As early as 1990, Medicaid, the principal public medical insurer of indigent patients in the USA, had become the principal insurer of patients hospitalized with AIDS in several cities . In a survey of 11 state and local health departments across the USA sponsored by the Center for Disease Control and Prevention and conducted in 1990-1992, 50% of patients with AIDS were insured by Medicaid . In Maryland, Medicaid was insuring 50% of patients with AIDS by 1991, an increase from 30% of patients 5years earlier . Given the epidemiologic trends in recent years toward HIV infection and the increasing prevalence of HIV infection in young women, racial/ethnic minorities, and injecting drug users, it is likely that Medicaid will continue to be the major insurer of patients with HIV infection and AIDS in the USA [4-6]. Medicaid costs have risen dramatically in the USA in recent years and have become the target of efforts to lower federal and state spending . A study of Medicaid insured patients, conducted at our institution in 1993-1995, estimated the cost of treating an HIV infected patient who presents with a CD4 cell count >500¥106 cells/l at $133500 over 8.3years .
Antiretroviral treatments with protease inhibitors have been shown to reduce rates of clinical disease progression, opportunistic infections, and death [9-12]. We wished to determine how the cost to Medicaid of the medical care of patients with HIV infection cared for in Maryland has changed since the introduction of protease inhibitor therapy. Our practice continues to provide both primary and sub-specialty care for a predominately indigent inner urban population of HIV infected patients and Medicaid is the principal insurer of our patients. This provides us with a unique opportunity to quantify how the costs of care have changed between 1995 and 1997.
The Johns Hopkins University AIDS Service provides comprehensive primary and sub-specialty care for HIV infected patients. At the time of registration in the clinic, patients undergo a comprehensive evaluation by a physician or physician‚s assistant, a social worker, and a nurse who use case report forms to collect extensive demographic, clinical, laboratory, pharmaceutical and psychosocial data. Additional data are collected every 6months by review of all ambulatory and inpatient medical records. This is supplemented by an automated laboratory database and all available records available from facilities other than Johns Hopkins Hospital. These data are abstracted by trained medical records technicians. Comprehensive HIV clinic medical records are maintained in the HIV clinic separately from the main hospital medical records. Validity checks are performed on a 10% sample of all collected data: concurrence is >98% on all abstracted fields. Systematic errors result in retraining of abstractors and re- abstraction on all records. To ensure a high rate of longitudinal follow-up, medical records from outside institutions are routinely sought on all patients. These include inpatient discharge summaries, outpatient clinic records, laboratory sheets, and home-health records. For patients who have not been seen in the last 12months, state and national vital statistics are surveyed; this process has been described previously .
Patients were included in this analysis if they gave informed written consent to link their clinical data with the state Medicaid claims database maintained by the Policy and Health Statistics Administration of the Department of Health and Mental Hygiene. Linkage was done using the patients‚ medical assistance numbers corroborated by their names and dates of birth. All patients insured by Medicaid were approached to request consent while they were in the clinic prior to seeing the medical care providers. Patients were included who had contributed fee-for-service Medicaid claims extending from 1 January 1995 to 31 December 1997. A comparison of the study sample with records of the overall HIV clinic database showed that 99% of patients insured by Medicaid during the time consent was being obtained were enrolled.
The Maryland Medicaid database maintains all inpatient, outpatient, and pharmacy claims data submitted to the state. These claims were stratified into five payment categories: (i) inpatient hospitalization (all bundled costs including professional fees for acute state and rehabilitation hospitals); (ii) outpatient clinic (including physician‚s professional fees); (iii) outpatient pharmacy; (iv) community care (including long-term care facilities, community, and home-health services); and (v) emergency department care. We analyzed only payments made by Medicaid, not charges to Medicaid, as the measure of costs to state and federal governments and ultimately to the taxpayer. All medical claims were analyzed whether for HIV related care or not. Any care received in the state of Maryland was captured, even if not delivered at Johns Hopkins. As far as we know, none of the study patients moved out of state, which is not unusual given that these HIV-infected patients are almost all lifelong residents of the Baltimore, Maryland area. Starting in the latter half of 1997, Maryland Medicaid converted from a fee-for-service to managed care model. All data used in this analysis were fee- for-service prior to conversion and were not affected by this change in provision of care.
Average payments made by Medicaid per month were calculated for each calendar year from 1995 through 1997. Payments per month were further stratified by CD4 cell count category (<50, 50-200, and 201-500¥106 cells/l), and by whether or not a protease inhibitor antiretroviral regimen was being used. For the CD4 cell count stratification, the CD4 cell count obtained closest to the time of the payment was used. A protease inhibitor regimen was defined as any antiretroviral regimen containing either a protease inhibitor and two nucleoside analogues or two protease inhibitors and either one or two nucleoside analogues. Average monthly payments were computed by dividing the payments made for each patient by the number of months of enrollment in Medicaid. Monthly payments are reported as mean±SEM, a measure of the precision of the estimate of the mean. Cost distributions are typically skewed toward higher costs, which tend to increase the mean to a value higher than the median, a measure of the middle-most cost when costs are simply ranked. However, insurers such as Medicaid must pay these higher costs, and the budgetary impact is arguably better represented by the mean rather than the median. We did not adjust for inflation since we felt that the most conservative comparison of annual costs would use unadjusted payment data.
Comparisons of costs between CD4 cell count or protease inhibitor use categories were performed using the Wilcoxon rank sum test. A multivariate repeated measures regression analysis was performed to assess the effect on Medicaid payments of age, race, sex, HIV risk group, CD4 cell count, and the use of protease inhibitor-containing regimens. The analysis was carried out by using general estimating equations repeated measures methodology (PROC GENMOD, SAS) to adjust the variance for possible multiple observations contributed by a single individual. Two models were assessed, the first with hospital inpatient payments as the dependent variable, and the second with total health care payments as the dependent variable. Log-transformed payments were used as the dependent variable for these analyses to provide an approximately normal distribution appropriate for the analyses.
In addition, the change in average Medicaid payments per month associated with inpatient hospitalization for opportunistic illness from 1995 through 1997 was analyzed comparing patients who used protease inhibitor therapy with those who did not. This was done by summing Medicaid inpatient payments when the principal ICD-9-CM discharge diagnosis corresponded to any one of the AIDS-defining infections, malignancies, or other illnesses and dividing the result by the number of months of enrollment when using or not using a protease inhibitor regimen. This computation was done for successive 6-month intervals from 1995 through 1997 and the results plotted for protease inhibiting using and non-using patients. This analysis was restricted to patients with CD4 cell counts <200¥106 cells/l, since opportunistic illness is unlikely to occur when the CD4 cell count is >200¥106 cells/l.
This analysis is based on data from 695 patients representing 9543 person-months of follow-up time. The demographics of the study population are shown Table 1. The patients were predominately African-American and male, with a median age of 34years. Injecting drug use was the principal risk factor for HIV transmission. The demography of the population did not change significantly from 1995 to 1997.
CD4 cell count stratified payments
The annual mean (SEM) monthly payments made by Medicaid are shown in Table 2. Total health care payments were unchanged from 1995 through 1997 when the CD4 cell count was <50 or 201-500¥106 cells/l. Total payments increased in 1997 compared with 1995 when the CD4 cell count was 50-200¥106 cells/l. Similarly, there was little change in payments made for hospitalization, for outpatient care, community care or emergency room use. There was, however, an increase in antiretroviral expense in all CD4 cell count strata from 1995 to 1997. In any of these years, the greatest increase in expense occurred when the CD4 count was <50¥106 cells/l.
Protease inhibitor use
The mean (SEM) monthly payments made by Medicaid in 1996 and 1997 stratified by CD4 cell count and use of protease inhibitors are shown in Tables 3 and 4. This analysis was not done for 1995 because there was almost no use of protease inhibitors. In both years, patients receiving a protease inhibitor regimen had lower hospital inpatient payments than patients not receiving a protease inhibitor regimen within each CD4 cell count stratum. However, protease inhibitor users had significantly higher ambulatory pharmacy costs than non-protease inhibitor users in all CD4 cell count strata. Total health care payments were similar for protease inhibitor using and non-using patients. These tables also show the demographic characteristics of the patients who were using or not using a protease inhibitor regimen. In both years, those using a protease inhibitor were more likely to be male, white, and homosexual.
The results of the multivariate analysis are shown in Table 5. Variables associated with higher inpatient hospital payments included a lower CD4 cell count, earlier year of treatment, being female, African-American, or having injecting drug use as a risk factor for acquiring HIV. Adjusting for these variables, protease inhibitor use was associated with significantly lower inpatient hospital payments. Variables associated with higher total health care payments included a lower CD4 cell count, early year, and older age, but no other factors including protease inhibitor use.
Plots of the average inpatient Medicaid payments for hospitalization with a principal discharge diagnosis of an opportunistic illness stratified by protease inhibitor use and non-use are shown in Fig. 1. The difference between those two therapeutic groups in the average monthly payments was significant in 1996 in the repeated measures regression analysis (P<0.001). By 1997, the difference between therapeutic groups was highly significant (P<0.0001).
HIV disease has been associated with frequent opportunistic infections, the use of multiple and often expensive drugs and other therapies, and the relatively frequent need for hospitalization. Early estimates of the costs of medical care for patients with AIDS have varied widely, with the expectation that most care would be given during inpatient hospitalization [14-17]. With the movement of HIV care into the outpatient arena, cost estimates for treating HIV infection decreased . We previously published an analysis of average monthly payments made by Maryland Medicaid to provide care for our HIV-infected patients from 1993 to 1995 . This analysis reflected an earlier therapeutic era when reverse transcriptase inhibitors were the only available antiretroviral therapy. We have now analyzed data from 1995 to 1997 to assess the economic impact of the introduction of protease inhibitor-containing antiretroviral regimens. Using the same CD4 cell count stratification as that used in the earlier paper, it was found that total health care payments either remained unchanged (CD4 cell count<50 and 201-500¥106 cells/l) or increased slightly (CD4 cell count 50-200¥106 cells/l) from 1995 to 1997. In patients with a CD4 cell count >50¥106 cells/l, there was a small decline in inpatient payments, but an increase in pharmacy (particularly antiretroviral) payments. In patients with a CD4 cell count <50¥106 cells/l, there was essentially no change from 1995 through 1997 in inpatient or total pharmacy payments, although there was a relative increase in antiretroviral compared with non- antiretroviral pharmacy payments. Outpatient payments generally remained stable from 1995 to 1997, as did community care and emergency room payments.
It was more revealing to further stratify payments by the actual use of a protease inhibitor-containing regimen. When this was done, it was found that in 1996 and 1997, patients who were using a protease inhibitor regimen had significantly lower inpatient payments according to CD4 cell count strata. They also had significantly higher pharmacy payments, though outpatient and emergency room payments were generally similar for protease inhibitor using and non-using patients. Notably, total health care payments tended to be lower for patients who used a protease inhibitor regimen, though they were significantly lower only in patients who had a CD4 cell count <50¥106 cells/l in 1997.
Patients who were using a protease inhibitor were somewhat more likely to be homosexual men and of white race. However, multivariate analysis demonstrated that protease inhibitor use was significantly and strongly associated with lower inpatient payments. Several other factors including the CD4 cell count, injecting drug use, African-American race, and female gender were associated with higher hospital costs. Fig. 1 demonstrating the differences in inpatient rates of payment for opportunistic illness between protease inhibitor using and non-using patients is consistent with the decline in opportunistic illness seen by others [10,12]. Total health care payments were also associated significantly with CD4 cell count; however, there was no association between total health care payments and demographic characteristics of the patients. In addition, protease inhibitor use was not associated with total payments indicating the increase in pharmacy payments was balanced by a decline in inpatient payments. It is also possible, however, that differences in the total costs of care were not detect because of the small sample size. Notably, the total costs were lower for protease inhibitor users compared with non-users in most CD4 cell count strata in 1996 and 1997.
Medicaid is the predominant medical insurer of persons with HIV infection and AIDS in the USA, and these results are probably applicable to this large group of patients. In Maryland, there is an all payer rate setting system in which Medicaid payment rates to hospitals are approximately equal to those of private insurers. Additionally, the actual cost of care for inpatient services is tied more closely to payments than in most states . Therefore, the inpatient data probably provide a good representation of the actual costs of care. From 1995 through 1997, the providers of primary and specialty care in our clinic have changed very little and the patient population has been demographically stable. Therefore, these changes in payments are likely to reflect the recent changes in their options for treatment. There was no adjustment for inflation; the US Consumer Price Index indicated an inflation rate of less than 3% per annum from 1995 to 1997, and the results would have changed very little with such adjustment.
Our results may not be generally applicable to HIV-infected patients in the USA with private insurance or to populations outside of the USA, particularly those in the developing world. In countries in which the costs of hospital and other health care are much less than in the USA, a decline in these costs may not balance the increase in pharmacy costs. As demonstrated by Hogg et al., the cost of combination antiretroviral therapy could comprise a substantial proportion of the gross domestic product  in many of the world‚s developing countries. These drugs may be out of reach for the majority of persons around the world with HIV.
We believe that there are several useful messages from this new analysis of 1995-1997 Medicaid payment data. First, approximately half the patients‚ follow-up time in 1996 and 1997 is assigned to protease inhibitor use. This indicates that in our Medicaid-insured population (approximately 60% of our total patient population), protease inhibitors are frequently not being used. Although this study was not designed to analyze reasons for use and non-use of protease inhibitors, we know anecdotally in our clinical practice that behaviors such as drug abuse, socioeconomic-related problems with housing stability, and other types of problems interfere with both the prescribing of and adherence to protease inhibitor-containing regimens. Second, when protease inhibitors are used, inpatient costs are lower. This is consistent with other data showing declines in hospitalization associated with protease inhibitor therapy [21-23]. As expected, it was also found that pharmacy costs are higher in these patients, consistent with the findings of others . Overall, total health care costs are similar in patients who use or do not use protease inhibitors. We consider this to be a favorable result in that patients have received the clinical benefit of the protease inhibitor combination antiretroviral regimen with a decline in both opportunistic illness and in the need for hospitalization. Other studies have also shown a decline in opportunistic illness [10,12] and improved survival [9,11]. Such a benefit may have been achieved without increasing overall health care costs. Although this is not a formal cost-effectiveness analysis, the implications are that the patients have received a very good value (decreased morbidity), while the health care payer maintained medical care costs at a relatively stable average monthly rate from 1995 through 1997. We believe that this is an important message for those who receive, provide, and pay for HIV care.
1. Green J, Arno PS. The ‚medicaidization‚ of AIDS: trends in the financing of HIV-related medical care. JAMA
2. Diaz T, Chu SY, Conti L, et al
. Health insurance coverage among person with AIDS: results from a multistate surveillance project. Am J Public Health
3. Bartnyska LM, Schactman, M Hidalgo. Patterns in Maryland Medicaid enrollment among persons with AIDS. Inquiry
4. Centers for Disease Control and Prevention. AIDS among racial/ethnic minorities - United States, 1993. MMWR
5. Centers for Disease Control and Prevention. AIDS among women - United States. MMWR
6. Centers for Disease Control and Prevention. HIV/AIDS surveillance Report.
7. Voelker R. States face rocky road to reform. JAMA
8. Moore RD, Chaisson RE. Costs to Medicaid of advancing immunosuppression in an urban HIV infected patient population in Maryland. J Acquir Immun Defic Syndr
9. Centers for Disease Control and Prevention. Update: trends in AIDS incidence, deaths, and prevalence - United States, 1996. MMWR
10. Palella Jr FJ, Delaney KM, Moorman AC, et al. Declining morbidity and mortality among patients with advanced human immunodeficiency virus infection. N Engl J Med
11. Hogg RS, Heath KV, Yip B, et al. Improved survival among HIV-infected individuals following initiation of antiretroviral therapy. JAMA
12. Mocroft A, Vella S, Benfield TL, et al. Changing patterns of mortality across Europe in patients infected with HIV-1. Lancet
13. Moore RD. Understanding the clinical and economic outcomes of HIV therapy: the Johns Hopkins HIV Clinical Practice Cohort. J Acquir Immune Defic Syndr
14. Hardy AM, Rauch K, Echenberg D, Morgan WM, Curran JW. The economic impact of the first 10,000 cases of acquired immunodeficiency syndrome in the United States. JAMA
15. Scitovsky AA, Cline M, Lee PR. Medical care costs of patients with AIDS in San Francisco. JAMA
16. Andrulis DP, Beers VS, Bentley JD, Gage LS. The provision and financing of medical care for AIDS patients in U.S. public and private teaching hospitals. JAMA
17. Bennett CL, Cvitanic M, Pascal A. The costs of AIDS in Los Angeles. J Acquir Immun Defic Syndr
18. Hellinger FJ. The lifetime cost of treating a person with HIV. JAMA
19. Anderson G, Chaulk P, Fowler E. Maryland: a regulatory approach to health system reform. Health Affairs
20. Hogg RS, Weber AE, Craib KJP, et al. One world, one hope: the cost of providing antiretroviral therapy to all nations. AIDS
21. Torres RA, Barr M. Impact of combination therapy for HIV infection on inpatient census. N Engl J Med
22. Keiser P, Kvanli MB, Turner D, et al. Protease inhibitor based therapy is associated with decreased HIV related health care costs. J Acquir Immune Defic Syndr Hum Retrovirol
23. Mouton Y, Alfandari S, Valette M, et al. Impact of protease inhibitors on AIDS-defining events and hospitalizations in 10 French AIDS reference centres. AIDS
24. Perdue B, Weidle P, Everson-Mays R, et al. Evaluating the cost of medications for ambulatory HIV-infected persons in association with landmark changes in antiretroviral therapy. J Acquir Immune Defic Syndr