Census Bureau: 'Evaluating Subannual Health Insurance Coverage Estimates in Current Population Survey Annual Social & Economic Supplement'
* * *
Here are excerpts of the report:
Although annual estimates of health coverage paint a broad picture of the health care landscape, estimates of within-year transitions in coverage or health insurance churning provides a more nuanced portrait of the nation's health and spending on healthcare. Understanding who loses coverage and for how long is crucial for measuring potential health needs and medical expenditures of the population (Graves & Swartz, 2013; Sudano & Baker, 2003). For example, people who lose health insurance coverage for one or more months are more likely to delay treatment (Aiken, 2004; Sudano & Baker 2004), experience negative health events, and face higher healthcare costs than those insured for the entire year (DeVoe et al., 2003; Schoen & DesRoches, 2000). Further, enrolling and re-enrolling individuals on plans increases administrative costs and aggregate health care spending. As health insurance instability is associated with higher costs (DeVoe et al., 2003; Schoen & DesRoches, 2000), and longer periods without health insurance coverage are associated with poorer health outcomes (DeVoe et al., 2008; Olson et al., 2005), additional research can inform which subgroups are most likely to experience transitions in coverage and the health policy implications of health insurance transitions.
This paper evaluates monthly health insurance data in the Current Population Survey Annual Social and Economic Supplement (CPS ASEC). Prior research indicates that the 2014 CPS ASEC questionnaire redesign helped to improve the quality of annual estimates of health insurance coverage in the CPS ASEC (e.g., Medalia, O'Hara, & Smith, 2015; Pascale, Boudreaux, & King, 2015)./2
Updates to the CPS ASEC processing system completed in 2019 further improved these estimates (Jackson and Berchick, 2020; Berchick and Jackson, 2019). Yet, no existing research has evaluated the sub-annual estimates of coverage available in the redesigned CPS ASEC. Using restricted-use CPS ASEC data, this analysis fills this gap by comparing 2018 and 2019 CPS ASEC health insurance spell and transition data with the 2017 and 2018 Medical Expenditure Panel Study (MEPS) and the 2018 Survey of Income and Program Participation (SIPP) Wave 1, two of the most commonly used datasets for examining sub-annual dynamics in health coverage (e.g., Graves & Mishra, 2016; Vistnes & Cohen, 2018).
* * *
Discussion
The recent redesign of the CPS ASEC questionnaire, together with an updated processing system that enable researchers to leverage the monthly level health coverage data collected by the redesigned questionnaire offer the potential for an additional source of data to examine health insurance dynamics. In this paper, we evaluate how subannual coverage estimates in the 2018 and 2019 CPS ASEC compare with these measures in SIPP and MEPS, the two leading sources of monthly health insurance data. Our results suggest that there are significant differences in estimates of health insurance dynamics between the surveys. In general, there were longer uninsured spells and fewer transitions between coverage states during 2017 and 2018 in the CPS ASEC (and in SIPP) compared with MEPS./18
Although there were some significant differences between the CPS ASEC and SIPP in terms of transitions in health coverage and spell characteristics, in general SIPP, like CPS ASEC, was marked by relatively few transitions to or from coverage over the course of 12 months.
Research suggests that there has been more stability in insurance coverage and less Medicaid churning since the implementation of the ACA (Vistnes and Cohen 2018; Goldman and Sommers 2020). Yet, there are differences across the surveys in health insurance dynamics. Differences in survey design likely contribute to the results presented here. The CPS ASEC is a cross-sectional supplement, whereas MEPS and SIPP are both panel studies designed to follow individuals over a longer time period. MEPS respondents are interviewed 5 times over approximately two years, reporting on coverage in the roughly 3 to 7 months prior to each interview, while respondents to the CPS ASEC are interviewed in February through April, and asked to report their coverage in the previous calendar year, extending the recall period to 14-16 months prior to the interview. With a longer recall period, it is not surprising that the CPS ASEC show more stability in coverage (or uninsured status) than MEPS.
Although SIPP is also a panel survey, in the 2018 panel respondents are also asked about health insurance coverage in each month of the previous calendar year. The 2018 SIPP Wave 1 interviews occurred between January and
This longer recall period likely results in the report of fewer transitions in both SIPP and the CPS ASEC surveys, all else equal, although more research is needed. As a result, several summary estimates reported in Tables 1 and 2 were not significantly different for CPS ASEC and SIPP. Although analyses revealed some significant differences between these two surveys in the cumulative risk of experiencing a transition to uninsured status, results were not substantively different between the surveys, compared with MEPS.
Second, differences in the sampling and oversampling strategies between the surveys (e.g., with respect to race/ethnicity) suggests that the MEPS sample population has a profile that is more likely to experience a transition (e.g., Kirby &
In addition, attrition is an issue with longitudinal surveys and, may have a greater impact on MEPS than on SIPP in this study. Although SIPP is also a longitudinal survey, we use the first wave of the 2018 SIPP, and therefore respondents have not had an opportunity to attrite. In contrast, because they may be interviewed more than once a year, MEPS respondents may attrite during the reference year. If respondents who attrite from the MEPS sample are more likely to be disadvantaged or more stably uninsured, then we would expect shorter uninsured spells or more transitions in MEPS than in the CPS ASEC or SIPP.
Seam bias may also contribute to more frequent transitions in MEPS-HC, although dependent interviewing attempts to mitigate this source of bias. Although beyond the scope of this paper to evaluate potential seam bias, an examination of reported spell beginnings in Appendix Figures A1 and A2 suggest a higher proportion of uninsured spells in MEPS beginning in March and April (at the beginning of a reference period for interviews 2 and 4), and in September and October (early in the reference period for interviews 3 and 5).
Notably, this analysis is limited to examining subannual coverage and transitions in coverage within a year. Most studies of health insurance dynamics using SIPP and MEPS have leveraged the length of a panel to examine transitions in coverage over a longer period of time. Thus, we may not expect to capture as many transitions within the course of a year as one would over a 24 or 48 month period.
Further, this paper evaluates transitions to and from coverage. Yet individuals may transition from one type of coverage to another. Future research will evaluate such transitions between coverage types.
Although it is beyond the scope of this paper to examine whether transitions in health insurance coverage are reliably reported across the surveys, researchers should use caution in interpreting subannual coverage and health insurance transitions in the CPS ASEC. Despite its larger sample size and rich demographic detail of the CPS ASEC, the infrequency of incident spells and stability in coverage retrospectively reported in the cross-sectional CPS ASEC may pose challenges for subgroup analysis or analyses of transitions between subtypes of coverage.
In addition to examining health insurance dynamics by coverage type and for broad subgroups, future evaluation of subannual coverage in the CPS ASEC will further explore explanations for differences in insurance spells and transitions by examining whether the imputation of missing data contributes to differences in health insurance dynamics between surveys. In addition, we plan to compare transitions in Medicaid coverage reported in Census households with administrative records (e.g. Medicaid enrollment data) to further evaluate subannual coverage in the CPS ASEC and SIPP.
* * *
View full report at https://www.census.gov/content/dam/Census/library/working-papers/2021/demo/sehsd-wp2020-21.pdf
Wellcare by Health Net Medicare Advantage Plans in California Earn 4-Star Rating in Annual CMS Star Quality Ratings
Journal of Disability Policy Studies Issues Research Articles in December 2021 Edition
Advisor News
Annuity News
Health/Employee Benefits News
Life Insurance News