EDITOR’S NOTE: What follows is the second piece of a two-part series examining how health risk, severity, and complexity impact healthcare policy, payment, and quality assessment.
There are currently many different risk adjustment models in use or under consideration in the healthcare industry. Each model has advantages and disadvantages. It is safe to say that no model adjusts risk at a level that always recognizes the actual clinical risk, severity, and complexity of the patient condition. Most of these models have been developed based on historical data.
Unfortunately, historical data does not come close to meeting the data requirements to accurately assign risk. This lack of data quality poses fundamental challenges to establishing a baseline moving forward in a value-based purchasing environment.
Many clinicians have not embraced the capture of complete, accurate, and specific codification of each patient’s health status; as significant as it is, they do not see the value in doing so. As a result, such clinicians will seek shortcuts to identify the easiest-to-use codes that will satisfy the requirements that result in appropriate payment. They historically have not had an incentive to perform complete, detailed, and accurate coding, as those activities previously rarely impacted payment dramatically. While some of those incentives are shifting more favorably for some providers, historical data has been lacking.
An analysis of three years of payer data demonstrates that more than half of all codes submitted on claims would be considered “unspecified” or a “symptom.” The tendency to assign such non-specific codes hinders one’s ability to determine the real financial risk and clinical risk of disease processes.
Hierarchal Condition Categories (HCCs), along with other demographic factors, have been a mainstay of risk adjustment since 2004 for the Medicare Advantage (Part C) program. These Centers for Medicare & Medicaid Services (CMS) HCCs include 79 categories that represent potential for risk adjustment. While there has been significant developed experience with the use of these categories for Medicare Advantage premium adjustment, they have significant limitations for use within models proposed by the Medicare Access and CHIP Reauthorization Act of 2015 (MACRA).
This is in contrast to the U.S. Department of Health and Human Services (HHS) HCCs that were introduced in 2014 to support Medicaid exchange risk adjustment under the Patient Protection and Affordable Care Act. These HCCs include 126 defined clinical categories that are proposed for use in risk adjustment models and provider resource use metrics under MACRA. The specifics of risk adjustment for outcome measures are not clear at this time, as they are continuing to evolve.
Challenges for proposed risk adjustment approaches
There are a number of challenges moving forward relative to financial and clinical risk adjustment under any of the current or proposed value-based purchasing or alternative payment models:
• Limited historical data quality – The data used to validate the predictive capabilities of risk categories as noted above lack sufficient data quality, by most applicable measures.
• ICD-9 to ICD-10 – The move to ICD-10 altered the landscape of health condition risk. There has been insufficient time to accumulate data to redefine risk as it pertains to the new ICD-10 codes. Many providers have not leveraged the ability of ICD-10 to capture greater levels of risk, severity, and complexity. Systems to support coding have tended to simply convert old, non-specific ICD-9 codes into equally vague ICD-10 codes. Since these ICD-10 codes are still relatively new to providers, accuracy in coding may be adversely impacted for some period of time.
• Insufficient granularity – The currently defined risk categories do not have the granularity to distinguish significant variation in risk within different clinical domains. The additional parameters related to risk have not been incorporated into these categories. For example, ICD-10 codes for open fractures have been defined based on the Gustilo classification, which recognizes dramatic differences in the potential costs and complications for open fractures. This difference and many others are not reflected in the proposed risk categories or adjustments for outcome measures. Organizations that take on high-risk cases may not have healthcare outcomes or revenue appropriately adjusted to reflect the risk severity and complexity of those patients as compared to other providers who are caring for similar clinical domains, but with a less complex population of patients.
• Incomplete definitions – There is a lack of clear definitions for currently defined risk categories. It is unclear which codes should or should not be included in these categories and why.
• Lack of clinical relevance – A clinical analysis of these categories calls into question the intent of defining the categories. From a clinical perspective, the codes included do not seem to align with the apparent definition of the category.
• Lack of homogeneity – Looking at the codes within each category, there appears to be a lack of homogeneity in the characterization of the disease process and/or the level of risk severity or complexity of any of the codes within.
• Focus on premium – The primary goal of most risk adjustment categories has been to focus on actuarial stratification of a population based on historical claims data. Application of these categories to the true risk, severity, and complexity of the patients’ conditions, being as it might impact the cost of care or the likelihood of positive outcomes, raises significant questions.
While there will never be a perfect model, there is a lot of work that needs to be done on both data quality and the methods for risk adjustment to achieve some level of confidence in this process for those impacted by these metrics.
Requirements to Succeed
Waiting for the realization of a perfect world of data does not address the challenges that healthcare entities currently face. A shift to value-based models has and will continue to move forward. It is important to have visibility at all levels of details into the methodologies that are implemented to attempt to influence these methodologies wherever possible. Regardless of the challenges however, change will occur.
All healthcare entities will need to prepare for these changes without necessarily knowing the eventual details of how they will be implemented. There are common steps that should be taken regardless of the exact approaches that may be implemented. There is little doubt that accurate, reliable, and complete data reflecting the nature of patients’ conditions, as well as the risk, severity, and complexity of those conditions, is essential to success in this new value-based purchasing environment.
Understanding your own data
Most provider organizations do not have the tools they need to fully understand their own experience as measured by the data they produce. Entities outside of their domain may know more about their data enterprise than they do. A recent news story that examined mortality rates for a Florida hospital suggested less-than-positive outcomes. When presented with this data, the hospital’s administration could not respond adequately with better data, and as a result, they changed their business practices based on that news report.
Healthcare entities should always seek to have a better understanding of their own data than any outside entity. Cost and quality measures should be known and addressed internally before organizations are forced to change based on outside audits or other disclosures. You can’t improve on what you can’t see. In a value-based purchasing world, you need to know and manage your own value.
Focusing on continuous data improvement
As previously noted, the reliability, accuracy, and completeness of data about patients’ health conditions are less than optimal for most entities. The incentives for achieving quality data have been lacking. Strategies for improving data quality must include the following:
• Establishing the case within the organization for the value of high-quality data and creating incentives for data excellence
• Leveraging all data to understand patterns of coding and documentation to improve data quality
• Creation of a data governance structure with all relevant stakeholders to ensure that the focus on data quality is empowered and crosses organizational boundaries
• Establishing mechanisms for operational data collection that reduces the burden on clinicians while assuring that all relevant parameters of the patient risk, severity, and complexity are captured
Leveraging actionable data to position for success
Data should not just be something that is reported, as this in of itself is an activity that does not result in meaningful change. Rather, data and its analysis should be a robust tool for answering questions and driving adaptation and course corrections to clinical activities and business operations within an organization. That analysis should go beyond merely raising questions and instead provide reliable information to drive the pursuit of opportunities. Those results cannot be achieved without first having reliable, accurate, and complete data about the nature of the clinical care the organization is providing, as well as the key parameters of the patients’ conditions that is the basis for that care.
Once data quality is improved, analytic knowledge and tools are needed to take advantage and to inform relevant stakeholders. Financial stakeholders need to understand their financial risk. Executives need to know how to change strategies and approaches. Clinical leaders need to understand where to focus clinical quality and efficiency efforts.
Data about patterns of coding and clinical documentation is needed for leveraging and focusing improvement. Data provides little value if it is not reliable, accessible, and leveraged for change.
Dramatic changes are occurring in how healthcare will be financed in the future. The evolving value-based purchasing environment can be overwhelming in its complexity. Proposed models lack directional clarity and stability. Preparing for this inevitable move toward value-driven versus Service-driven healthcare policy means understanding the driving forces and the common infrastructure requirements that will persist, independent of the details of any given model or methodology.
There is no doubt that healthcare is becoming progressively more data-driven, and that high-quality data about the parameters of patient conditions will be essential to success.
A special thank you to Bob Perna, director of healthcare economics at the Washington State Medical Association, for his support, advice, and input into this paper.