By Steven Lang, Senior Vice President of Sales, SyTrue, shortlisted for Best SaaS Product for Healthcare at 2022 SaaS Awards
The healthcare system is built on a foundation of massive, disorganized, messy “photo albums” of data that depict patients’ stories throughout their medical histories. Too often, these stories become garbled and confused, preventing an accurate retelling, because patient data is so often fragmented and incomplete.
Specifically, the majority of healthcare data is stored as huge collections of unstructured information within medical records, diagnostic images, lab reports, and pathology notes. Almost 80% of healthcare information is unstructured, making it difficult for providers and researchers to access. Because patient records are often scanned and faxed, healthcare organizations are left with billions of paper files, PDFs, and images that most intelligent healthcare information systems cannot make sense of or access in any meaningful way.
To overcome the challenges of dirty healthcare data, many payer and provider organizations are looking to Natural Language Processing (NLP) solutions built on Software-as-a-Service (SaaS) architecture that can read, extract and produce clean, accurate data for a wide variety of use cases, such as payment integrity, risk adjustment, and population analytics.
What is NLP?
As healthcare organizations seek to improve finances, manage costs and increase operational efficiencies, they’re realizing the impact of unstructured data and the potential that NLP solutions offer. For example, the global market for NLP in healthcare and life sciences was valued approximately at $1.5 billion in 2020 and is anticipated to increase at a compound annual growth rate of 19% between 2021 and 2027, according to Market Insight Reports.
By giving computers the ability to read, understand, and interpret clinical language, NLP can extract and organize data from an individual’s episodic health record. By strategically presenting these data points, organizations can modernize the chart review process and eliminate antiquated and bloated workflows associated with manually reading a medical record. NLP organizes disparate, unstructured data into a comprehensive, longitudinal, and semantically interoperable patient record that captures all key health information, such as treatments, diagnoses, and care plans.
Additionally, NLP gives organizations the power to retrospectively analyze longitudinal health data to find just one particular piece of clinical information about a single patient. It also supports stratification of populations that require further exploration, providing significant time savings from removing the need for clinicians to read through pages and pages of medical documents.
A significant advantage of NLP is that chart reviews can be performed just one time to extract all key patient data, as opposed to multiple reviews by multiple clinicians, researchers, or auditors who are all looking for different nuggets of information. Further, NLP technology is a key component of interoperability between healthcare information systems, structuring and standardizing health information from a wide variety of sources including claims and patient charts.
Solving real-world issues with NLP and SaaS platforms
Leading NLP solutions are often built on SaaS platforms to deliver more transparency, scalability, security, and flexibility.
Below are some of the critical real-world challenges of analyzing health documents, the benefits of provisioning a solution across a SaaS delivery model, and the benefits of NLP.
- Highly irregular document format issues: NLP solutions leverage machine learning to ensure capture of the essence of document styles and formats, but do not require absolute adherence given the variable nature of documentation. Good NLP can differentiate the History of Present Illness (HPI) and Review of Systems (ROS) from a consult note and apply proper rules to map terminologies appropriate to each.
- Endless synonyms and abbreviations: Leading NLP solutions use models to predict the document types being processed to allow sorting into specific groups and manage the lexical uniqueness by specialty or healthcare domain.
- The massive scale problem of ingesting hundreds of million files: As cloud-based platforms that are architected for massive scale, leading NLP solutions anticipate and stay ahead of queues and surge swings in processing volume.
- Go-Live, Customizations and Updates: Natural Language Processing Engines truly encapsulate everything that a SaaS delivery model has to offer. From multi-tenant architecture for managing organizational ontology preferences, to user management, access privileges and seamless updates, delivering NLP solutions in a SaaS framework allows for transparency, flexibility and efficiency of onboarding.
Use cases: What NLP can do for healthcare
Given the flexibility and scalability that NLP solutions deliver, as well as healthcare’s exceptional data challenges, there are myriad use cases for the technology across the industry by payers, providers, life sciences, technology providers, and more. Following are three common healthcare use cases for NLP:
- Risk adjustment: To succeed in value-based care payment models, it is essential that payers and providers accurately assess risk at the patient and population levels. When risk is not accurately measured, providers may leave money on the table, patients potentially suffer worse care and outcomes, and payers may jeopardize their bottom lines by underestimating premiums. The key to successful risk adjustment is to accurately identify patients’ full disease burden with substantiated data and documentation.
Historically, this process requires human reviewers to comb through thousands of pages of medical documentation to find key diagnoses, assign the correct ICD-10 code and correlate the appropriate HCC code for risk calculations. With NLP, accurate capture of risk is streamlined by processing unstructured and structured documents, automatically extracting diagnoses and applying the appropriate codes based on rules and specifications.
Risk adjustment is necessary to ensure that payers receive appropriate compensation for assuming responsibility for high-risk patients, and providers are compensated for accurately documenting and reporting patients’ conditions and treatment plans. The higher a patient’s risk, the greater the payment to payers and providers, so accurate risk capture is critical.
- Payment integrity: To successfully manage the challenges posed by dirty data, healthcare organizations must replace time-consuming and expensive manual processes with AI-based tools that comb patient records to determine payment accuracy. In the past, payers depended on expensive and time-consuming chart reviews to find and extract key unstructured data from patient records and claims, but more recently NLP has played an increasingly important part in helping confirm payment integrity.
In cases when payment integrity is in question, there is often a pattern of repeatability in the data, such as a large number of patients meeting the same prior authorization requirements. NLP helps payers detect these patterns that lack the natural variability found in legitimate patient records.
In the same respect, NLP can help payers spot unusual data that may be representative of fraud, such as expensive tests for which there is no medical necessity. With its ability to accurately analyze unstructured data to identify anomalies within records, NLP can quickly verify the presence, or lack of, critical data.
While even the most hard-working humans possess limitations on their ability to perform a high amount of chart reviews in a narrow timeframe, NLP automates the process, enabling substantial improvements in scalability. Because some complex medical records may consist of thousands of pages, NLP can drive significant savings in time and money in reviews.
- Population Health Analytics: Many healthcare organizations have leveraged NLP to advance population health initiatives by surfacing new insights from diagnoses buried in patient records. To improve the population of a subset of patients affected by a certain condition, it is essential to identify that condition as early as possible to prescribe the proper interventions before the disease escalates into a more costly issue.
NLP enables speedy and accurate symptom and diagnosis capture, helping healthcare organizations fully understand the disease burden for a given patient. For example, consider a population health initiative aimed at early identification and treatment of patients who are experiencing the beginning phases of chronic obstructive pulmonary disease (COPD). Early symptoms of COPD may include shortness of breath, lack of energy, and swelling in the lower extremities, but this information is often lost in the free text portions of patient records. By using NLP to analyze patient records, a healthcare organization can pinpoint these early signs of COPD and design appropriate care plans for these at-risk patients, leading to population health improvements.
Healthcare’s dirty data problem cannot be solved by manual processes and armies of chart reviewers. To properly perform risk adjustment and payment integrity functions, healthcare organizations must turn to technology to gain efficiencies and cut costs. SaaS-enabled NLP solutions provide the flexibility, scalability, transparency, and security to transform dirty healthcare data into useful, actionable insights.