Managing Medicare’s HIV Caseload in the Era of Suppressive Therapy

Abstract

Objectives. The 1996 introduction of antiretroviral medications changed Medicare’s role in providing HIV care. We analyzed Medicare’s patient database in an effort to document the new HIV therapies’ effects on expenditures and outcomes.

Methods. We examined the medical billing records of a 5% national Medicare sample from 1997 through 2003. The cohort was stratified by year and categorized by age, race/ethnicity, gender, and Medicare status. Population summaries were categorized according to presence of major chronic diseases and HIV-related conditions.

Results. The number of Medicare beneficiaries with HIV increased from 42520 in 1997 to 76500 in 2003, whereas mortality among this group fell by 35%. HIV-associated infections declined by as much as 43% (mycoses). Heart and liver disease and diabetes increased by more than 50%. Adjusted annual per person Medicare expenditures fell 28%; expenditures were 49% higher for Blacks than for Whites.

Conclusions. Improved HIV medical management has led to fewer deaths and has shifted treatment toward chronic care. However, successful management is complicated by conditions that have not been historically linked to HIV and whose effects vary according to race/ethnicity.

Between 1995 and 2004, the number of Americans living with AIDS grew from 217 000 to 415 000.1^,2 Deaths from AIDS among US residents fell by two thirds over the same period, but AIDS incidence rates dropped by only one third. These changes are primarily attributable to the introduction of highly active antiretroviral therapy (HAART) in 1996. HAART has reduced disease progression and prolonged life, leaving patients in need of ongoing disease management.

This transformation has been especially significant for HIV-positive individuals enrolled in Medicare, which covers persons with disabilities in addition to the elderly. Among those reaching the clinical (i.e., physically disabling) stage of AIDS, median survival rates were 20 months or less through the early 1990s,3^,4 and this low life expectancy greatly restricted the number of individuals who qualified for Medicare coverage. HAART induced a sudden change in this picture, even among patients with the most advanced disease. The death rates of people with advanced HIV disease fell 62% after starting the new antiretroviral therapy.5 Not only was the number of people with AIDS increasing, but they were surviving longer. As a result, there was the potential for significant increases in numbers of HIV-positive Medicare beneficiaries as well as increases in the expenditures associated with their care.

From 1991 to 1996, Medicare spent $3.8 billion on approximately 95 000 individuals with HIV.6 Half of this population was still alive at the end of 1996. Medicare’s claim records document the experiences of these patients between 1997 and 2003 as well as the experiences of patients new to the system. In the United States, with its multitude of separate health care systems, the extensive Medicare data available allowed a rare opportunity for us to make a detailed national analysis of trends in HIV prevalence rates, comorbidities, mortality rates, and expenditures in the HAART era.

METHODS

Study Population

There are several different categories of Medicare beneficiaries with HIV. One group is composed of people disabled as a result of HIV, that is, individuals who advanced to AIDS before or despite the availability of HAART. Five months after being classified as having a long-term disability, qualified individuals can apply for Social Security disability benefits. Twenty-four months later, they are eligible for Medicare. Another group is made up of Medicare beneficiaries who are disabled as a result of a condition other than HIV and who are coincidentally HIV positive. Yet another group is elderly individuals with HIV who may or may not have advanced to AIDS. A 1996 study showed that, overall, only 56% of Medicare’s HIV-positive beneficiaries (about 44 000 patients) had progressed to AIDS as of that year.7

Data Collection

Our source data were derived from Medicare’s yearly nationwide sample containing claims records for 5% of the program’s beneficiaries. We used the records for the years 1997 through 2003 to evaluate expenditures and disease frequencies among Medicare beneficiaries receiving treatment for HIV and its associated conditions. We based our inclusion of patients on the repeated appearance in their claim history of International Classification of Diseases, Ninth Revision (ICD-9),8 treatment diagnosis code 042, which covers symptomatic HIV infection. (A treatment diagnosis, as represented by the appropriate ICD-9 code, is the diagnosis entered by a service provider to justify a line-item charge for treatment.) Only diagnoses linked directly to a hospital or physician service were used in identifying patients.

In the case of each annual Medicare 5% sample, we isolated patients with HIV (ICD-9 code 042) diagnoses on their claim forms. These annual records were then integrated into longitudinal diagnosis and treatment records for each patient. (There is a loss of about 0.5% each year from the panel sample as a result of changes in beneficiaries’ account numbers, which are based on heads of households’ Social Security numbers.) The cohort was stratified by year and categorized according to age, race/ethnicity, gender, and Medicare status. Population summaries were further stratified by major chronic disease status and presence of HIV-related conditions. We multiplied our results by a factor of 20 so that we could present estimates for the overall Medicare population from the 5% sample.

We focused on patients who received at least 2 HIV-related hospital or physician services in a given calendar year, designating them in our data with a “confirmed HIV” flag. We chose this moderately restrictive case definition to minimize the number of unidentified cases while achieving a high degree of year-to-year continuity in our study population. At least some level of error is involved in using diagnosis codes derived from claims data to identify study populations, because of the inclusion of erroneous or misreported diagnoses. Algorithms stringently designed to filter these false positives also lead to the exclusion of potentially true cases. Along with these technical considerations, there are clinical ones as well. A patient’s care may vary from year to year, sometimes falling to a level considered negative according to the case definition algorithm. We wanted to ensure that most patients with confirmed HIV cases maintained that designation even in years when their care was sporadic.

We did not divide the HIV-diagnosed Medicare population according to AIDS status. Conceptually, such distinctions between disease stages are now difficult to make given HAART’s impact on immune recovery. The Medicare program continues to include a large number of beneficiaries with long-term, disabling HIV. Although still suffering from a chronic, serious illness, many of these people have now recovered their health at least to the extent that they no longer meet the Centers for Disease Control and Prevention’s criteria for officially reportable “AIDS” (these criteria include CD4+ T-cell counts below 200 cells/mm³, which would not appear in Medicare claims records, and such recordable diagnoses as pneumocystis and cytomegalovirus infections). We considered as equivalent all Medicare beneficiaries meeting our “confirmed HIV” definition. We did not distinguish between those who had been officially diagnosed with AIDS and those who had originally qualified for Medicare for reasons unrelated to HIV.

We did not have records of outpatient pharmacy drug payments because Medicare did not cover these expenditures during our study period. However, patients are required to visit physicians to obtain prescriptions and to monitor drug effects. Patients who had undergone successful antiretroviral therapy appeared in our records because they had visited physicians or outpatient clinics for reasons related to HIV even though they did not have serious HIV-associated conditions.

Most patients were enrolled in the standard fee-for-service Medicare system in which care providers seek payment for the specific services they performed. During part or all of the study period, however, some patients were enrolled in Medicare health maintenance organizations (HMOs) or had only partial Medicare coverage (Medicare part A or part B only). We excluded data on months of HMO or partial enrollment because we could not obtain complete diagnosis or payment records for those periods. These months amounted to only 7.7% of our population’s total Medicare enrollment time.

Statistical Analysis

To analyze the evolution of Medicare expenditures, we examined annual per person payments among disabled beneficiaries younger than 65 years after adjusting for inflation. Medicare payments are regulated by government fee schedules and do not follow typical health care inflation patterns. In fact, the Balanced Budget Act of 1997 (HR 2015) imposed Medicare fee reductions: in 1998, limitations were imposed on home health benefits; in 1999, limitations on skilled nursing facility expenditures were introduced; and, in 2000, limitations on outpatient expenditures were implemented and new home health payment rules were formalized. Between 1997 and 2000, increases in hospital reimbursements were limited to levels at or below the overall consumer price index.

To compensate for changes in Medicare expenditures, we calculated the percentage difference between 1997 and each subsequent study year in average annual payments disbursed to all disabled Medicare beneficiaries. By applying the resulting adjustment factors to annual payments within the Medicare HIV cohort, we were able to convert payments to constant 1997 dollars.

We then developed a multivariate regression model using a log + 1 transformation of adjusted annual payments. The model controlled for gender, age, race/ethnicity, county of residence (urban vs rural), US region of residence, partial year eligibility status, mortality, and the interaction between mortality rate and months of eligibility within a given year. We used the parameter estimates associated with each year to estimate the underlying proportional change in payments for that year in comparison with 1997.

In addition, we constructed a classification system sensitive to both disease and Medicare administrative status to track trends by HIV subpopulations. This taxonomy included 3 categories. The first was “preexisting HIV.” Included here were individuals who were enrolled in Medicare and had confirmed HIV in the study year in question and the 1 preceding it. The second category was “incident HIV.” These beneficiaries had Medicare eligibility in the year in question and the preceding year but confirmed HIV status only in the former. The third category was made up of “newly eligible” beneficiaries. These individuals were in their first year of Medicare coverage and also met the confirmed HIV definition. We were not able to ascertain their HIV history because data on their status over the preceding years were not included in the Medicare records.

This approach allowed us to distinguish between established Medicare beneficiaries with newly identified HIV and beneficiaries who were already receiving treatment for HIV during the year they entered Medicare. The latter status is a strong indication that disability among these beneficiaries was attributable to AIDS. Our taxonomy supported categorizations from 1998 onward. We could not categorize beneficiaries in 1997, because we did not have the necessary 1996 records.

Finally, we analyzed ICD-9 diagnostic codes to identify HIV-related conditions, including conditions for which the HIV-diagnosed Medicare population had an odds ratio at least 3 times higher than the general Medicare disabled population (after adjustment for age and gender) and conditions historically associated with HIV, such as opportunistic infections and wasting. We consolidated these codes into the following broad categories: immunological (including HIV and lymphoma), diverse infections, liver, renal, heart and vascular, gynecological, pregnancy, metabolic, adverse effects, and substance abuse. We also used specific filters for patients with diabetic, renal, hepatic, and cardiovascular conditions, which allowed us to analyze these patients as separate subgroups. Presence of a condition was determined annually on the basis of at least 1 physician or hospital treatment diagnosis not linked to a laboratory or other diagnostic test.

RESULTS

Demographic Characteristics

The calendar year 2003 demographic characteristics of the study population are shown in Table 1 ▶. Most notably, men were predominant, and Blacks were overrepresented relative to the disabled Medicare population as a whole. Over the 7-year study period, HIV-treated beneficiaries entitled to Medicare as a result of disability (including end-stage renal disease) accounted for 94% of the total number of months of Medicare eligibility. The elderly made up only a small portion of the confirmed HIV population (5580 in 2003, or 7.3% of all HIV-treated beneficiaries).

TABLE 1—

Demographic Characteristics of Medicare Beneficiaries With HIV and Comparisons With Medicare Disabled Beneficiaries and Elderly Populations: United States, 2003

		Disabled Beneficiaries				Elderly Beneficiaries
Characteristic	Beneficiaries With HIV	HIV	Non-HIV	χ2	P	HIV	Non-HIV	χ2	P
Male, %	80.2	80.6	54.1	993.9	< .001	74.8	41.4	129.6	< .001
Age, y, %				845.9	< .001			178.3	< .001
< 45	49.1	53.0	30.4
45–64	43.9	47.0	69.6			5.3a	0.8a
65–74	5.9					79.8	51.9
≥ 75	1.1					14.9	47.3
Race/ethnicity, %				947.7	< .001			539.3	< .001
Black	38.0	37.6	19.4			43.3	7.9
Hispanic	7.1	7.1	3.4			7.1	1.8
White	52.1	52.5	73.5			47.9	87.5
Other/Unknown	2.7	2.8	3.7			1.8	2.8
Urban county of residence, %	86.3	86.3	68.7	502.6	< .001	86.5	71.2	32.1	< .001
Mortality by age group, y, %				164.5	< .001			5.1	.025
< 45	5.3	5.3	1.1
45–64	6.1	6.1	2.9			6.7	2.1
65–74	7.6					7.6	2.3
≥ 75	11.9					11.9	8.4
Overall	5.9	5.7	2.4			8.2	5.2

^aElderly individuals can obtain Medicare coverage at 62 years of age.

Despite the fact that HAART had been in existence for 7 years, mortality in 2003 remained significantly higher among beneficiaries with confirmed HIV than among the non-HIV population for virtually every major age category. Mortality among disabled beneficiaries with confirmed HIV was 2.4 times higher than in the overall disabled Medicare population (P≤ .001) (Table 1 ▶).

New Enrollments

A total of 5.4 million disabled individuals had complete Medicare coverage (parts A and B) in 2003,9 up 26% from 4.3 million in 1997.10 From 1997 through 2003, the HIV prevalence rate among disabled beneficiaries increased 40% as a result of the accumulating number of preexisting HIV cases (Table 2 ▶). There was a 47% decline in the percentage of HIV cases among disabled individuals newly eligible for coverage (the category including individuals entering Medicare with an HIV-related disability).

TABLE 2—

Confirmed HIV Prevalence and Incidence Rates per 100000 Medicare Beneficiaries: United States, 1997–2003

HIV Population	1997	1998	1999	2000	2001	2002	2003
Disabled	939	1121	1206	1236	1251	1284	1312
Preexisting HIV	. . .a	763	924	985	1009	1037	1080
Incident HIV	. . .a	191	177	161	152	156	148
Newly eligible	1530	1675	1122	948	933	916	814
Elderly	8.8	9.2	10.3	13.0	15.9	17.8	18.8

Note. Beneficiary records for the HIV population contained 2 or more HIV-related physician or hospital nondiagnostic interventions in a given year.

^aPreceding year data required for classification.

Medicare’s overall HIV caseload grew 80% over the study period, with the increase much more pronounced among women (129%) than among men (71%). The HIV prevalence rate among elderly beneficiaries initially was extremely low. It more than doubled from 1997 to 2003.

The percentage of new cases was higher among Blacks than among Whites. In 2003, rates of preexisting HIV were 73% among disabled Black beneficiaries and 81% among disabled White beneficiaries. Also in 2003, 8.4% and 5.0% of Black and White beneficiaries with HIV, respectively, were newly eligible for Medicare coverage.

Annual Expenditures

Table 3 ▶ shows how the size of and expenditures associated with the HIV-diagnosed Medicare population grew over the 7-year study period. Mortality rates declined continually over this period, whereas numbers of hospitalizations declined and then leveled off during 2000 through 2002. Total 7-year Medicare expenditures amounted to $5.97 billion for the study population. Payments for acute hospitalizations amounted to $2.93 billion, 49% of all Medicare payments.

TABLE 3—

Selected Characteristics of and Expenditures Related to Medicare Confirmed HIV Population: United States, 1997–2003

	1997	1998	1999	2000	2001	2002	2003
Overall HIV population
Months of enrollment × 103	453.3	555.0	625.6	674.7	730.8	795.2	863.6
Beneficiary count	42 520	51 400	56 500	60 440	64 960	70 800	76 500
Hospitalizations per 1000 beneficiaries	801	759	725	684	692	669	631
Mortality rate per 1000 beneficiaries	90.3	87.9	83.5	78.8	69.0	65.5	58.8
Total payments, $, × 106	608.0	706.0	761.4	787.6	909.3	984.0	1 218
Annual per beneficiary expenditure, 1997 $a	14 299	13 835	13 535	12 646	12 501	11 718	12 725
Disabled HIV population < 65 y
Months of enrollment × 103	427.0	527.4	594.9	636.0	680.2	738.8	801.5
Beneficiary count	40 000	48 840	53 640	56 820	60 420	65 580	70 860
Total payments, $, × 106	569.9	657.1	718.0	736.5	841	891	1 125
Annual per beneficiary expenditure, 1997 $a	14 247	13 551	13 445	12 575	12 435	11 456	12 697
Whites	11 198	10 691	10 901	9 601	9 529	8 662	9 342
Blacks	18 811	18 002	17 797	17 436	16 904	15 785	17 690

Note. Beneficiary records for the HIV population contained 2 or more HIV-related physician or hospital nondiagnostic interventions in a given year.

^a The adjustment to 1997 dollars was based on the yearly change in average annual per beneficiary payments observed in the complete Medicare disabled population.

In terms of 1997 dollars, mean annual payments for beneficiaries with confirmed HIV declined only slightly during this period. In contrast, our multivariate analysis focusing on disabled beneficiaries with confirmed HIV revealed significant reductions for each year relative to 1997 (Table 4 ▶). However, as with hospitalizations, reductions in annual payments plateaued after 1999, leaving mean annual payments about 25% lower than in 1997.

TABLE 4—

Multivariate Analysis of Annual Medicare Expenditures Among Disabled Beneficiaries With HIV: United States, 1997–2003

Variable	Parameter Estimate (SE)	t	Prob(t)
Intercept	6.43383 (0.09099)	70.71	<.0001
Male (female)	−0.05258 (0.03001)	−1.75	.0798
Age, y (<35)
35–44	0.03717 (0.03289)	1.13	.2585
45–54	0.07643 (0.03561)	2.15	.0318
55–64	0.19429 (0.04900)	3.96	<.0001
Race/ethnicity (White)
Black	0.40077 (0.02501)	16.02	<.0001
Hispanic	0.13844 (0.04562)	3.03	.0024
Other/unknown	0.02818 (0.06187)	0.46	.6488
Urban county of residence (rural/unknown)	−0.05893 (0.03468)	−1.70	.0892
Region (Northeast)
Midwest	−0.27929 (0.03865)	−7.23	<.0001
South	−0.15842 (0.02920)	−5.42	<.0001
West	−0.15549 (0.03357)	−4.63	<.0001
Outlying area/unknown	−0.70765 (0.15039)	−4.71	<.0001
Partial year eligibility adjustment factora	0.18004 (0.00654)	27.54	<.0001
Died during year (did not die during year)	3.00619 (0.10562)	28.46	<.0001
Mortality rate × months of eligibility interactionb	−0.06795 (0.01399)	−4.86	<.0001
Year of expenditures (1997)
1998	−0.12763 (0.04714)	−2.71	.0068
1999	−0.15066 (0.04631)	−3.25	.0011
2000	−0.28499 (0.04585)	−6.22	<.0001
2001	−0.29990 (0.04542)	−6.60	<.0001
2002	−0.33083 (0.04483)	−7.38	<.0001
2003	−0.32346 (0.04435)	−7.29	<.0001

Note. Analyses were based on log + 1 transformed annual Medicare expenditures adjusted to 1997 dollars. The reference category for each variable is in parentheses.

^aAmong beneficiaries with fewer than 12 months of eligibility in a given year, the change in annual expenditures for each additional Medicare month.

^bAdditional adjustment factor per added month of eligibility for beneficiaries who died in a given year.

Death was the greatest elevator of annual expenditures. For example, among individuals with 6 months of eligibility in a given year, expenditures for those who died were approximately 14 times higher than expenditures for those who did not die. Other significant contributors to expenditures included region (expenditures were higher in the Northeast than in other regions) and age (expenditures were higher among those older than 45 years). No significant differences were observed between expenditures for men and women (P = .08).

In a separate model that excluded 1997 observations and included our taxonomic prevalence and incidence categories (data not shown), expenditures for beneficiaries with incident HIV cases were found to be about 14% lower than expenditures for beneficiaries with preexisting HIV cases. No significant difference was found in expenditures between those with HIV who were newly eligible for coverage and those with preexisting HIV.

Expenditures for disabled Black beneficiaries were notably higher than expenditures for disabled White beneficiaries (Table 3 ▶). Our multivariate analysis showed that, across the study period, expenditures for disabled Blacks with confirmed HIV were 49.3% higher than those for disabled Whites with confirmed HIV (P< .001). Expenditures for Hispanics (whose numbers were lower) were 14.9% higher than those for Whites (P= .002). Unlike Hispanics, Blacks also had consistently higher mortality than Whites: 2003 rates were 6.9% and 4.8% among disabled Black and White beneficiaries, respectively.

HIV-Related Comorbidity Rates

Table 5 ▶ presents 7-year trends for the diagnosis categories identified as significantly correlated or of special concern. Rates for many HIV-related conditions declined through 2000, after which they subsequently slowed or halted. However, heart disease, diabetes, and liver disease prevalence rates all rose substantially (by more than 50%) during the 7-year period. (Including both treatment and secondary diagnoses in this analysis increased the reported prevalence rates considerably but did not change the trend patterns.)

TABLE 5—

Trends in Rates of HIV-Associated Conditions Among the Medicare Population: United States, 1997–2003

Condition Category	1997, %	1998, %	1999, %	2000, %	2001, %	2002, %	2003, %
Diverse infections	34.6	34.1	30.8	27.0	26.6	24.8	23.0
Neurological	14.9	15.1	14.4	12.2	13.9	13.1	11.9
Metabolic	13.9	14.7	15.0	13.5	12.1	12.3	13.2
Hepatic	5.6	7.2	7.6	8.9	10.0	10.5	10.5
Diabetes	6.0	5.5	6.5	6.9	8.1	9.6	9.1
Substance abuse	10.2	9.4	9.6	7.4	7.2	6.7	7.2
Renal	6.3	7.3	7.0	7.3	6.9	7.5	7.6
Cardiac	4.9	5.1	6.8	6.6	6.5	7.6	7.9
Respiratory	7.6	7.5	7.2	6.4	6.4	6.6	6.3
Adverse effects	6.7	6.5	7.5	6.8	6.3	6.0	6.4
Fungal	6.8	6.0	5.0	4.8	5.0	4.4	3.9
Dermatological	3.4	2.8	3.3	2.4	2.7	2.7	2.6
Intestinal	2.2	2.2	2.5	2.1	2.2	1.6	1.5
Sexually transmitted disease	1.6	1.8	1.4	1.4	1.3	1.3	1.5
Tuberculosis	1.4	1.0	0.7	0.7	0.6	0.5	0.3
Gynecologicala	6.6	11.6	9.9	9.1	7.7	7.9	7.9

Note. Data are based on at least 1 treatment or primary diagnosis from a physician or hospital claim in the year in question.

^aAmong women only.

The rate of renal disease increased 19% over the 7 years. Renal impairments, up to and including end-stage renal disease, were associated with extremely high expenditures. In terms of 1997 dollars, annual Medicare payments for those with renal conditions averaged $46 912 during the 1997 through 2003 period. Significant portions of these expenditures were because of the high cost of dialysis during end-stage renal disease. When beneficiaries who qualified for Medicare as a result of end-stage renal disease were omitted, the mean annual Medicare expenditure still amounted to $42 853 per renal patient.

Renal disease and its attendant expenses were concentrated among Blacks. An average of 14.2% of Black beneficiaries with confirmed HIV had renal disease, compared with only 2.8% of White beneficiaries and 4.4% of Hispanic beneficiaries. Again excluding those who entered Medicare as a consequence of end-stage renal disease, expenses associated with renal disease were 2.3 times higher among Black patients than among White patients during 1997 through 2003 (data not shown).

On average, diabetes and heart disease were 2.4 times and 3.2 times more common, respectively, among those for whom renal disease was not their initial Medicare qualification. Mortality averaged 5.5 times higher in this group. Removal of all Medicare beneficiaries with renal disease from the analysis eliminated some but not all of the race/ethnicity-based expenditure and mortality differences. Yearly mortality rates among those without renal disease averaged 7.3% among Blacks and 5.6% among Whites. Per-patient Medicare expenditures were still an average of 31% higher for Blacks than for Whites.

DISCUSSION

The population of Medicare beneficiaries with confirmed HIV increased 80% from 1997 through 2003, driven by high annual incidence rates and reduced mortality. Our findings illustrate how expenditures for Medicare’s HIV population are changing as the number of new beneficiaries with HIV decreases and the number of beneficiaries with preexisting cases increases. These effects may be the result of better disease management in general and the introduction of potent anti-retroviral agents in particular.

If disease management is to be improved further, it is vital to establish the extent to which antiretroviral medications have been successful in altering the clinical and payment characteristics of those who were previously seriously ill. The Medicare cohort presents a unique opportunity to do so, because members of Medicare’s disabled population by definition have had disabling disease for at least 29 months. In addition to allowing assessments of racial differences, the size and detail of Medicare’s patient database allow analyses of the current status and payment histories of HIV-treated individuals to an extent not before possible. Other studies of patient status and expenditures have been limited by a focus on narrow geographic areas or time periods.11^,12 They also have lacked exact data on expenditures, relying instead on imputed or estimated values. A recent international review of 543 potentially useful cost studies revealed only 9 with sufficient data to make meaningful statements about post-HAART expenditures.13

We found that mortality greatly decreased after the introduction of HAART, by 35% during our 7-year study period. A previous study6 reported that about 20% of Medicare’s symptomatic HIV beneficiaries died in 1995. This figure was only 5.9% in 2003. The apparent 70% decrease parallels the 65% decline in the overall HIV-diagnosed population after HAART became available.1^,2 Yet, death rates among Medicare beneficiaries with HIV or HIV-related conditions remained high relative to rates in the overall US population with HIV and in the Medicare population as a whole.

In 2003, according to the Centers for Disease Control and Prevention, the death rate among people with AIDS (i.e., those with the most advanced HIV disease) was 45.9 per 1000,1 in contrast to the Medicare figure of 58.8 found in the present study. Also in 2003, disabled patients receiving treatment for HIV died at a rate 2.4 times higher than that of disabled beneficiaries without HIV. Mortality increases Medicare expenditures by a large magnitude (Table 4 ▶).

Rates of diabetes, heart disease, and liver disease are increasing among HIV-treated Medicare beneficiaries. Single-organ disease, cardiovascular disease, and diabetes have received new attention in the era of HAART. Antiretroviral drugs clearly have effects on lipid and glucose processing,14 as do chronic HIV infection, comorbidities, aging, and environmental factors. Rapidly progressing nephropathy associated with advanced HIV has been a concern since the beginning of the epidemic,15 but recent studies have highlighted more indolent renal conditions even among patients whose HIV has been suppressed.16 Notably, these renal conditions are seen predominantly in Blacks.

In the absence of a drug benefit that covers HAART, hospitalization accounts for half of Medicare expenditures for HIV patients. Other reports17 as well as our data indicate that the introduction of HAART quickly led to a general reduction in hospitalizations. This trend, initially much sharper outside of Medicare, evened out after 1998 in the general population. According to our data, the plateau occurred 2 years later in the case of Medicare. Another study focusing on the period between 2000 and 2002 period showed that the adjusted hospital admission rate was 1.7 times higher among HIV-treated Medicare patients than among patients with private insurance.11 The inability to further reduce hospital stays among Medicare beneficiaries has contributed to the stability of expenditures associated with this group in recent years even as rates of HIV-related diseases have continued to decline.

When expenditures related to self-administered drugs are considered ($10 000–$15 000 per year for antiretroviral drugs alone),18 it is quite possible that total health care expenditures for the HIV-treated Medicare population increased after HAART’s introduction, along with life expectancies. This is what occurred in Canada according to 2 studies that included data on pharmacy payments.19^,20 These studies showed that per-patient expenditures for antiretroviral drugs tripled during the transition to HAART, more than outweighing reductions in expenditures associated with care and other drugs.

The clear benefits of HAART in terms of increased life expectancy, reductions in permanent disability, and decreased demand for medical care argue for the value of ensuring that Medicare beneficiaries with HIV have access to drug therapies. The new Medicare part D pharmacy benefit promises to coordinate financing of Medicare-covered health care and currently noncovered pharmaceuticals. However, Medicaid was already supporting drug expenses for two thirds of the Medicare beneficiaries with HIV in our study, and these individuals now face a more challenging drug payment system managed by a multitude of private entities with varying formularies.

The part D patient copayment and the gaps in drug coverage may lead to increases in personal expenditures among beneficiaries who previously had private pharmacy insurance. At both population and personal levels, barriers to expensive but beneficial therapies such as HAART may broaden as patients live longer and total expenditures further increase. Taking advantage of these drugs’ demonstrable ability to extend survival and limit the use of nondrug medical services requires sound financial planning of the type this study supports.