Assessing the Impact of the Affordable Care Act on Health Insurance Coverage of People with HIV

APPENDIX A: METHODS

Data source:

This analysis relies on data from the Medical Monitoring Project (MMP), as recommended by the Institute of Medicine in a report commissioned by the White House to help identify data sources for monitoring the effects of the Affordable Care Act (ACA) on HIV care and coverage in the United States.¹ MMP is a supplemental HIV surveillance system designed to produce nationally representative estimates of behavioral and clinical characteristics of HIV-infected adults receiving medical care in the United States.^2,3 MMP is a complex-sample, cross-sectional survey. For the 2009 data collection cycle, first, U.S. states and territories were sampled, then, facilities providing HIV care, and finally adult persons aged 18 years or older receiving at least one medical care visit in participating facilities between January and April 2009. Data were collected via face-to-face interviews and medical record abstractions from June 2009 to May 2010. All sampled states and territories participated in MMP (California, Delaware, Florida, Georgia, Illinois, Indiana, Michigan, Mississippi, New Jersey, New York, North Carolina, Oregon, Pennsylvania, Puerto Rico, Texas, Virginia, and Washington). Of 603 sampled facilities within these states or territories, 461 participated in MMP (facility response rate 76%), and of 9,338 sampled persons, 4,217 completed both an interview and a linked medical record abstraction (adjusted patient-level response rate 51%) for a combined response rate of 39%. Data were weighted based on known probabilities of selection at state or territory, facility, and patient levels. In addition, data were weighted to adjust for non-response using predictors of patient-level response including facility size, race/ethnicity, time since HIV diagnosis, and age group. This analysis includes information on 4,067 participants, who, after weighting for probability of selection and non-response, are estimated to represent a population of 406,970 HIV-infected adults aged 19-64 years receiving medical care in the United States between January and April 2009.

Data for this analysis were collected in 2009 – prior to the passage of the ACA.  They do not, therefore take into account increases in in the number of people living with HIV (estimated to increase by about 3% annually),⁴ and as such likely underestimate the number who may be newly eligible for coverage as of 2014. They also assume that the income and insurance profile of people with HIV remains the same between 2009 and 2013, just before the implementation of the major ACA coverage expansions examined here.

Analysis:

This analysis focuses on individuals between age 19 and 64, the age group targeted for coverage expansions by the ACA.⁵  For all respondents in MMP, we examined self-reported insurance coverage as well as payment source for antiretroviral medicines using responses to the following questions “During the past 12 months, what were all the kinds of health insurance or health coverage you had?” and “During the past 12 months, what were the ways your antiretroviral medicines were paid for?”  Response options included insurance programs (Medicaid, Medicare, private insurance, Veteran’s Administration, Tricare or CHAMPUS coverage, other public insurance, and other unspecified insurance) as well as medical care, medications and other services paid for by the Ryan White HIV/AIDS Program (Ryan White or the AIDS Drug Assistance Program). It is important to note that HIV patients may not be aware of all the services they receive that are paid for by the Ryan White HIV/AIDS Program (the program provides funding directly to service organizations in many cases) and therefore, the estimates of the number of individuals who receive Ryan White HIV/AIDS Program services is likely an underestimate. Because respondents in MMP may indicate more than one type of coverage, we relied on a hierarchy to group people into mutually-exclusive coverage categories. Specifically, the hierarchy groups people into coverage types in the following order:

1. Medicaid coverage, except for those dually eligible for Medicare
2. Dually eligible for Medicaid and Medicare coverage
3. Private coverage
4. Medicare coverage only
5. Other public coverage, including Tricare/CHAMPUS, Veteran’s Administration, or city/county coverage

In most cases, this hierarchy classifies individuals according to the coverage source that serves as their primary payer. The exception is Medicaid, which captures anyone with Medicaid coverage (even if Medicaid is not the primary payer) in order to account for the undercount of Medicaid coverage in population-based surveys.⁶  People who do not report any of the sources of insurance coverage above are classified as uninsured.  As noted above, we separately assess whether each respondent received assistance through the Ryan White HIV/AIDS Program.

Lastly, we examine income for each respondent as a share of the federal poverty level (FPL). MMP captures income in terms of either monthly or annual dollar income, and it also measures how many people are supported by that income. We use these two variables to translate dollar income to corresponding share of FPL, using 2008 HHS Federal Poverty Guidelines for persons interviewed in 2009 and 2009 HHS Federal Poverty Guidelines for persons interviewed in 2010. Because the income measure in MMP is categorical, rather than continuous, we assign each respondent the mid-point of the category for their income. We conducted a sensitivity analysis (explained below) to assess to what extent using the mid-point could lead us to over or under-estimate the number gaining coverage under the ACA. We group individuals into income categories that correspond with income eligibility for coverage under the ACA: less than 100% FPL, between 100 and 138% FPL, between 139 and 399% FPL, and greater than or equal to 400% FPL.

Respondents for whom income or insurance coverage information were missing (3.2% and 0.2% of the unweighted sample, respectively) were excluded. However, in reporting total number of people with HIV in care in each group, we assumed that the insurance and income profile of those with missing data was the same as for those with complete data. In reporting estimates, we exclude point estimates where the relative standard error is greater than 30% or where the unweighted cell size was less than 5. These restrictions affected some of the estimates for the population at or above 400% FPL.

Limitations:

MMP is nationally representative only of those with HIV who are in care.  The survey does not include those who are diagnosed but not in care, nor does it capture people with HIV who are not diagnosed (see Figure 2).  People with HIV who are not in care or who have not been diagnosed also will be affected by ACA coverage provisions and are likely to come into care as a result of outreach efforts. However, we have limited information on the basis of which to extrapolate MMP findings to this population. Although we do not attempt to estimate coverage changes for people with HIV not in care, we do discuss the potential impact of the ACA on the population of all people infected with HIV regardless of diagnosis or care status, using CDC estimates of that population.

MMP also only allows for analysis at the national level.  Because not all states are moving forward with the Medicaid expansion, our national estimates of the impact of ACA coverage provisions will over-estimate the number affected. We conducted a secondary analysis to account for states not expanding Medicaid by scaling the national findings to the share of people in treatment for HIV who live in states that are expanding Medicaid. Specifically, using the NCHHSTP Atlas and the Kaiser Family Foundation’s health reform resources, we calculated the shares of people living with diagnosed HIV in the states expanding and in the states not expanding Medicaid.  We then multiplied these shares by the MMP findings to estimate the number of people in care for HIV who will be eligible for Medicaid coverage and Marketplace subsidies, given the current state decisions regarding Medicaid expansion.  In doing this, we make the assumption that the national income and insurance distribution is equal to the income and insurance distribution in the expansion states as well as the non-expansion states.  Since actual coverage patterns and income distributions are likely to vary across states, the state-level analysis should be interpreted with caution. While imprecise, the state-level analysis still provides a better approximation of the current landscape than using only the national data, given that all states are not expanding Medicaid. Although Puerto Rico is included in our national estimates, we do not account for it when conducting this secondary analysis.  However, the effect is minimal, as Puerto Rico accounts for only 2% of people living with diagnosed HIV in the United States, according to the NCHHSTP Atlas data.

As described above, MMP does not collect data on the actual income amount for each individual but rather categorizes the respondent’s annual income into categories at intervals of $5,000.  We assigned the middle point of the income category to the individual in calculating where they fall with respect to the FPL, which could incorrectly estimate their potential eligibility for Medicaid or Marketplace subsidies. For example, an individual who is just above 138% of the poverty level ($14,945 in 2009) would be classified as at or below 138% FPL using the mid-point of income ($12,250) for their category ($10,000 to $14,999). We conducted sensitivity analysis of our approach by assigning both the minimums and maximums of the income categories in calculating income with respect to the FPL; we then compared how the population redistributed itself using both the minimum and maximum levels. We found that if we had used the lower bound of the income category, we would have estimated the share of the population falling into the Medicaid eligibility group to be 2 percentage points higher than we estimated using the midpoint. Had we used the upper bound, the share of the population falling into the Medicaid eligibility group would be 11 percentage points lower than what we estimated using the midpoint. Thus, while we are likely capturing most people who would fall into the Medicaid eligibility group, we may be capturing some people whose income would in fact be above Medicaid eligibility.  These individuals would still be eligible for new coverage under the ACA via Marketplace subsidies.

We are also unable to identify which individuals are undocumented residents in the US and consequently ineligible for Medicaid or enrollment in the Health Insurance Marketplaces in 2014. In 2009, 12.3 percent of respondents in MMP were not born in the US, though many of these individuals are likely lawfully present (or subsequently gained citizenship) and would be eligible for coverage. Lawfully present noncitizens who have been in the country for less than five years are also barred from Medicaid coverage.  Of the foreign born in MMP, only 5 percent have been in the country for less than five years, and these individuals account for less than one percent of the total sample. Thus, inclusion of foreign-born individuals is unlikely to substantially affect our estimates of eligibility for Medicaid coverage.

Our estimates do not account for participation rates in coverage. Not everyone who is newly-eligible for Medicaid or Marketplace subsidies will enroll in coverage, and some people who are currently uninsured may already be eligible for Medicaid or other coverage.  Uptake is dependent on several factors, including ease of applying, knowledge of coverage options, other coverage options available, personal utility of insurance coverage, and a host of other factors.  Other analyses that have used micro-simulation models to predict coverage changes under the ACA estimate that approximately 60% of newly-eligible individuals will enroll in Medicaid;¹⁰ given the high need for medical services among the population with HIV, it is likely that uptake among this group would be much higher.