Patricia D. Franklin, MD, MBA, MPH; Brocha Z. Stern, PhD, MOT , 2025-04-16 15:02:00
April 16, 2025
4 min read
Click here to read the Cover Story, “Surgeons navigate patient-reported outcome mandate for TJA.”
Risk-adjustment is useful
In business, the focus on consistent outcomes includes managing “inputs,” such as uniform materials and well-trained employees, and consistent “processes” or work practices.
For example, to ensure flight safety, airlines manage consistent inputs and cancel flights if any defect is observed in the plane or if the pilots have flown too many consecutive hours. Flights are also canceled if the process is threatened by inclement weather. In contrast, the health care environment necessitates different approaches to quality. For example, diverse patients with varied clinical and social risks are regularly treated across hospitals. Preoperative patient optimization in total joint replacement is one attempt to minimize presurgical risks so patients are more similar at the time of surgery, but optimization cannot remove all risks. Thus, if one hospital cares for more patients with diabetes than another and patients with diabetes have a higher risk for infection, differences in infection rates across hospitals may be related to patient risks and not to the quality of care.
Patricia D. Franklin
In designing fair comparisons of the quality of care provided across multiple hospitals, analysts must consider varied inputs or the unique attributes of patients. Case mix or risk adjustment are statistical processes to minimize the effect of the variation in attributes of patients treated across hospitals. The goal is to improve the likelihood that outcome variation reflects the quality of care, rather than differences in patients who are treated. To effectively risk-adjust, the data from all sites must include the same patient risk factors. Many established risk factors, such as older age or comorbid medical conditions, are documented in the electronic medical record using standardized codes. Today, when comparing traditional clinical outcomes, such as post-TJR readmission rates or implant revision rates, analysts adjust for patient demographic and clinical data included in the billing data set before comparing outcomes.
As quality outcome analysis expands beyond clinical events to compare patient-reported outcomes (PROs) at 1 year following TJR, it is possible that additional risk-adjustment variables are needed. For example, researchers have documented that arthritis in multiple anatomic locations (eg, lumbar spine or nonoperative knee or hip) is associated with poorer functional gain after TJR. However, since multisite arthritis is not captured in a single arthritis billing code, it is incumbent upon health systems to capture PROs as well as new risk factors. Furthermore, social factors (eg, primary language and insurance coverage for physical therapy) may influence functional outcomes after TJR. Thus, risk-adjustment prior to comparing hospital-specific PROs after TJR will require careful re-review of key patient demographic, clinical and social risks that may vary across health care settings. Additional research is needed to define the optimal risk factors to ensure that comparative outcome data reflect variation in the quality of health care delivered, as opposed to variation in which patients were treated.
Effective risk adjustment is critical both for fair comparison of health system outcomes, as well as to minimize incentives for selection biases of patients based on pre-existing risks. Last, effective outcome comparisons are only the first step. If varied outcomes are uncovered, careful evaluation is needed to translate the analyses to improved health outcomes for all.
References:
For more information:
Patricia D. Franklin, MD, MBA, MPH, a professor in the department of medical social sciences at Northwestern University Feinberg School of Medicine, can be reached at patricia.franklin@northwestern.edu.
Caution is needed
Case mix adjustment specifically for social risk remains a topic of debate in quality measurement. Not adjusting can contribute to unfair penalties for hospitals (or surgeons) who treat more patients with social factors demonstrated to be associated with poor outcomes. Conversely, adjusting for social factors such as socioeconomic status or educational attainment can mask disparities if adjustment “excuses” inferior care quality that contributes to poorer outcomes. David R. Nerenz, PhD, and colleagues comprehensively summarized the nuanced arguments for and against social risk adjustment in quality measurement and advocate such adjustment as the “default” for health equity.
Brocha Z. Stern
However, patient-reported outcome-based performance measures (PRO-PMs) have unique considerations and may require a different default practice. For example, beyond the direct impact of social factors on outcomes, such risk could impact the measurement of outcomes if PROs have measurement invariance (ie, responses differ by social factors and not only the health outcome). In addition, PROs may be evaluated using thresholds for within-patient improvement. Since within-patient improvement already partially accounts for differences in preoperative health status, which may be closely tied to social factors, additional adjustment for social risk factors may only pose concerns of masking disparities without improving the statistical model. In contrast, if the quality outcome is based on the proportion of patients achieving a specific postoperative PRO score (instead of meeting an improvement threshold), the need for social risk adjustment may be heightened, given a likely worse preoperative starting point combined with barriers to postoperative recovery.
Beyond PROs as outcome measures for quality, it is important to note that PROs may also be framed as process metrics. In fact, payment penalties for the CMS total hip/knee arthroplasty PRO-PM are currently linked to a reporting vs. improvement threshold. However, many safety-net hospitals have decreased resources for robust PRO collection and may serve larger proportions of patients who are less likely to respond to PROs. Some level of facility-level social risk adjustment (or stratification with peer hospitals such as in the Hospital Readmissions Reduction Program) may be needed to avoid disproportionate payment penalties linked to PROs as a process metric for quality.
More research is needed to determine whether social risk factors should be included in the adjustment of PROs in quality measures, including which social factors and under what circumstances. Efforts are needed to simultaneously minimize disproportionate payment penalties linked to barriers to PRO collection in lower-resource settings, avoid incentivizing “lemon dropping” of patients with increased social risk for poorer outcomes and avoid adjustments that mask disparities in care quality.
References:
For more information:
Brocha Z. Stern, PhD, MOT, an assistant professor at Icahn School of Medicine at Mount Sina, can be reached at brocha.stern@mountsinai.org.