Process measures for quality are subject to unintended consequences such as system gaming. As we have seen in diabetes care, this quality measure for bronchitis appears to have been manipulated by clinicians as well.
Conventional wisdom backed by extensive data attributes the over prescription of antibiotics for the treatment of acute respiratory tract infections (ARIs) to emerging antibiotic resistance, increased cost of care, and an increased risk of side effects. Of particular concern are antibiotic prescriptions for acute bronchitis.
In 2006, the National Committee for Quality Assurance (NCQA) estimated that 71 percent of commercially insured patients received antibiotics for the treatment of acute bronchitis. In 2007, avoidance of the use of antibiotics for treatment of adults diagnosed with acute bronchitis was added as a component of the Healthcare Effectiveness Data and Information Set (HEDIS) reported to the NCQA. The health system in this study responded by implementing a health-record based intervention with real-time physician feedback tied to diagnostic codes. The intervention changed display texts for code 466.0 to ACUTE BRONCHITIS (ABX NOT INDICATED) and 490 to ACUTE BRONCHITIS (COMPLICATED), but it was unclear whether the intervention would deter the misuse of antibiotics for acute bronchitis.
This three year, retrospective study reviewed electronic health records from 33 primary care practice sites within a large, integrated network across inpatient and outpatient settings. Visits for ARIs were identified through the primary diagnosis code, but analysis was limited to diagnosis codes 466.0 acute bronchitis and 490 bronchitis NOS in adults between 18 and 65 years old.
Results show that throughout the study period, 74.9 percent of all patients were prescribed antibiotics for acute bronchitis (466.0 and 490 combined). Analyzed by diagnostic code per study year, prescriptions for antibiotics were consistently high above 80 percent with visits coded 490, but the percentage of patients prescribed antibiotics for diagnostic code 466.0 had a dramatic third year drop falling from over 70 percent in 2006 and 2007 to 27 percent in 2008.
Eligible visits for acute bronchitis identified by ICD-9-CM codes 466.0 or 490 totaled 18,542. These visits showed little change for the study period between 2006 and 2009. Between 2006 and 2008 approximately 98.5 percent of these were coded 466.0. After the diagnosis display text changed in 2008, the used of code 466.0 was reduced, and 490 coding increased significantly to account for 84.6 percent of visits for bronchitis (p<0.001).
The evidence presented suggests that HEDIS measures combined with clinical reminders built into electronic medical records were able to decrease the use of prescription antibiotics for the treatment of acute bronchitis, but these results were more likely the product of diagnostic coding shifts rather than reduced use of antibiotics.
One common theme in quality improvement efforts is that providers are tasked with changing their behavior or the way they function within the larger system. Whether it is through team-based care, coordination across a continuum of care, or adhering to clinical practice guidelines, physicians in particular primary care physicians are charged with a host of new responsibilities which they, and their practices may not be ready or willing to face. A second theme emerging is that physicians, who are part of a systemic redesign of support systems (infrastructure, staffing, technology and training) which adapt to meet new responsibilities, tend to improve the quality of care they deliver.
This study demonstrated an example of system gaming, where physicians shifted their method of coding, rather than the intended behavioral change, which was to alter their method of diagnosis and treatment. The systemic support offered these physicians was to provide them with a choice when choosing the diagnostic code. One choice reminded them they may be penalized if they prescribe antibiotics, the other allowed them to proceed without penalty regardless of treatment decision.
The success of programs designed to alter physician behavior depends greatly on an investment in a system that enables physicians to change their behavior. This type of investment requires up-front financial support and sufficient time for cultural redesign. The results of this study suggest that simply telling physicians to change, even to behaviors that are widely accepted in the medical community, cannot be readily achieved without the proper backing from the entire system.