Making the Most of Structured Data

Manual processes have bogged down payers in risk adjustment for years – if not decades. Now, an innovative approach to leveraging data enables a deeper understanding of member needs, increased revenue, and enterprise-wide benefits stemming from automation.

Most payers reveal that they capture 95% to 97% of their hierarchical condition category (HCC) codes correctly, a decent percentage of accuracy given the complexity of the task. Second- and third-level HCC code reviews by employees can be tedious, consuming a significant number of labor hours as coding specialists toil over PDFs pulled together from disjointed systems. While the 3% to 5% of not accurately captured codes may not seem like much, this additional revenue can translate into millions of dollars annually for a health plan.

Given current staffing shortages in healthcare and the complexity of member bases, the opportunity to identify and streamline missed revenue through automation is a sensible and worthwhile pursuit. Health plans can approach compliance more efficiently by implementing a strategy that leverages structured data about the patient’s condition. Further, securing concrete, actionable patient data enables payers to understand the needs of their member base more deeply and provide the appropriate level of support. As the industry strives for data use that is standardized, consistent, and fluid, payers that automate data flow, in the long run, will be able to discern the risk of each member and prioritize their needs to drive member services for outcomes-based care delivery.

Acknowledging the PDF problem

In the collection of patient information for risk adjustment, the portable document format (PDF), albeit imperfect, reigns supreme. A PDF captures the exact elements of a printed document to be viewed and shared consistently, regardless of systems integration. These documents, or electronic images, were made for humans to read—which they can do very easily. However, vital patient information stored in PDFs is unstructured, rendering it difficult for computers to interpret. Patient medical records are often long and complex, meaning that when health plans want to process the data digitally, they must turn the PDF into structured data. This reverse engineering approach, based on Optical Character Recognition and Natural Language Processing  can yield data inaccuracies and is generally inefficient.

Payers must embrace the value of compartmentalized structured data, leveraging the value of its harmonization to create a clearer, longitudinal record of each member. While PDFs provide valuable information for human risk adjusters, they do little to support data fluency, analytics, or the understanding of members that can be leveraged enterprise-wide. To effectively manage member medical complexities, health plans require accurate and comprehensive data about demographics, test results, and diagnoses in forms that computers can ingest and process. If they’re not there yet, payers will be heading there soon. PDFs will not go away but will be used in conjunction with structured data, pairing the efforts of humans and computers to drive the future of risk adjustment.

Growing confidence in data

The value of data in risk adjustment to support members is evolving in new applications.  As noted, using structured data in second-and third-level audits of HCC coding is an effective means of marrying human and computer strengths. Once reviewed by coding experts, computers can cost-effectively identify codes to add or delete during code review to help close gaps in justifiable revenues. Today, humans check the small subset of differences that the computer has identified to ensure compliance. Ultimately this automated identification process allows payers to receive enough money from CMS to support all members, yielding a significant return on investment.

Structured data also helps risk adjustment specialists optimize the sequence of charts to be coded. Due to the time constraints of human PDF review, they don’t have the resources to code every chart that comes across their desks. Instead, they have to make intelligent decisions about which members they’ll code right away and which members they’ll code when time permits. If they had structured data at their disposal and algorithms built off of that data, the decision would be supported by historical information about the presence of chronic conditions to facilitate member prioritization.

Thirdly, data can help with risk stratification of members, enabling health plans to have deeper insights into clinical risks and offerings to support in-need patients. Beyond clinical risks, demographic data or information relating to social determinants can help payers get ahead of the curve and eliminate barriers that impede successful outcomes.

Finally, structured data availability improves clinical document exchange with providers, enabling a streamlined approach to coding exactly what is presented in the member record.

Valuing a future of automation

Considering the value of structured data within the risk adjustment environment, it will soon be viewed as a necessity in the provision of modern, member-centric health plans. Data transformation indeed takes time, and despite the availability of structured data, risk adjustment teams are generally accustomed to workflows built upon unstructured data in the form of PDFs. There is also anxiety around change management: “upsetting the apple cart,” failing an audit, or eliminating employee positions in favor of machines. Alas, humans are imperfect, as are computers that perform code reviews independently with varying levels of accuracy. That is why human quality assurance is built into the process of automated code review and why computers are also performing related tasks like coding prioritization. Health plans are still relying on employees predominantly, using automation as an aid to process high volumes of member data.

Algorithms will grow stronger, instilling confidence in automated code reviews. Further in the future, both structured and unstructured data will be mined and harmonized across disparate systems to ascertain a more complete picture of patient health. Today, payers can glean value from structured data, breaking down data silos to understand the complexity of members and enabling physicians to provide optimal care for the system’s most vulnerable patients.