An investigation of the association between deteriorating economic environment and utilization of postlumpectomy radiation therapy (PLRT) using a difference-in-differences (DID) model.
Debra Nana Yeboa, MD, Xiao Xu, PhD, James B. Yu, MD; Yale School of Medicine
Introduction: Our prior work indicates that the receipt of postlumpectomy radiation therapy (PLRT) for early-stage breast cancer decreased from 80.7% in 2004 to 76.8% in 2009 among women over 40 years of age. Whether this decrease in the receipt of appropriate radiotherapy care for breast cancer was associated with economic factors is unknown. We therefore investigated the association between deteriorating economic environment and utilization of PLRT using a difference-in-differences (DID) model.
Methods: Using the Surveillance, Epidemiology, and End Results (SEER) registry database, we identified 21,816 women aged 40–85 years undergoing lumpectomy in 2004 or 2008. County-level unemployment rate was obtained from the US Department of Labor, Bureau of Labor Statistics and linked to SEER data by each patient’s zip code of residence. We used county unemployment rate as a marker for the economic environment of the patient’s residence area and compared 2004 and 2008 data, because the most recent economic recession began in 2007. To assess the impact of increasing unemployment, we focused our analysis on the 12,568 patients residing in counties that had an unemployment rate below 5.57% in 2004 (ie, the mean unemployment rate of the cohort in 2004). Counties with unemployment rates that rose from below 5.57% in 2004 to above 5.57% in 2008 were defined as having “increasing unemployment,” while those that consistently had rates below 5.57% in 2004 and 2008 were defined as having “stable unemployment.” Via a logistic regression, the DID model assessed the impact of “increasing unemployment” on the likelihood of receiving PLRT, while data on patients from counties with “stable unemployment” were used as a control condition to account for secular changes that occurred between those years.
Results: Receipt of PLRT decreased from 80.3% in 2004 to 79.8% in 2008 among women who lived in counties with increasing unemployment (P = not significant [NS]). In contrast, in counties where the unemployment rate stayed consistently low, the receipt of PLRT increased from 81.7% to 82.4% (P = NS). Unadjusted analysis suggests that patients living in counties that had increasing unemployment were significantly less likely to receive PLRT, compared with those living in counties that had consistently low unemployment (odds ratio [OR] = 0.88; 95% confidence interval [CI], 0.80–0.96). However, in the DID model, the interaction term between increasing vs stable unemployment and year was not statistically significant, indicating that increasing unemployment was not associated with an additional negative impact beyond other unmeasured variables that also changed from 2004 through 2008 (OR = 0.93; 95% CI, 0.78–1.11). In the adjusted model, being Medicare-age-eligible was associated with higher likelihood of PLRT receipt, while black race and estrogen receptor–negative status were associated with lower receipt of PLRT.
Conclusion: The impact of increasing vs stable unemployment rate in patients’ residence areas was not significantly associated with lower PLRT receipt. This indicates that the lower rate of PLRT receipt in 2008 in comparison with 2004 was due to multiple factors not completely encompassed by increasing unemployment alone.