Introduction

It has been over 20 years since disabled beneficiaries first appeared on the Social Security Disability Insurance (DI) and Supplemental Security Income (SSI) rolls as a result of Acquired Immunodeficiency Syndrome (AIDS). Since the first several hundred of such beneficiaries began receiving payments in 1982, the Social Security Administration has monitored the epidemic's impact on expenditures from the DI Trust Fund as well as expenditures of Federal general revenue under the SSI program.

This monitoring process has involved periodic examinations of agency administrative records for cases on the DI or SSI program rolls that could be identified as receiving disability benefits, at least in part, due to infection with the Human Immunodeficiency Virus (HIV). The results of these analyses were used to produce rough estimates of historical and projected program expenditures due to HIV. Modeling the impact of HIV on program expenditures required an understanding of the circumstances causing HIV-infected persons to apply for benefits, the factors involved in favorable disability determinations, and the reasons that HIV beneficiaries leave the rolls.

The primary reason for termination of benefits payable to HIV beneficiaries has been death, so a careful examination of mortality experience is clearly a key to understanding duration on the rolls. Prior evaluations of HIV mortality had been done on a rough aggregate basis. The purpose of this note is to present the detailed results of the first person-based mortality study of the extensive data that we have collected. The current study is based on a total of roughly 1.4 million life-years of exposure distributed over two 5-year periods: 1992-96 and 1997-2001, and represents the most extensive case study of HIV mortality produced from administrative records.

History of the HIV data collection project

As indicated, this study represents the culmination of two decades of efforts to track the impact of HIV on the cash benefit programs administered by the Social Security Administration (SSA). The agency's collection of administrative data that provide the basis for this study was developed over time as the world's scientific community gained a greater understanding of the effects of HIV infection on diverse body systems. Those findings have, in turn, come to affect SSA's evaluation of claims for disability benefits based on the presence of the infection. To fully appreciate the value of the current study, as well as its limitations, it is important to have some understanding of how SSA's collection of HIV data has evolved over the past 20 years.

In the early years of the HIV epidemic, infected persons who presented themselves as claimants for DI or SSI benefits, most often did so as a result of one or more diseases, such as Kaposi's sarcoma and pneumocystis carinii pneumonia. These particular diseases appeared to be present primarily due to an impaired immune system which resulted from the presence of an HIV infection. Such claims were tentatively identified as being associated with HIV, and the claims folders were forwarded to SSA headquarters for confirmation. The first data collection on HIV claimants was the result of this manual examination of claims folders.

As the HIV epidemic spread, and the number of disability claims grew correspondingly, SSA could no longer manage the "hand-selection" of cases for its compilation of HIV data. As a proxy for the manual identification process, the Office of the Chief Actuary compiled a list of certain impairment codes¹ commonly assigned to HIV cases, which was then used to electronically retrieve the corresponding administrative records from SSA's two main payment record files—the Master Beneficiary Record (MBR) for DI and the Supplemental Security Record (SSR) for SSI.

However, since no single impairment code yet existed for HIV, and since the assignment of impairment codes can be subject to some misclassification in the claims taking process, the selection of cases included in SSA's compilation of HIV data was exposed to both type I and type II error.² Nevertheless, the overall characteristics of the cases identified seemed consistent with the cases that had been identified in the earlier manual folder evaluation, providing some assurance that the cases under study were to a large degree correctly identified.

When the World Health Organization introduced ICD series 042.0-044.9 specifically for HIV-related diseases³, SSA followed with a corresponding series of impairment codes. At that time all HIV-related impairments were to be assigned a diagnosis code in that range. As a result, the collection of data shifted to identifying new cases using the now-familiar HIV diagnosis codes of 042, 043, and 044, while continuing to track the experience of HIV cases that had been identified under prior evaluations.

Soon after, some information was added to the data collection process from the agency's transactions file for disability claims processed by the State Disability Determinations Services—known as the 831 file. In addition to the primary diagnosis code—which is carried on the MBR and SSR—the 831 file also contains a secondary diagnosis code. This code allowed us to identify additional cases involving HIV as a contributing (if not the primary) reason for impairment.

Since 1990, our process for collecting HIV data has remained relatively unchanged. Administrative data on previously identified cases is updated every six months, in June and December. At the same time, newly-identified HIV beneficiary cases are added to the data collection.

SSA's evaluation of HIV-related impairments

In addition to changes in the way we have collected data on HIV cases, the standards used to evaluate potential HIV disability claims have also been modified over time as medical knowledge of HIV has progressed. While SSA has tried to remain consistent with the criteria and rationale of the Centers for Disease Control and Prevention (CDC) and World Health Organization, the agency has found it necessary to define its own guidelines for evaluating HIV-related impairments from the perspective of administering the DI and SSI programs. The SSA definition and medical listings with regard to HIV have evolved from the earliest definitions provided by CDC when the disease was first listed under infectious/parasitic category, to the time when HIV received its own regulatory listing in the Listing of Impairments⁴ in June 1993.

As the spectrum of manifestations of HIV-infection became better defined, it became apparent that some progressive and seriously disabling conditions were not included in the definition of AIDS. In addition, clinicians began to identify a group of individuals with a variety of signs and symptoms which were thought to be caused by the HIV virus. The collection of these symptoms came to be known as AIDS-Related Complex (ARC). Effective September 1987, the CDC began using a revised criteria for determining which cases involving HIV should be reported for AIDS statistical surveillance purposes. The revision expanded the definition to include HIV Dementia and HIV Wasting. To remain consistent, SSA revised its definition for AIDS, but discontinued the automatic link with the CDC definition. The revised criteria lead to reclassifying as AIDS a number of then-current beneficiaries who were on the disability rolls due to ARC, rather than a definitively diagnosed case of AIDS. During this time, the number of AIDS-related impairments on the DI rolls grew rapidly from 5,700 at the end of 1986 to 17,400 by the end of 1988.

In 1990, SSA issued new guidelines to include criteria for symptomatic HIV which is not AIDS, including all lymphomas and other disease manifestations coupled with 200 CD4 count and marked functional limitations. The expanded guidelines resulted in a new class of HIV impairments, and by 1992, the DI program had experienced its most critical year for new HIV entitlements as an additional 33,000 workers began receiving payments. However, the death rate among those afflicted was so high that monthly benefit payments were often made only for a relatively short period of time, if at all. Many never received payments, failing to survive the requisite 5-month waiting period under the DI program. One-third of those who made it onto the rolls because of HIV impairments had died by the end of the calendar year in which they became entitled; two-thirds had died by the end of the following year.

By 1996, CDC began reporting sharp declines in AIDS incidence as a result of public awareness and widespread use of highly-active antiretroviral therapy (HAART) which slows the reproduction of the virus and thus the progression of HIV infection to AIDS.⁵ By 1998, CDC reported a leveling of incidence and essentially no change from 1999 to 2001 with roughly 41,000 new AIDS cases reported each year over that period. DI entitlements followed the decline in population incidence as infected workers remained employed for longer periods of time. Since peaking in 1992, the number of workers becoming entitled to DI benefits based on HIV has fallen in recent years to roughly 10,000 annually.

Mortality 1997-2001

As of December 31, 2002, the CDC reported nearly 385,000 persons in the U.S. living with AIDS, and an estimated 41,000 new diagnoses of AIDS infections occurring annually. From the beginning of the epidemic through 2002, CDC estimates that roughly 502,000 persons have died with AIDS in the U.S.⁶

Social Security records on HIV impairments show that by the end of 2002, there were nearly 98,500 workers receiving DI benefits—with approximately 10,500 becoming newly entitled in that year⁷—and roughly 72,000 individuals receiving SSI benefits.⁸ Over the 10-year period covered by this study (1992-2001), approximately 204,100 HIV beneficiaries were terminated from the DI and SSI rolls as a result of death.

Results found in this study reflect DI (worker only) and SSI disability experience. We caution against viewing these results as a proxy for the HIV/AIDS mortality of the population being monitored by CDC. Only a fraction of those diagnosed with the HIV infection or full-blown AIDS actually become eligible for DI or SSI benefits. Many remain in the work force for extended periods of time, delaying, perhaps indefinitely, pursuit of disability benefits. For those who attain beneficiary status—and subsequently come under observation—the infection has progressed to the point of being disabling. Consequently, the make-up of the HIV population that is the basis for this study is quite different from the overall AIDS population tracked by CDC. Furthermore, the primary variable of interest for this mortality study is duration since becoming entitled to disability benefits. The time of initial diagnosis of HIV may not be known, and therefore the amount of time from onset of the infection to death cannot be measured.

As mentioned earlier, the expansion of the guidelines in 1990 more clearly defined the aspects of disability for individuals who had HIV, but had not progressed to the point of having AIDS or showed other severe manifestations of the infection.⁹ For purposes of this study, HIV infection must be present and would be a contributing factor (if not the primary reason) for a favorable disability determination.

The main section of this study presents HIV mortality experience for the DI and SSI rolls over the period 1997-2001. For comparison, similar actuarial tables are provided in appendix A for 1992-96. Dramatic differences in mortality exist between the two 5-year periods. Among male beneficiaries in early durations, HIV mortality for 1997-2001 ranges from 30-50 percent of HIV mortality for 1992-96. Mortality improvements in later durations are less dramatic, but still significant, ranging from 50-75 percent of the earlier period. A similar comparison for females shows mortality for 1997-2001 is 40-60 percent of the previous period in early durations, and 60-80 percent in later durations.

Much of the decline in mortality experience among DI and SSI recipients appears to be consistent with the wider use of HAART. To a lesser extent, the improvement may also be the result of differences in the composition of the rolls between the two periods. Over the period 1992-96, the number of individuals that appeared on the disability rolls with HIV as the primary reason for impairment was greater than those for whom HIV was the secondary reason by a ratio of 10-to-1. During 1997-2001, this ratio had dropped to less than 7-to-1. Furthermore, the ratio of symptomatic HIV cases to asymptomatic HIV cases¹⁰ fell from roughly 6-to-1 in the earlier period, to 4-to-1 in the later period. Data suggest that cases having HIV impairment as the secondary diagnosis rather than the primary diagnosis, or an asymptomatic HIV diagnosis rather than symptomatic HIV diagnosis, exhibit inherently lower mortality. Consequently, a relatively higher concentration of lower-mortality individuals on the rolls contributes to lower overall mortality.

Tables 1A and 1B show select-and-ultimate probabilities of death for male and female HIV disabled beneficiaries, by select age (that is, age at entitlement/eligibility to disability benefits) and duration since selection. Data reflect the combined actual experience of the DI and SSI rolls from January 1, 1997 through December 31, 2001. The probability of death among HIV beneficiaries is generally highest within the first several durations, then typically decreases in later durations. In comparing HIV mortality to general disability mortality, we see several common characteristics found in the overall disability population do not necessarily hold for HIV experience.¹¹ Deviations may be due to uncertainty over the length of time individuals have been infected, and the degree of therapy received. Certain characteristics generally seen in mortality patterns in the overall disability population appear to be less influential in the HIV disability population. These include:

Over the period 1997-2001, male HIV mortality ranges from one-and-a-half to three times that of overall disability mortality for early durations, and up to four times as high for later durations. Female HIV mortality can be three to six times as high as overall disability mortality over the same period. Several unique circumstances were encountered in the data that affect mortality estimates. These include death within the disability waiting period, and cases where HIV is present but is not material to benefit allowance. These and other data considerations are discussed in appendix B.

A survival table is a concise way of representing the probabilities of a particular population living to a particular age. Tables 2A and 2B show the progression of a series of cohorts—each for a given select age—reflecting the probabilities of death shown in tables 1A and 1B. See appendix B for details on table construction and usage.

Tables 3A and 3B show the expected future lifetime of male and female HIV disabled beneficiaries. Females have a higher future lifetime than males. As with general disability mortality, HIV beneficiaries often exhibit a shorter life expectancy in the first several years of entitlement than in later durations. This is due to higher mortality in those years.

Tables 4 and 5 show aggregate probability of death and expected future lifetime, by select and attained ages. Probabilities are exposure-weighted averages of those found in tables 1A and 1B. They represent the average probability of death, within one year, for those originally entitled to disability benefits at a particular select age (table 4), or those entitled to disability benefits who have attained a particular age (table 5). Similarly, aggregate future lifetime represents the life expectancy for those of a particular select or attained age. These values are exposure-weighted averages of the select-and-ultimate future lifetimes shown in tables 3A and 3B.

Table 6 shows aggregate results based on years since selection, or duration. Probabilities are based on aggregate counts of exposure and deaths across all select ages, and represent the average probability of death within the next year of entitlement to disability benefits. Aggregate future lifetime represents the average life expectancy for all those who have been entitled to disability benefits for the stated number of years.

Tables 7A and 7B show select-and-ultimate probabilities of death for HIV disabled beneficiaries by select age and duration. These tables are similar to tables 1A and 1B, however they reflect only the experience of the DI rolls. As such, beneficiaries may be concurrently eligible for DI and SSI benefits, but those eligible for SSI only are not considered.

When comparing tables 1A-1B to tables 7A-7B, we observe that the mortality of the combined DI and SSI rolls is higher than that of the DI rolls in roughly two-thirds of all attained ages; this proportion increases to roughly three-fourths when considering only durations 5 and later. This indicates somewhat higher mortality among those receiving only SSI benefits (whose experience is included in tables 1A-1B, but not in tables 7A-7B) than among DI beneficiaries, particularly for older attained ages. Reasons for higher SSI mortality may be related to the means-tested nature of the SSI program itself. By definition, SSI-only recipients are of lesser economic means and typically uninsured for DI benefits and, hence, ineligible for any Medicare benefits that are available under the DI program. Generally, SSI recipients are categorically eligible for Medicaid. However, prior to SSI eligibility, medical assisitance through Federal or State sponsored programs may not have been available. The lack of medical treatment—both prior to and during eligibility—may contribute to higher mortality rates.

Similar tables for combined experience from January 1, 1992 through December 31, 1996 are presented in appendix A.

Notes:
1. Select age denotes age last birthday at entitlement to disability benefits. Duration measured in years since selection. Attained age calculated as sum of select age and duration. Results do not include auxiliary beneficiaries payable under the DI program.
2. The value q_[x]+t at duration t represents the probability of death—in a multiple-decrement environment—during the (t+1) year of entitlement for those originally entitled to disability benefits at select age [x] who have attained age [x]+t.
3. Select-and-ultimate table is read across the row for 0-10 years since selection, and down the last (ultimate) column for 10 or more years since selection.
4. Results have been graduated using the Whittaker-Henderson Type B two-dimensional method.

Notes:
1. Select age denotes age last birthday at entitlement to disability benefits. Duration measured in years since selection. Attained age calculated as sum of select age and duration. Results do not include auxiliary beneficiaries payable under the DI program.
2. The value q_[x]+t at duration t represents the probability of death—in a multiple-decrement environment—during the (t+1) year of entitlement for those originally entitled to disability benefits at select age [x] who have attained age [x]+t.
3. Select-and-ultimate table is read across the row for 0-10 years since selection, and down the last (ultimate) column for 10 or more years since selection.
4. Results have been graduated using the Whittaker-Henderson Type B two-dimensional graduation method.

Notes:
1. Select age denotes age last birthday at entitlement to disability benefits. Duration measured in years since selection. Attained age calculated as sum of select age and duration. Results do not include auxiliary beneficiaries payable under the DI program.
2. The value l_[x] at duration 0 represents the assumed number of lives originally entitled to disability benefits at select age [x]; the value l_[x]+t at duration
t > 0 represents the number of lives remaining from the original l_[x] who have attained age [x]+t. Lives are decremented using probabilities from table 1A.
3. Select-and-ultimate table is read across the row for 0-10 years since selection, and down the last (ultimate) column for 10 or more years since selection.