br Measures br The primary outcomes were U S Preventive
The primary outcomes were U.S. Preventive Services Task Force−recommended CRC and BC screening. Up-to-date (UTD) CRC screening was defined as having Protease Inhibitor Cocktail colonoscopy in the past 10 years, sigmoidoscopy in the past 5 years, or a stool test in the past year.21 CRC testing in the past 2 years was used to capture recent procedures and individual tests were also considered. UTD BC screening was defined as having a mammogram in the past 2 years.22 Prostate cancer screening was not examined because routine prostate-specific antigen testing was not recommended by the Task Force during the study period.23 Cervical cancer screen-ing was not considered because co−human papillomavirus and Pap testing was recommended beginning in 2012, but questions on human papillomavirus testing were not included in BRFSS until 2016.24 Questions on low-dose computed tomography for lung cancer screening were not included in BRFSS.25
The primary predictor variables were Medicaid expansion sta-tus and survey year. Respondents were grouped into the following categories based on the timing of their states’ expansion status: very early (VE; six states expanding March 1, 2010−April 14, 2011), early (21 states expanding January 1, 2014−August 15,
2014), late (five states expanding January 1, 2015−July 1, 2016), and not expanding (19 states not expanding as of January 1, 2017) (Appendix Figure 1, available online). Because of the differences in expansion timing, survey year was used to examine the matu-rity of expansion, shown in Appendix Figure 2 (available online). For example, 2012, 2014, and 2016 data represented the initial (1−2), 3−4, and 5−6 years following expansion for VE expansion states, respectively. For early expansion states, 2012, 2014, and 2016 data represented the pre-expansion, initial (up to the first year), and up to 2 years following expansion, respectively. For late adopters, 2012 and 2014 data represented the pre-expansion period, whereas the 2016 data captured the initial (1−2 years) fol-lowing expansion.
Absolute differences in crude screening prevalence estimates according to year and Medicaid expansion were compared using chi-square tests (a<0.05). Difference-in-differences (DDs) were used to determine whether changes in screening were greater in magnitude among Medicaid expansion states compared with non-expansion states (ref) and computed with logistic regression
models with predicted marginal probabilities including year, Med-icaid expansion, and Medicaid expansion X year terms.26,27 Abso- lute adjusted DD (aDD), additionally accounting for state, year, age, race/ethnicity, and sex, were similarly computed. BRFSS sampling weights and absolute changes were used to estimate the number of screened individuals.
Several additional analyses were conducted. First, screening among adults aged 50−64 years with medium/high incomes (≥$25,000) according to expansion and year was assessed. To
determine if disparities between low- and medium/high-income adults narrowed more rapidly in expansion states, adjusted differ-ence-in-difference-in-differences (aDDDs) were computed in models with a three-way interaction between Medicaid expansion X year X income. Second, changes in CRC and BC screening among low-income elderly respondents aged 65−75 and 65 −74 years, respectively, were assessed, as these mostly Medicare-insured groups were not anticipated to benefit from Medicaid expansion. To explore whether patterns in 2012−2016 were a continuation of past trends, pre-ACA CRC and BC screening among low-income adults aged 50−64 years were computed using 2006, 2008, and 2010 BRFSS data; however, owing to changes in survey design and questions on CRC screening testing, these estimates could not be directly compared to those ≥2012. Analyses were conducted in 2018−2019 using SAS-callable SUDAAN, version 9.4, and accounted for complex survey designs and non-response.
Among low-income respondents aged 50−64 years, there were a disproportionate number of Hispanic respondents in VE and non-Hispanic blacks in non-expansion states (Table 1). Patterns were similar among women aged 50−64 years (Table 1). Age increased slightly throughout the study period among early and non-expansion states, but other factors were unchanged (Appendix Table 2, available online).
Colonoscopy was the most commonly used test and past 10-year and 2-year use increased only among low-income residents in VE expansion states by 7.8% (p=0.001) and 5.1% (p=0.006), respectively (Appendix Table 3, available online). Stool-based test-ing increased nonsignificantly in VE states (by 3.8%, p=0.059) and increased significantly in non-expansion states (by 1.7%, p=0.033). Sigmoidoscopy use was uncommon (<5%) and changed little over time (data not shown).