Abstract
BACKGROUND: Using a health systems approach to investigate low-value care (LVC) may provide insights into structural drivers of this pervasive problem. OBJECTIVE: To evaluate the influence of service area practice patterns on low-value mammography and prostate-specific antigen (PSA) testing. DESIGN: Retrospective study analyzing LVC rates between 2008 and 2018, leveraging physician relocation in 3-year intervals of matched physician and patient groups. SETTING: U.S. Medicare claims data. PARTICIPANTS: 8254 physicians and 56 467 patients aged 75 years or older. MEASUREMENTS: LVC rates for physicians staying in their original service area and those relocating to new areas. RESULTS: Physicians relocating from higher-LVC areas to low-LVC areas were more likely to provide lower rates of LVC. For mammography, physicians staying in high-LVC areas (LVC rate, 10.1% [95% CI, 8.8% to 12.2%]) or medium-LVC areas (LVC rate, 10.3% [CI, 9.0% to 12.4%]) provided LVC at a higher rate than physicians relocating from those areas to low-LVC areas (LVC rates, 6.0% [CI, 4.4% to 7.5%] [difference, -4.1 percentage points {CI, -6.7 to -2.3 percentage points}] and 5.9% [CI, 4.6% to 7.8%] [difference, -4.4 percentage points {CI, -6.7 to -2.4 percentage points}], respectively). For PSA testing, physicians staying in high- or moderate-LVC service areas provided LVC at a rate of 17.5% (CI, 14.9% to 20.7%) or 10.6% (CI, 9.6% to 13.2%), respectively, compared with those relocating from those areas to low-LVC areas (LVC rates, 9.9% [CI, 7.5% to 13.2%] [difference, -7.6 percentage points {CI, -10.9 to -3.8 percentage points}] and 6.2% [CI, 3.5% to 9.8%] [difference, -4.4 percentage points {CI, -7.6 to -2.2 percentage points}], respectively). Physicians relocating from lower- to higher-LVC service areas were not more likely to provide LVC at a higher rate. LIMITATION: Use of retrospective observational data, possible unmeasured confounding, and potential for relocating physicians to practice differently from those who stay. CONCLUSION: Physicians relocating to service areas with lower rates of LVC provided less LVC than physicians who stayed in areas with higher rates of LVC. Systemic structures may contribute to LVC. Understanding which factors are contributing may present opportunities for policy and interventions to broadly improve care. PRIMARY FUNDING SOURCE: National Cancer Institute of the National Institutes of Health.