Over the past 25 years, we've seen a dramatic shift in how patients' pain has been assessed and treated. The drive to capture pain as a vital sign drew attention to the need for better control of patients' pain. The coincident release and aggressive marketing of new types of opioid pain medications dramatically increased opioid use. This tragically led to an epidemic of overdoses and opioid use disorder. The pendulum has now swung back towards a more conservative approach to therapy. This also led to a realization that a more nuanced approach to assessing pain is critical: the many biopsychosocial factors that contribute to a patient's experience of pain make distilling it down to a single number unrealistic.
In this new era of pain assessment and management, good performance measures are needed in order to benchmark how physicians, medical groups, and health plans perform, and to track improvements. To that end, the Performance Measurement Committee (PMC) of the American College of Physicians (ACP) of existing performance measures in pain management, and made a recommendation for a proposed measure concept that we feel would be helpful in moving patient care forward in this arena.
Our Methodology
The PMC is composed of 12 internal medicine physicians with diverse professional and geographic backgrounds. We use a standardized methodology to search for applicable metrics promulgated by national quality organizations, and assess their quality on several axes. These include clinical importance, appropriate resource use, evidence base, specification adequacy, and feasibility and usability.
For each measure, we evaluate whether it's applicable at each of three levels of attribution -- that is, if it is appropriate to measure a health plan, a medical group, or an individual physician. We look for adequate testing of the measure at each of these three levels before we will recommend it to be used in high-stakes programs that may impact physicians' job satisfaction, income, or reputation. If the measure was not designed for or tested at a given level of attribution, we do not recommend that measure, but we may offer comments on how it could be improved, once tested appropriately.
While performance measures should be based on evidence-based clinical guidelines, they are different animals. Guidelines can allow for clinician judgment and patient preferences (i.e., shared decision-making), whereas measures must be quantitative, and based upon data in medical records or structured data fields for electronic data reporting. Non-compliance with guidelines has little downside, but poor performance on measures can impact reimbursements and reputations for physicians or health plans. Poorly designed measures can also lead to patient harm if they create perverse incentives that lead to under- or overtreatment.
Evaluating Performance Measures
In the case of pain, we identified six performance measures, three of which the PMC deemed valid. Due to this small number, we then outlined a proposed performance measure concept that addresses a gap in measurement in the appropriate management of acute pain; it is based on a recent ACP clinical guideline.
Measures the PMC Supported
CBE 2950 measures situations in which a patient has opioid prescriptions from multiple providers, a signal of potential misuse of opioids by the patient. We support the use of this measure at the health plan level.
CMIT 746 measures overuse of imaging in acute (28 days after first diagnosis) low back pain. This measure has been tested and is valid at the health plan level. A similar measure designed and tested at the individual physician or medical group level would also reduce the overuse of imaging in this patient population, as physicians are the loci of control for appropriate imaging.
CMIT 748 measures the prescription of high dose opioids (over 120 morphine milligram equivalents [MME] for 90 days) in patients without cancer. We support the use of this at the health plan level.
Measures the PMC Did Not Support
CBE 2951 measures overuse of opioids at high dose and from multiple providers/pharmacies. Combination measures (CMIT 248 and CBE 2950) like this conflate many clinical variables, which can lead to unintended consequences. The complexity could lead to errors in implementation. As a result, we rated it uncertain validity at the health plan level, and we instead recommend the measurement of the subsidiary concepts.
CBE 150 attempts to measure concurrent use of opioids and benzodiazepines. Although taking these together is a risk for overdose, the measure specifications do not distinguish between concurrent prescriptions for the two classes and patients actually taking them together. There is no accommodation for shared decision-making. For these reasons, we do not give our support.
CBE 3666 measures patient satisfaction for pain-related visits to palliative care. We were concerned about denominator specifications excluding inappropriate visits, burden of collecting data from surveys, response bias, and the risk of bias towards medical groups with less resources. Implementation of this measure can lead to overuse of unnecessary treatments and increase the risk of harms.
A New Concept
Our analysis revealed a gap: appropriate treatment for acute pain. We built upon a recent ACP Clinical Practice Guideline for acute pain treatment and proposed a new feasible measure concept that could be implemented at the physician level and help address the problem of opioid overuse.
We propose an overuse measure concept in the context of patients with acute (<4 weeks) non-low back, musculoskeletal pain. The numerator is simply the number prescribed an opioid acutely with a denominator including patients 18 years and older with acute non-low back musculoskeletal pain. We do have some exclusions, such as patients with low back pain or with active cancer. This measure concept would not be appropriate for accountability but for quality improvement purposes only.
As AI tools become more commonplace in routine care, we look forward to the time when natural language processing extracts medical concepts from clinical documentation and stores it in standardized formats. With more structured data to incorporate into the logic of performance measures in standardized ways, we will be able to account for many of the factors that lead clinicians to feel like the performance measures don't accommodate the richness of their interactions that demonstrate commitment to the principles of quality improvement.
Overall, we recommend more testing and modification as needed. We also recommend a new measure concept for acute pain. Taken together, these could provide fair and useful measures to compare performance and drive clinical changes for more appropriate use of opioids in acute and chronic pain patients.
Scott MacDonald, MD, is affiliated with UC Davis Health System and specializes in internal medicine and clinical informatics in Sacramento, California. Amir Qaseem, MD, PhD, MHA, is senior vice president of clinical policy and chief science officer at the American College of Physicians.