extraction of meaningful information for patient care, analytics, research, and reporting.
I would be remiss if I did not mention the impact the pandemic is having on our healthcare system due to COVID-19. The plethora of cases and the high acuity of many of the patients has significantly changed the ways our colleagues are caring for patients and organizations are operating. Rapid changes to workflows, equipment procurement, reporting, and policies and procedures are impacting so many areas, including the content of patient records and the urgent need for sharing this important information. Here are some examples of the many challenges our colleagues are experiencing today.
- Rapid and iterative creation and revision of care guidelines, decision support rules, triage procedures and protocols due to the continually changing COVID-19 symptom list and knowledge of the virus and disease.
- Care providers working outside their specialty areas. For example, OB working in ICU, informatics nurses working at the bedside. The staff may be unfamiliar with the patient population and equipment, e.g., patient isolation, PPE, ventilators, patients with acute respiratory distress, and what and how to properly document the care provided.
- Fast on-boarding of staff from different hospitals, new graduates, or those coming out of retirement.
- Swift uptake in telehealth and caring for patients in alternative care sites, e.g., drive-through testing. There are reports of varying availability of technology in these areas. Manual flowsheets and other documents need to be scanned into the record.
- Surge response and care of high acuity patients.
- Reporting requirements to public health departments, Centers for Disease Control and Prevention (CDC), and registries.
- New COVID-19 related standardized codes such as, ICD-10, CPT®, SNOMED-CT® and LOINC®.
Narrative notes likely abound in many systems. cNLP is a solution that can help organizations extract information needed for COVID-19 analytics and reporting. It can also be used at the point of care to identify patients with signs and symptoms that then alert the care team to patient decline or decompensation for rapid care intervention.
Quality improvement and measures
Payers, providers, vendors, and patients are all key stakeholders in the quality process. Quality information from notes that are complete and accurate can be parsed and extracted to help identify opportunities for improvement, analyze performance, and report out to external agencies and registries. Subpar data can lead to risks such as missed identification of care gaps, patient safety issues, a perception of poor patient outcomes, and decreased reimbursement affecting the bottom line.
With the continued growth of our Medicare population and transition to value-based care, quality is at the forefront of healthcare. Finding ways to manage structured and unstructured data effectively and more efficiently is critical. Having the ability to automate the extraction of key clinical data and normalize to standard terminologies such as SNOMED CT can help organizations better understand where they may be falling short on meeting measure thresholds. Clinical NLP can help organizations accomplish these goals by reducing the ambiguity in the notes by analyzing context, leveraging clinical synonyms, and normalizing the information.
Below are two examples where the Heath Language cNLP solution is helping organizations deliver on HEDIS and CMS quality measures focused on colorectal cancer prevention:
Colonoscopy Cancer Screening (COL) – this HEDIS measure lists several screening procedures and tests in the inclusion criteria along with associated time frames for each such as, colonoscopy in the past 10 years or Fecal Occult Blood Test (FOBT) annually. Also included in the specifications are exclusions to the measure such as, total colectomy and frailty and advanced illness. Data for these inclusion and exclusion criteria may be in current or past notes, may live in structured or unstructured data, and will need to be extracted and normalized for analysis and reporting.
Comprehensive Diabetes Care (CDC) – this measure focuses on assessing adults age 18-75 with Type 1 and Type 2 Diabetes. It measures Hemoglobin A1c testing and in certain populations identifies the level of control. Also included are eye exams, blood pressure control and identifies patients seen for nephropathy. Leveraging the value of cNLP, organizations can quickly and effectively extract data from multiple records, normalize or codify these data to standard terminologies, and then use the output for analytics and reporting.
Beyond these two examples, cNLP can be leveraged to optimize workflows and increase data accuracy that can empower a variety of high value use cases. Speak to an expert to learn more about how the Health Language cNLP Solution can help your organization leverage unstructured data.
Be sure to keep your eye out for our next blog in the cNLP series, where we will dive into how cNLP can be used to improve the efficiency and accuracy of medical necessity review.
CPT® is a registered trademark of the American Medical Association (AMA).
SNOMED CT® is a registered trademark of the International Health Terminology Standards Development Organisation (IHTSDO).
LOINC® is a registered trademark of Regenstrief Institute, Inc.