Pre-Existing Condition Prevalence for Individuals and Families

The impending decision by the Fifth Circuit Court of Appeals in the Texas v. Azar case raises the prospect that insurers will once again be able to return to using people’s health status in determining their eligibility and premiums for health insurance, at least for coverage obtained from the non-group, or individual insurance, market.  In the case, the plaintiff states’ Attorneys General and the Trump Administration are arguing that the Affordable Care Act is unconstitutionally structured and should be invalidated in its entirety.  This would include overturning provisions that guarantee that people with pre-existing health conditions cannot be denied coverage or charged higher premiums due to their health status.

Given the significant barriers to coverage that may reemerge if these provisions in the ACA were to be invalidated, we are updating our prior work looking at the share of nonelderly adults with health conditions that would likely to have caused them to be denied coverage if they applied for non-group health insurance prior to the effective date of the ACA.  And because the financial consequences of these changes would potentially affect the whole family, we extend our analysis to estimate the percent of nonelderly families with at least one adult who has one or more declinable conditions.

Consistent with our previous analysis, we estimate that 27% of nonelderly adults have a declinable health condition, which is about 53.8 million people in 2018.  We further estimate that 45% of nonelderly families have at least one nonelderly adult member with a declinable health condition.  Finally, we update our state-based estimates of the prevalence of declinable pre-existing conditions with the most current data available, showing that the share of non-elderly adults with pre-existing conditions ranges from 22% in Colorado to 37% in West Virginia.

People with pre-existing health conditions were often denied coverage or charged higher premiums for individual market coverage before the ACA took effect in 2014.  While most of people with pre-existing health conditions are covered currently by employer-based coverage or public programs, such as Medicaid, the non-group market is where they may need to look for coverage in times of transition, for example, if they lose a job, change jobs, start a business, divorce, age-off of a parent’s policy, retire before age 65, leave employment due to serious illness, get a job and lose Medicaid, or otherwise lose their eligibility for work-based or public coverage. While we cannot predict how the court would fashion relief if these ACA provisions were overturned, access to individual market insurance for people with pre-existing conditions could be seriously reduced.

Use of Health Status in Underwriting and Rating Before the ACA

Estimates of the Share of Adults with Pre-Existing Conditions

We used data from the National Health Interview Survey (NHIS) to estimate that 27% of nonelderly adults had a declinable health condition in 2018, the same percentage that we found in our earlier analysis for 2015. The NHIS has a number of questions about whether the respondent has ever been diagnosed with a number of the health conditions that would have been declinable in the pre-ACA non-group market. While we cannot duplicate the underwriting processes carried out by insurers, we feel that our approach is reasonable and may be conservative because the NHIS does not contain information about all of the conditions (e.g.,  AIDS/HIV) used by insurers and does not provide information on prescriptions that insurers also used to decline applicants for coverage.

Although each family member would have been separately underwritten in the pre-ACA non-group market, the economic consequences of having a member of the family denied coverage or surcharged due to their health would likely be felt by all members of the family. To look at the number of people that might be affected, we extended our previous methods and estimate that, in 2018, 45% of non-elderly families included a non-elderly adult with a declinable condition. Individuals living in households without a relative are considered to be a family of one person for this analysis.

A larger share of non-elderly adult women (30%) than men (24%) have declinable pre-existing conditions in 2018, unchanged from 2015.  We estimate that 23.7 million men have a pre-existing condition that would have left them uninsurable in the individual market pre-ACA, compared to 30.1 million women. Pregnancy explains part (about 2 million women) but not all of this difference.

The prevalence of declinable conditions also increases with age among non-elderly adults: ranging from 18% of those in the 18-34 age group to 44% for those in the 55-64 age group.

The rates of declinable pre-existing conditions continue to vary from state to state. On the low end, in Colorado, at least 22% of non-elderly adults have conditions that would likely be declinable if they were to seek coverage in the individual market under pre-ACA underwriting practices.  Rates are higher in other states – particularly in the South – such as Arkansas (34%), Kentucky (34%), Mississippi (34%), and West Virginia (37%), where at least a third of the non-elderly population would have declinable conditions.

Discussion

Since the effective date of the ACA market changes in January of 2014, people with pre-existing health conditions have not had to worry about their health conditions affecting their access to health insurance or increasing the premiums that they pay. The legislation assures people access to individual market coverage with comprehensive benefits through a variety of changes in their work and life circumstances.  This could change quite quickly if the ACA market protections for people with pre-existing conditions were invalidated.  While many adults with pre-existing conditions have Medicaid or employer coverage that would still provide protection, over a quarter of nonelderly adults have a health condition that would jeopardize their access to non-group coverage without the ACA market protections, potentially affecting almost one-half of non-elderly families in the country.  For these families, an invalidated ACA could fundamentally affect future access to health care.

Methods

To calculate nationwide prevalence rates of declinable health conditions, we reviewed the survey responses of nonelderly adults for all question items shown in Methods Table 1 using the CDC’s 2018 National Health Interview Survey (NHIS).  Approximately 27% of 18-64 year olds, or 54 million nonelderly adults, reported having at least one of these declinable conditions in response to the 2018 survey.  The CDC’s National Center for Health Statistics (NCHS) relies on the medical condition modules of the annual NHIS for many of its core publications on the topic; therefore, we consider this survey to be the most accurate means to estimate both the nationwide rate and weighted population.

Since the NHIS does not include state identifiers nor sufficient sample size for most state-based estimates, we constructed a regression model for the CDC’s 2018 Behavioral Risk Factor Surveillance System (BRFSS) to estimate the prevalence of any of the declinable conditions shown in Methods Table 1 at the state level.  This model relied on three highly significant predictors: (a) respondent age; (b) self-reported fair or poor health status; (c) self-report of any of the overlapping variables shown in the left-hand column of Methods Table 1.  Across the two data sets, the prevalence rate calculated using the analogous questions (i.e. the left-hand column of Methods Table 1) lined up closely, with 21% of 18-64 year old survey respondents reporting at least one of those declinable conditions in the 2018 NHIS and 23% of 18-64 year olds in the 2018 BRFSS.  Applying this prediction model directly to the 2018 BRFSS microdata yielded a nationwide prevalence of any declinable condition of 29%, a near match to the NHIS nationwide estimate of 27%.

In order to align BRFSS to NHIS overall statistics, we then applied a Generalized Regression Estimator (GREG) to scale down the BRFSS microdata’s prevalence rate and population estimate to the equivalent estimates from NHIS, 27% and 54 million.  Since the regression described in the previous paragraph already predicted the prevalence rate of declinable conditions in BRFSS by using survey variables shared across the two datasets, this secondary calibration solely served to produce a more conservative estimate of declinable conditions by calibrating BRFSS estimates to the NHIS.  After applying this calibration, we calculated state-specific prevalence rates and population estimates off of this post-stratified BRFSS sample.

To calculate nationwide prevalence rates of declinable health conditions at the family-level, we imputed person-level presence of any condition onto the NHIS person file using the main 2018 NHIS sample adult estimate.  This model relied on three highly significant predictors: (a) respondent age; (b) self-reported fair or poor health status; (c) self-report of any health-related activity limitation, disability, hospitalization, or high rate of physician visits.  Since all individuals responding to the NHIS sample adult questionnaire also respond to the NHIS person component of the survey, these factors produced a reasonably predictive estimate, matching 27% of non-elderly adults with pre-existing conditions for all individuals participating in the survey.  Unlike the NHIS sample adult file, the NHIS person file allows for analyses of family-wide characteristics.  This prediction yielded 53% of non-elderly adults having a declinable condition themselves or co-habiting with a non-elderly adult family member with a declinable condition; using the NHIS family weights, this results in 45% of non-elderly families (families having at least one non-elderly adult family member) having one or more adults with a declinable condition.  In total, approximately 54 million non-elderly adults may have a pre-existing condition and almost as many non-elderly adults without pre-existing conditions live with a family member that does.

The programming code, written using the statistical computing package R v.3.6.1, is available upon request for people interested in replicating this approach for their own analysis.