Measurement Error in Public Health Insurance Reporting in the American Community Survey: Evidence from Record Linkage

Self Cite

Objective To measure the accuracy of survey‐reported data on features and type of health insurance coverage. Data Source Enrollment records from a private insurer were used as sample for primary survey data collection in spring of 2015 using the Current Population Survey health insurance module. Study Design A reverse record check study where households with individuals enrolled in a range of public and private health insurance plans (including the marketplace) were administered a telephone survey that included questions about general source of coverage (eg, employer), program name (eg, Medicaid), portal, premium, and subsidies. Data Collection/Extraction Methods Survey data were matched back to enrollment records, which indicated coverage status at the time of the survey. Concordance between the records and survey data was assessed. Principal Findings Correct reporting of general source of coverage ranged from 77.8 percent to 98.3 percent across coverage type, premium ranged from 91.6 percent to 96.4 percent, and subsidy ranged from 83.0 percent to 91.0 percent. Using a conceptual algorithm to categorize coverage type resulted in sensitivity of 98.3 percent for employer‐sponsored enrollees, and 70.6 percent‐77.6 percent for the other coverage types, while specificity ranged from 93.9 percent to 99.4 percent across coverage types. Conclusions Survey reports of features of coverage suggest they are viable items to include in an algorithm to categorize coverage type. Findings have implications beyond the CPS, particularly for marketplace enrollees.

Section: Methodsmentioning

confidence: 99%

Assessing the accuracy of survey reports of health insurance coverage using enrollment data

Pascale

Fertig

Call

2019

Self Cite

“…Survey errors arise from sampling, sample coverage, response, and data processing errors. The most important of these errors is thought to stem from the response rather than data processing, although less is known about the potential impact of sample coverage error . A description of the design features and respondent factors that shape health insurance responses is reviewed elsewhere …”

Section: Introductionmentioning

confidence: 99%

Medicaid expansion and the Medicaid undercount in the American Community Survey

Boudreaux

Noon

Fried

et al. 2019

Self Cite

Objective To measure discordance between aggregate estimates of means‐tested coverage from the American Community Survey (ACS) and administrative counts and examine the association of discordance with ACA Medicaid expansion. Data Sources 2010‐2016 ACS and counts of Medicaid and Children's Health Insurance Program enrollment from the Centers for Medicare & Medicaid Services. Study Design State‐by‐year counts of means‐tested coverage from the ACS were compared to administrative counts using percentage differences. Discordance was compared for states that did and did not adopt expansion using difference‐in‐differences. We then contrasted the effect of expansion on means‐tested coverage estimated from the ACS with results from administrative data. Data Collection/Extraction Survey and administrative data. Principal Findings One year before expansion there was a 0.8 and 4 percent overcount in expansion and nonexpansion states, respectively. By 2016, there was a 10.64 percent undercount in expansion states vs a 0.02 percent undercount in nonexpansion states. The ACS suggests that expansion increased means‐tested coverage in the full population by three percentage points, relative to five percentage points suggested by administrative records. Conclusions Discordance between the ACS and administrative records has increased over time. The ACS underestimates the impact of Medicaid expansion, relative to administrative counts.

“…This question design aligns more closely with surveys such as the American Community Survey (ACS) and the National Health Interview Survey which ask about health insurance coverage at the time of the survey. Whereas the CPS ASEC has consistently produced a Medicaid undercount, the ACS has been shown to produce an overcount of Medicaid recipients …”

Section: Introductionmentioning

confidence: 99%

Response error and the Medicaid undercount in the current population survey

Noon

Fernández

Porter

2018

Objective To measure the Medicaid undercount and analyze response error in the 2007‐2011 Current Population Survey Annual Social and Economic Supplement (CPS ASEC). Data Sources/Study Setting Medicaid Statistical Information System (MSIS) 2006‐2010 enrollment data linked to the 2007‐2011 CPS ASEC person records. Study Design By linking individuals across datasets, we analyze false‐negative error and false‐positive error in reports of Medicaid enrollment. We use regression analysis to identify factors associated with response error in the 2011 CPS ASEC. Principal Findings We find that the Medicaid undercount in the CPS ASEC ranged between 22 and 31% from 2007 to 2011. In 2011, the false‐negative rate was 40%, and 27% of Medicaid reports in CPS ASEC were false positives. False‐negative error is associated with the duration of enrollment in Medicaid, enrollment in Medicare and private insurance, and Medicaid enrollment in the survey year. False‐positive error is associated with enrollment in Medicare and shared Medicaid coverage in the household. Conclusions Survey estimates of Medicaid enrollment and estimates of the uninsured population are affected by both false‐positive response error and false‐negative response error, and these response errors are non‐random.