By: Dr. Victor Lee
Value sets are collections of coded clinical and administrative concepts that are critical components of analytics, research, clinical decision support, and many other health IT use cases. The accuracy of insights and the overall value derived from these initiatives is often proportional to the quality of the value sets. Let’s explore a fictional case series that illustrates some unintended consequences of poor-quality value sets. Note that while the specific details have been made up, the shortcomings of the value sets are firmly grounded in reality.
Case 1: Inaccurate Analytics
A multi-facility health system has 5 different electronic health record systems in use due to mergers and acquisitions. An analytics team has been charged with tracking longitudinal sepsis diagnoses, treatments, and outcomes over a 10-year period. Diagnosis codes were extracted from billing data across the 5 facilities. The analytics team developed a sepsis value set consisting of ICD-10-CM codes which contained all of the concepts under “Streptococcal sepsis” (A40), “Other sepsis” (A41), and various organism-specific sepsis codes. However, it accidentally omitted codes under “Puerperal sepsis” (O85) which were deemed to be in scope based on the project charter. Due to the incompleteness of the value set, many patients that should have been captured for this analysis were omitted, so the initiative failed to represent an important subpopulation.
Case 2: False Positive Cohort Identification
A research team secured a grant from the National Institutes of Health to study the comparative effectiveness of various multi-drug regimens for the treatment of essential hypertension. The researchers created comprehensive value sets to represent various classes of antihypertensive agents. After several months of data collection and analysis of the antihypertensive regimens, a research assistant noticed that some patients were incorrectly identified as having a multi-drug regimen consisting of minoxidil in combination with other agents. The error was traced back to the inclusion of topical minoxidil in a value set. While minoxidil was originally developed as an antihypertensive agent, its use nowadays is often reserved for patients with resistant hypertension. However, the topical formulation of the drug is used exclusively for hair loss, and the oversight resulted in significant rework.
Case 3: Outdated Clinical Decision Support
An ambulatory clinic had implemented rules-based alerts in an effort to optimize heart failure outcomes. A clinical decision support rule was designed to identify patients with heart failure who were not prescribed either an angiotensin-converting enzyme inhibitor (ACEI) or an angiotensin receptor blocker (ARB), both of which had been shown to improve various heart failure outcomes. Comprehensive value sets for ACEIs and ARBs were developed by a clinical informatics team and validated by pharmacists to ensure optimal performance, and the first year after go-live was a success. However, physician staff members had recently submitted multiple complaints about the alerts malfunctioning and wasting their time. It was discovered that a relatively new ACEI-diuretic combination drug had become a popular choice for patients due to its convenient dosing, and the alert failed to recognize that these patients were on proper ACEI therapy because the dependent value sets had become outdated. Although the ACEI value set was subsequently updated to include the new ACEI-diuretic combination, the delay in accounting for this change had frustrated many physicians and eroded trust in their technology investment.
Many other examples of suboptimal outcomes related to inaccurate value sets are discussed by Wright et al. Fortunately, there are things that can be done to ensure that your value sets satisfy your business objectives. Our value set experts at Clinical Architecture have many years of experience with defining best practices for value set authoring and maintenance, and their collective wisdom has been distilled into The 6 Cs Framework for Assessing Value Set Quality. As the name implies, there are 6 dimensions, each starting with the letter C, and they are categorized under accuracy and usability domains. Download the white paper, apply the framework to your value sets, and maximize the value of your health IT investments.