The Paradox of Coding: Policy Concerns in the Move to Risk-Based Provider Contracts

Health care providers submit medical claims to insurers that generally include a Current Procedural Terminology (CPT) code, which describes the medical, surgical, or diagnostic service that was provided to a patient as well as a series of International Classification of Disease (ICD-9 or -10) diagnostic codes that describe the conditions that prompted the treatment. Under fee-for-service reimbursement, the focus of health care organizations and insurers has been on accurate CPT coding to ensure that reimbursement accurately matched the services provided. In recent years, however, the medical diagnoses contained in claims have taken on increasing importance because diagnosis codes have become the principal basis for risk adjusting payments to providers and health plans under capitated or risk-based payment arrangements. The purpose of such adjustments is to fairly compensate health plans and providers for caring for sicker patients. For instance, health plans participating in the Medicare Advantage (MA) program receive higher rates for enrolling individuals with multiple conditions and more complex diagnoses. Without risk adjustment, plans would have strong incentives to enroll healthier patients. Indeed such “cream skimming” has been a major concern of policy makers throughout the history of Medicare managed care.¹

The growing use of risk adjustment has led health care organizations to increasingly focus on capturing all the relevant medical diagnoses of their enrollees in order to maximize risk-adjusted reimbursements. This represents a stark contrast to the past when capturing all diagnoses was not a priority for physicians because the diagnoses did not affect the fee received for a specified CPT code. For instance, health plans participating in the MA program now employ nurses or outside firms to identify and document missing or inaccurate diagnosis codes in order to increase their capitated payments. Risk scores for Medicare Advantage members have risen substantially faster than those for comparable beneficiaries in traditional Medicare. Beginning in 2010, CMS implemented an across the board reduction in MA risk scores to account for increasing coding intensity over time but it has not kept up with actual risk score growth.²

Although much of the attention on coding practices has focused on Medicare Advantage plans, diagnostic coding is taking on increasing importance for physician practices and health systems participating in global budget arrangements such as Medicare’s Accountable Care Organization (ACO) initiatives or commercial programs like Blue Cross and Blue Shield of Massachusetts’ Alternative Quality Contract. Global budget targets primarily are set based on prior years spending by an enrolled or statistically assigned population and targets are updated annually. However, the patients assigned to a particular ACO may vary by 30% or more from year to year, and continuously enrolled patients may develop new health conditions.³ Consequently, in most commercial global payment programs, each medical group’s target budget is adjusted by the relative change in its risk score compared to changes across the entire network.

To the extent that such adjustments reflect coding practices rather than true changes in patient complexity, however, these adjustments can have large implications. For instance if a group’s relative risk score falls by three percentage points without a real change in patient acuity, its global budget will be adjusted down by the same amount and vice versa. Since, on average, early evaluations of performance by both Medicare and commercial ACOs show savings in the range of 1-2% per year, a 3% downward adjustment could easily overwhelm the actual savings of any particular ACO and lead to aggregate losses. Similarly, if its relative risk score increased by 3% with a true change in patient acuity, an organization could earn large surplus payments without improving operating efficiency. Risk adjustment also generates financial uncertainty for medical groups because there is a substantial lag in information about the groups’ relative risk scores, which impedes groups’ ability to budget accurately and could lead to year-end shortfalls if projections are inaccurate.

For managers of health systems participating in global payment, improving risk score profiles is often seen as low hanging fruit for boosting financial results. Changing care patterns and lowering the costs of care is difficult, and most organizations are not very good at it. In contrast, improving coding acuity provides almost immediate returns and is easy relative to changing care patterns. Consequently, organizations participating in ACO contracts or Medicare Advantage may see much greater returns on coding initiatives than on investments to manage care more effectively. This is potentially even more important for physician organizations participating in risk contracts that may have more direct control over their physicians’ coding practices, in contrast to Medicare Advantage plans that generally are one step removed from coding process. Although as noted above, MA plans can and do submit supplemental codes that they obtain outside of routine medical care.

These issues are not just theoretical. After the introduction of inpatient prospective payment, the average hospital case-mix index, which directly reflects diagnostic coding, was much higher than expected, resulting in much higher payments to hospitals than Medicare had projected.⁴ Similarly, groups participating in CMS’s Physician Group Practice demonstration, an early test of ACO-like arrangements in Medicare, showed much higher growth in coding intensity than was seen in matched control groups.⁵

Indeed, one motivation for investing in upgraded electronic health records (EHR) systems is that they help organizations to capture codes more effectively. This raises equity concerns to the extent that safety net providers or less affluent providers lack the most sophisticated EHR systems or the resources to hire third-party vendors for “coding” initiatives and might suffer relative pay cuts simply because their coding practices are not “keeping up” with others in the market. In addition, the need for such investments might also contribute to consolidation, as independent community hospitals and physician groups will increasingly have difficultly competing with large corporate systems that have invested in systems that maximize coding.

Although accurate diagnostic coding is important and proper risk-adjustment is needed for global payment, we believe that the increasing impact of diagnostic coding on provider finances distracts physicians and health care organizations in risk arrangements from their core mission of improving the value of care. Hence, the paradox of coding. The question is what is the alternative? Of course, simply abandoning risk adjustment would bring us back to a system that encourages selection (“cream skimming”). One possibility is to use an entirely different approach to risk adjust payments. For instance, samples of enrollees of MA plans and ACOs are surveyed annually about their care experiences, and these surveys include some information such as global ratings of health that might be used for risk adjustment.⁶ A second option is to downweight the importance of risk adjustment in determining budgets such as limiting annual variation due to risk scores.

CMS currently controls risk score growth in its Medicare Shared Savings Program by incorporating risk scores for new enrollees but not allowing changes for continuously enrolled beneficiaries because of concerns about igniting the same risk score growth observed in Medicare Advantage. ACOs predictably object to this arrangement – particularly since ACO budget targets can be adjusted downward if risk scores of continuously enrolled members fall. This policy has been dropped in the Next Generation ACO initiative and instead the annual change in risk score is limited to 3%, which still could still result in large target budget adjustments.

As global budget arrangements continue to proliferate, diagnostic coding practices within healthcare organizations will take on increasing importance for determining financial performance. Many economists and policy makers have promoted the theory of managed competition where risk adjustment is critical element for establishing a level playing field among competing delivery systems or health plans.⁷ Under current systems of risk adjustment, however, effective coding practices provide a distinct competitive advantage, even if they may wane over time. Moreover, there is a fine line between reasonable improvements in coding accuracy and abusive or even overtly fraudulent coding practices. Although some form of risk adjustment is needed to mitigate incentives to avoid high risk patients, current systems have the potential to create further disparities between well-financed and struggling health systems and create further stimulus for consolidation. Of equal concern is its potential to shift management resources towards better documentation rather than initiatives to improve care. This issue deserves increased policy attention as it detracts from a core goal of health care reform – increasing the value of care.