Use of Protease Inhibitors and Non-Nucleoside Reverse Transcriptase ...

 RESEARCH 

Use of Protease Inhibitors and Non-Nucleoside Reverse Transcriptase Inhibitors Among Medicaid Beneficiaries With AIDS | Usha Sambamoorthi, PhD, Patrick J. Moynihan, PhD, Elizabeth McSpiritt, MD, MPH, and Stephen Crystal, PhD Studies of HIV care have often documented differential access to new drug therapies across sociodemographic subgroups. For example, early access to zidovudine, the first antiviral therapy for HIV disease, was found to be less prevalent among women, members of racial minorities, adolescents and young adults, active injection drug users, and the uninsured.1–4 After the introduction of zidovudine, several years passed before rates of use across demographic subgroups converged.1,5 Since early 1996, combination therapies that include new, more effective antiviral drugs— protease inhibitors (PI) and non-nucleoside reverse transcriptase inhibitors (NNRTI)—have become available for treatment of HIV infection.6 Some studies suggest that during the years immediately after the introduction of protease inhibitors, women, racial minorities, and injection drug users were significantly less likely to use these new drug therapies.7–11 Also, a handful of published papers—including those from the HIV Cost and Services Utilization Study (HCSUS), which involved a representative sample of the adult US population infected with HIV—have documented a rapid increase in the use of protease inhibitors.8,10,12,13 In addition, some of these studies have examined diffusion of the new therapies among subgroups defined by sex, race, and other characteristics.8,10 Although these studies have reported a tendency toward a decrease over time in racial and ethnic differences in use of antiretroviral regimens involving protease inhibitors, it has typically been found that African Americans and women continue to lag behind nonminority groups in use of such therapies.8,10 However, the studies were unable to track recent changes in protease inhibitor use owing to data limitations. Messeri and colleagues explored the diffusion issue through 1997,8 and the HCSUS addressed PI/NNRTI use only through January 1998.10 In this article, we examine the determinants of PI/NNRTI use over time among

Objectives. This study compared the use of new antiretroviral treatments across sociodemographic subgroups during the 3 years after the introduction of these treatments and examined diffusion of the therapies over time. Methods. Merged surveillance and claims data were used to examine use of protease inhibitors and non-nucleoside reverse transcriptase inhibitors (PI/NNRTIs) among New Jersey Medicaid beneficiaries with AIDS. Results. In 1996, there were sharp disparities in use of PI/NNRTI therapy among racial minorities and injection drug users, even after control for other patient characteristics. These gaps had decreased by 1998. Higher PI/NNRTI treatment rates were also observed among beneficiaries enrolled in a statewide HIV/AIDS-specific home- and community-based Medicaid waiver program. Conclusions. Even within a population of individuals similar in regard to health coverage, there were substantial sociodemographic differences in use of PI/NNRTIs during the early years after their introduction.These differences narrowed as new treatments became standard. Participation in a case-managed Medicaid waiver program seems to be associated with a more appropriate pattern of use. These results suggest a need to address nonfinancial barriers to care. (Am J Public Health. 2001;91:1474–1481)

persons with AIDS, using more recent data. We analyze claims data from 2089 adults with AIDS who received Medicaid benefits in New Jersey between January 1996 and December 1998. Such an analysis within a single payer source is critical because, despite financial eligibility, disadvantaged subpopulations may differ in their access to outpatient health care services. A number of studies have shown that among persons with AIDS, women, members of racial minorities, and injection drug users receive fewer medical care services than non-drug-using White men, even after insurance differences have been controlled.2,14,15 Because New Jersey is among the topranking states in regard to number of AIDS cases, it is an important state in which to study this issue.16 In addition, the state’s HIV/AIDS registry data have been merged with Medicaid claims data, allowing for better identification of HIV-infected individuals (in contrast to diagnostic screening approaches to case identification used in other studies).17 This study estimated crude and adjusted rates of PI/NNRTI use among HIV-infected

1474 | Research Articles | Peer Reviewed | Sambamoorthi et al.

Medicaid recipients in New Jersey. The objectives included comparing patterns of PI/ NNRTI use across demographic subgroups (e.g., sex, race/ethnicity, risk group, and geographic residence), examining diffusion of use over time, and identifying correlates of PI/NNRTI use over time.

METHODS Study Population The population for the present study consisted of adult Medicaid participants who were diagnosed with AIDS in New Jersey between January 1991 and December 1998. Three sources of data were combined into client-level files: HIV/AIDS registry data from the New Jersey Department of Health and Senior Services, paid Medicaid claims for medical care and prescription drugs from the Division of Medical Assistance and Health Services of the New Jersey Department of Human Services, and the Division of Medical Assistance and Health Services Medicaid eligibility file. The HIV/AIDS registry and the Medicaid file were linked by the Department of Health

American Journal of Public Health | September 2001, Vol 91, No. 9

 RESEARCH 

and Senior Services by identifying fields common to both files, such as name, date of birth, sex, and Social Security number. In linking these databases, the Department of Health and Senior Services used a match-scoring procedure to determine database intersections. Once the link was established, we were provided with a master file in which clients were identified only by scrambled Medicaid identification numbers. HIV/AIDS registry. The HIV/AIDS reporting system maintained by the Department of Health and Senior Services, designed and supported by the Centers for Disease Control and Prevention (CDC), contains all New Jersey AIDS cases diagnosed since the beginning of the epidemic and all HIV infections reported since 1992. Definition changes were accommodated with software supplied by CDC that automatically reclassifies cases from HIV to AIDS status if the new criteria are met. In New Jersey, active investigation was done to reclassify individuals on the basis of the definition revisions in both 1987 and 1993. Rates of completeness of reporting estimates based on validation and follow-up through independent data sources (death certificates, hospital discharge records, AIDS Drug Distribution Program) ranged from 90% to 95%. We obtained information on demographic characteristics (i.e., sex, race, county of residence at diagnosis), exposure category, and date of AIDS diagnosis from the surveillance data. Medicaid paid claims. Medicaid claims histories contained all processed claims for services and pharmacy prescriptions between January 1988 and April 1999. We included paid claims from January 1996 through December 1998, because protease inhibitors were first introduced during the final months of 1995. To allow for time lags between receipt of services, billing, payment, and appearance of paid claims in the computerized database, we included services received through December 1998 in the analyses. The non–drug claims file provided information on claim type, diagnosis, category of service, dates of service, and actual amount paid by Medicaid for each of the services. The pharmacy claims file contained information on National Drug Codes, dispensing date, and actual amount paid by Medicaid.

Medicaid eligibility file. The Medicaid eligibility file contained the Medicaid enrollment and termination dates for each individual. Using these dates, we excluded individuals who left Medicaid during or before March 1996. In addition, we constructed a variable to indicate interruptions in Medicaid enrollment and excluded individuals with breaks in Medicaid enrollment after March 1996 because of the possibility of incomplete data on protease inhibitor use among these individuals. We combined information on dates of death from Medicaid eligibility file and surveillance data to classify respondents as decedents or nondecedents. The study population comprised individuals diagnosed with AIDS between January 1991 and March 1996 who participated in the New Jersey Medicaid program between January 1996 and December 1998. We included only those individuals who received Medicaid services during or after 1996 (based on the availability of protease inhibitors). To better control for disease stage, we included only individuals diagnosed with AIDS. Also, respondents had to be 18 years or older at the time of AIDS diagnosis and enrolled in Medicaid for at least 90 days during the study period. We excluded individuals participating in managed care because encounter data for these individuals might not be complete. We excluded a small number of patients with data inconsistencies (e.g., patients who had PI/NNRTI claims but no accompanying physician claims). In the final stage, we identified 2089 beneficiaries who met the study criteria. In addition, we required that individuals be followed for at least 90 days during the year of observation to be included in year-specific analyses.

Measures PI/NNRTI use. As of summer 1999, 14 antivirals for treatment of HIV disease had been approved by the Food and Drug Administration18: 6 nucleoside analogue reverse transcriptase inhibitors (abacavir, didanosine, lamivudine, stavudine, zalcitabine, and zidovudine), 3 non-nucleoside reverse transcriptase inhibitors (nevirapine, delavirdine, and efavirenz), and 5 protease inhibitors (indinavir, ritonavir, saquinavir, nelfinavir, and amprenavir). We used National Drug Codes

September 2001, Vol 91, No. 9 | American Journal of Public Health

recorded in pharmacy claims to identify PI/ NNRTI use. Ritonavir, indinavir, saquinavir, and nelfinavir were the protease inhibitors available during the observation period; all 3 of the non-nucleoside reverse transcriptase inhibitors just mentioned were available. It should be noted that the protease inhibitor amprenavir was approved in April 1999 (after the end of our study period) and therefore was not included. For each individual, we constructed an indicator variable in which 0 indicated no PI/ NNRTI use and 1 indicated the use of either a protease inhibitor or a non-nucleoside reverse transcriptase inhibitor. It was assumed that most patients using PI/NNRTIs were on combination regimens. Demographic characteristics. Demographic characteristics (i.e., sex, race, county of residence at diagnosis), exposure category, and date of AIDS diagnosis were obtained from the surveillance data. Race/ethnicity was characterized as White, African American, or Latino. In multivariate analyses, White was used as the reference group. Exposure category was based on information on injection drug use history obtained from the AIDS registry; patients were classified as either users or nonusers. Age at diagnosis was categorized into the following groups: 18 to 29 years (the reference group in multivariate models), 30 to 39 years, 40 to 49 years, and 50 years or older. We also compared treatment rates between individuals residing in the highest prevalence counties (i.e., those near New York City and Philadelphia) and those residing elsewhere. Illness stage. To better control for disease stage, we included dummy variables for year of AIDS diagnosis and decedent status as controls in the regression analysis. On the basis of dates of death available from both the AIDS registry and the Medicaid eligibility file, respondents were classified as decedents or nondecedents. Decedent status was also assessed for each year of observation. Because AIDS diagnoses can involve wide ranges of disease severity, we also included the presence of opportunistic infections (e.g., Pneumocystis carinii pneumonia, Kaposi’s sarcoma, disseminated infection with Mycobacterium avium complex) as a marker of severity. (A complete list of the opportunistic infections

Sambamoorthi et al. | Peer Reviewed | Research Articles | 1475

 RESEARCH 

included is available upon request.) These infections were identified from International Classification of Diseases, Ninth Revision, Clinical Modification diagnostic codes for medical care episodes contained in administrative Medicaid claims data. Medicare participation. The recent introduction of highly active antiretroviral treatments, combined with aggressive use of prophylactic therapies for opportunistic infections, has extended the life expectancy of HIV-infected individuals.9,19 This has enabled more Medicaid beneficiaries with HIV to survive the waiting period for Medicare eligibility and to become dually eligible for Medicare and Medicaid.20 Medicare participation may provide broader access to health care providers such as office-based physicians. Therefore, we included Medicare coverage as a covariate. Medicare coverage was assessed for each year of observation in the study on the basis of the claim type recorded in paid Medicaid claims data. Waiver status. Many states have turned to Medicaid waivers for home- and communitybased care to improve access to and coordination of care for individuals with HIV/ AIDS.21 In New Jersey, a portion of the Medicaid population is enrolled in an HIV/AIDSspecific Medicaid home- and communitybased care waiver program that offers a variety of home care services, including mandatory case management with monthly visits by case managers. The waiver program is available to individuals with income levels above the regular Medicaid income threshold, up to the income level at which they would become financially eligible if institutionalized. Eligibility requirements of the program and the services provided have been documented elsewhere.22 Access to the home- and communitybased services provided by the waiver program has been shown to be associated with different patterns of service use (S. Crystal, U. Sambamoorthi, and A. T. LoSasso, unpublished data, 1998). Therefore, we also used waiver status as a covariate in all of our analyses. We determined participation in the waiver program by using the procedure codes in Medicaid claims for waivered services. These codes were provided by the

Division of Medical Assistance and Health Services.

Analytic Procedures In bivariate analyses, we tested for subgroup differences in the use of PI/NNRTIs with χ2 statistics. Logistic regression analyses were estimated to predict the probability of PI/NNRTI use and to isolate the effects of characteristics such as sex and race. Parameter estimates from logistic regressions were transformed into odds ratios (ORs) associated with each independent variable, representing relative risk ratios for a 1-unit change in the variable in question. Odds ratios exceeding 1 indicated an increased likelihood of PI/NNRTI use relative to the comparison group, whereas odds ratios below 1 indicated a decreased probability of use. To explore the extent of PI/ NNRTI use, we examined patterns of use across Medicaid subgroups for 1996, 1997, and 1998. Bivariate analyses and logistic regressions were conducted separately for each year of observation. We used nested logistic regression models for use of PI/NNRTIs to investigate the possibility that some of the differences in use could be due to disease severity. In stage 1, we entered only demographic characteristics. In stage 2, we included waiver program participation along with demographic characteristics. In stage 3, we included demographic characteristics, waiver program participation, illness stage, and Medicare coverage. In stage 4, in addition to the variables entered in stage 3, we included a further marker of illness severity (presence or absence of opportunistic infection). Findings from the models controlling for this additional illness marker (available upon request) were similar to the findings (reported subsequently) from the models without the indicator variable in terms of the presence of opportunistic infections.

RESULTS Table 1 describes the study population and compares the characteristics of patients with no PI/NNRTI use and those with PI/NNRTI use at any time during the follow-up period. The study population was 64% male, 57%


African American, 19% Latino, and 24% White. A majority of the patients (68%) were alive as of December 1998. Sixty-nine percent of the study cohort lived in the highHIV-prevalence urban areas of the state (near New York City or Philadelphia) at the time of initial diagnosis. Roughly one quarter of the respondents were enrolled in the HIV-specific waiver program, and just over a third were eligible for both Medicare and Medicaid. Results from the bivariate comparisons of PI/NNRTI use during the follow-up period are also reported in Table 1. Of the 2089 patients receiving Medicaid services between January 1996 and December 1998, 1365 (65%) received at least 1 PI/NNRTI prescription at some point during this period. In the bivariate comparisons, we found significant differences in use of PI/NNRTIs by race/ethnicity, age at diagnosis, illness stage, vital status as of 1998, Medicare participation, and waiver program participation. A significantly higher percentage of Whites (75%) than African Americans (62%) or Latinos (65%) had at least 1 paid prescription for PI/NNRTIs. In comparison with the other age groups, significantly lower proportions of the youngest and oldest groups received PI/NNRTIs. Those with Medicare as their primary insurance were more likely to receive PI/NNRTIs than Medicaid-only beneficiaries (76% vs 59%). Sixty-nine percent of waiver enrollees and 64% of nonwaiver participants used the new drugs. Logistic regression findings were similar to bivariate findings (Table 1). After control for other characteristics, African Americans were less likely to receive a prescription for PI/NNRTIs than Whites (OR = 0.58, 95% confidence interval [CI] = 0.45, 0.75). The same was true for Latinos (OR = 0.65, 95% CI = 0.47, 0.88). Patients living in the high prevalence urban areas were more likely than patients living elsewhere to have received at least 1 prescription for the new drugs. Medicare beneficiaries were twice as likely to be using PI/NNRTIs as Medicaidonly beneficiaries. Similarly, participants in the waiver program were more likely than traditional Medicaid beneficiaries to use the new drugs. Table 2 reports patterns of diffusion in the use of PI/NNRTIs from 1996 through the


 RESEARCH 

TABLE 1—Crude and Adjusted Rates of Protease Inhibitor/Non-Nucleoside Reverse Transcriptase Inhibitor (PI/NNRTI) Use Among New Jersey Medicaid Beneficiaries With AIDS: 1996–1998

Sex Male Female Race/ethnicity White African American Hispanic Age at diagnosis, y 18–29 30–39 40–49 ≥50 Exposure category Injection drug user Nonuser County of residence Urban Elsewhere Vital status as of 1998 Nondecedent Decedent Year of diagnosis 1991 or 1992 1993 or 1994 1995 or 1996 Waiver program participationb Nonparticipant Participant Medicare coverageb No Yes Total

Sample, No. (%)

PI/NNRTI Use, %

Odds Ratio

95% Confidence Interval

1334 (63.9) 755 (36.1)

64.2 67.4

... 1.34*

1.09, 1.65

501 (24.0) 1195 (57.2) 386 (18.5)

74.7a 61.7 64.5

... 0.58* 0.65*

0.45, 0.75 0.47, 0.88

276 (13.2) 1027 (49.2) 650 (31.1) 136 (6.5)

61.6a 66.9 66.5 55.9

... 1.24 1.27 0.77

0.92, 1.67 0.92, 1.75 0.49, 1.21

1272 (60.9) 581 (27.8)

65.2 66.3

1.03 ...

0.82, 1.29

1435 (68.7) 654 (31.3)

65.9 64.2

1.36* ...

1.09, 1.69

1414 (67.7) 675 (32.3)

73.2a 48.9

... 0.33*

0.27, 0.40

412 (19.7) 1136 (54.4) 541 (25.9)

65.5 65.5 64.9

... 1.03 0.95

0.85, 1.25 0.41, 2.26

1604 (76.8) 485 (23.2)

64.2a 69.3

... 1.46*

1.13, 1.89

1295 (62.0) 794 (38.0)

59.0a 75.7

... 2.12*

1.71, 2.62

2089 (100.0)

65.3

...

Note. Odds ratios were estimated from logistic regressions on PI/NNRTI use anytime between January 1996 and December 1998. The regression also included an intercept term. a Significant group difference in PI/NNRTI use at the 5% significance level. b Participation anytime between 1996 and 1998. *P ≤ .05 (relative to reference group).

end of 1998. Bivariate comparisons are presented for each of the years separately. The demographic composition of the study population remained essentially unchanged in 1996, 1997, and 1998. PI/NNRTI use increased from 45% in 1996 to nearly 70% by the end of 1998. In each year of observation, rates of use were lower among mem-

bers of racial minorities, Medicaid beneficiaries not enrolled in the case-managed waiver program, and patients with only Medicaid coverage. Although there were statistically significant differences in rates of PI/NNRTI use between Whites and members of minority groups during each year of observation,


these between-group differences had narrowed by the end of 1998. For example, the difference in the use of PI/NNRTIs between Whites and African Americans was 23.7% in 1996 but only 7.2% at the end of 1998. However, this was not the case for either Medicare participation or waiver participation. The initial difference of 14% between dually eligible beneficiaries and Medicaid-only beneficiaries had increased to 23% by the end of 1998. The difference in the rate of PI/NNRTI use between waiver enrollees and traditional Medicaid beneficiaries increased slightly, from 16% in 1996 to 21% by the end of 1998. In comparison with other groups, a significantly greater proportion of non–injection drug users received the new drugs in 1996; however, this difference decreased over time and was no longer significant by the end of 1998. Relative risk ratios from the simple logistic regression analyses of predictors of PI/NNRTI use for each year of observation are reported in Table 3. Differences in PI/NNRTI use for some subgroups were no longer significant by the end of 1998. For example, in 1996, injection drug users were significantly less likely than nonusers to receive the new treatments (OR = 0.73). However, by the end of 1998, this difference was no longer significant. Similarly, there were associations between race/ ethnicity and PI/NNRTI use that were significant in 1996 (African American OR = 0.43) and 1997 (African American OR = 0.58) but by the end of 1998 were no longer significant in the multivariate models. To explore the disparities in results between bivariate and multivariate models with regard to race, we also examined the use of PI/NNRTIs in 1998 with nested logistic regression models. In stage 1, we entered only the demographic characteristics. In stage 2, we included waiver program participation along with demographic characteristics. In stage 3, we included demographic characteristics, waiver program participation, illness stage, and Medicare coverage. Our results revealed that in stage 1, African Americans were significantly less likely to receive PI/NNRTIs (OR = 0.66, 95% CI = 0.49, 0.90). In the second stage, however, the difference between African Americans and Whites was no longer significant (OR =


 RESEARCH 

TABLE 2—Rates of Protease Inhibitor/Non-Nucleoside Reverse Transcriptase Inhibitor (PI/NNRTI) Use Among New Jersey Medicaid Beneficiaries With AIDS, by Year of Observation 1996

Sex Male Female Race/ethnicity White African American Hispanic Age at diagnosis, y 18–29 30–39 40–49 ≥50 Exposure category Injection drug user Nonuser County of residence Urban Elsewhere Vital statusa Nondecedent Decedent Year of diagnosis 1991 or 1992 1993 or 1994 1995 or 1996 Waiver program participationa Nonparticipant Participant Medicare coveragea No Yes Total

1997

1998

Sample, %

PI/NNRTI Use, %

Sample, %

PI/NNRTI Use, %

Sample, %

PI/NNRTI Use, %

63.9 36.1

45.3 45.0

63.5 36.5

67.5 69.2

63.5 36.5

68.4 72.3

24.0 57.2 18.5

61.9* 38.2 44.8

24.0 57.4 18.4

77.4* 65.4 64.3

23.8 57.1 19.0

75.9* 68.7 65.8

13.2 49.2 31.1 6.5

46.0 45.7 45.1 40.4

12.3 49.3 32.4 6.0

65.6 68.9 68.6 64.2

12.1 49.7 32.5 5.7

62.8 72.0 68.7 72.7

60.9 27.8

42.1* 51.3

62.2 26.8

67.6 70.0

62.0 26.9

69.4 70.7

68.7 31.3

43.7* 48.5

69.8 30.2

68.7 66.7

70.1 29.9

69.9 69.6

82.9 17.1

48.7* 28.3

90.5 9.5

69.0* 60.0

94.1 5.9

70.1 65.0

19.7 54.4 25.9

46.4 45.5 43.6

19.9 53.5 26.5

66.5 68.9 67.9

19.8 53.7 26.5

68.7 71.3 67.8

79.8 20.2

41.9* 58.2

82.6 17.1

65.5* 80.7

84.1 15.9

66.5* 87.5

69.4 30.6

40.9* 54.8

63.8 36.2

62.4* 78.2

63.3 36.7

61.3* 84.5

100.0

45.2

100.0

68.1

100.0

69.8

Note. Numbers of observations were as follows: 1996, 2089; 1997, 1584; 1998, 1356. a As of year of observation. *P ≤ .05.

0.77, 95% CI = 0.56, 1.05), suggesting that waiver participation explains at least part of the racial difference in the odds of receiving PI/NNRTIs. For some groups, the gap in PI/NNRTI use actually widened over time. Among waiver participants, the odds ratio for using PI/NNRTIs increased from 1.64 in 1996 to 2.74 by the end of 1998. Similarly, the gap widened between dually eligible beneficiar-

ies and Medicaid-only recipients for each year of observation. The odds ratio for dually eligible beneficiaries increased from 1.58 in 1996 to 3.25 by the end of 1998. No difference was noted in the use of PI/NNRTIs between patients living in highprevalence areas and others in 1996; by the end of 1998, however, patients living in high-prevalence areas were more likely to be using PI/NNRTIs. Similarly, although


no sex differences were found in PI/NNRTI use in 1996 and 1997, women were more likely to be using PI/NNRTI in 1998 (OR = 1.51, 95% CI = 1.17, 1.97).

DISCUSSION Use of PI/NNRTIs in New Jersey increased markedly between 1996 and 1998. During 1996, less than half of our study population


 RESEARCH 

TABLE 3—Logistic Regression Analysis of Protease Inhibitor/Non-Nucleoside Reverse Transcriptase Inhibitor (PI/NNRTI) Use Among New Jersey Medicaid Beneficiaries With AIDS, by Year of Observation 1996

Sex Male Female Race/ethnicity White African American Hispanic Age at diagnosis, y 18–29 30–39 40–49 ≥50 Exposure category Injection drug user Nonuser County of residence Urban Elsewhere Vital statusa Nondecedent Decedent Year of diagnosis 1991 or 1992 1993 or 1994 1995 or 1996 Waiver program participationa Nonparticipant Participant Medicare coveragea No Yes

1997

1998

Odds Ratio


Odds Ratio


Odds Ratio


... 1.08

0.89, 1.31

... 1.24

0.99, 1.56

... 1.51*

1.17, 1.97

... 0.43* 0.54*

... 0.35, 0.54 0.41, 0.72

... 0.58* 0.56*

0.43, 0.78 0.39, 0.80

0.81 0.75

0.58, 1.13 0.50, 1.11

... 1.05 1.07 0.82

0.79, 1.39 0.79, 1.46 0.53, 1.27

... 1.13 1.13 0.82

0.79, 1.60 0.77, 1.64 0.48, 1.42

... 1.43 1.23 1.37

0.97, 2.10 0.81, 1.85 0.74, 2.61

0.73* ...

0.59, 0.91

0.98 ...

0.75, 1.28

1.13 ...

0.84, 1.52

1.10 ...

0.90, 1.35

1.47* ...

1.14, 1.89

1.44* ...

1.08, 1.91

... 0.68*

0.56, 0.82

... 0.97

0.73, 1.30

... 0.62

0.39, 1.01

... 1.01 0.56

0.84, 1.21 0.22, 1.29

... 1.06 0.45

0.85, 1.32 0.18, 1.09

... 1.22 1.30

0.95, 1.56 0.41, 4.96

... 1.64*

1.29, 2.09

... 1.84*

1.31, 2.63

... 2.74*

1.76, 4.39

... 1.58*

1.30, 1.93

... 2.01*

1.57, 2.58

... 3.25*

2.43, 4.39

Note. Odds ratios were estimated from separate logistic regressions on PI/NNRTI use in 1996, 1997, and 1998. The regression also included an intercept term. a As of year of observation. *P ≤ .05 (relative to reference group).

received at least 1 prescription for PI/NNRTI regimens. This rate had increased to nearly 70% by the end of 1998. When use of PI/ NNRTIs was defined as any use during the entire follow-up period, we found significant inequities in use of PI/NNRTI therapy for racial minorities and for injection drug users. Our findings on racial disparities are consistent with studies identifying disparities in such areas as cardiac care23–25 and cancer

care.26–28 Some of these differences in use might be related to differences in treatment preferences and in knowledge and positive impressions of protease inhibitor use. However, existing data on HIV patients suggest that knowledge and positive impressions of protease inhibitor use do not differ substantially by race/ethnicity.29 Examination of PI/NNRTI use by year of observation revealed that during the first 2


years after the introduction of PI/NNRTI therapies, there were inequities in use of these medications. These inequities had disappeared by the end of 1998 for subgroups defined by race and mode of transmission. Such differences in use of innovations by socioeconomic status have often been documented in social research. For instance, one review of diffusion research suggested that those who demonstrate greater innovativeness and have a


 RESEARCH 

higher likelihood of adopting new ideas tend to be of higher socioeconomic status.30 Again, consistent with earlier research on zidovudine, there is some evidence that use differences in these new treatments abate over time.5 We found no differences in incidence of protease inhibitor use between men and women during 1996 and 1997, consistent with findings from earlier studies.9,12,31,32 However, this result was in contrast to results from the nationally representative HCSUS, which revealed less PI/NNRTI use among women. The differences in findings between HCSUS and our study may be due to the differing time periods and the inability to apply statistical controls for stage-of-illness differences in our population. For example, Shapiro and his colleagues examined PI/NNRTI use with HCSUS data as of January 1998, whereas our study followed individuals through the end of 1998.10 The absence of access disadvantages among women could be due to the active efforts of the state’s public health community, including state agencies, community-based organizations, and health care providers, to help HIV-positive women obtain care.33 In addition, New Jersey requires testing and counseling of all pregnant women. The Department of Health and Senior Services implements this law through frequent contact with health care facilities (including emergency departments) and provision of education to clinicians. Sex differences in PI/NNRTI use over time in the Medicaid population may also reflect differing disease severity between the 2 groups. Men enrolled in Medicaid may be sicker and at more advanced stages of illness than women, who are more likely to qualify for Medicaid insurance earlier in the illness stage through programs such as Temporary Assistance to Needy Families. However, our results indicated that, even after severity-of-illness markers had been controlled, women were more likely to receive PI/NNRTIs over time. Noteworthy among our findings is the statistically significant relationship between waiver program participation and PI/NNRTI use. In addition to the availability of home care services, mandatory comprehensive case management is a major component of the

waiver program. Previous studies suggest that case managers serve as negotiators of the health care system on behalf of their clients.22 Our current findings suggest that case management strategies in populations of HIV patients may be associated with more appropriate patterns of treatment. By the end of 1998, waiver program participation appeared to be associated with reduced racial differences in use of these new drug therapies. This finding is consistent with earlier research showing that, even among waiver program participants, racial parity was achieved only several years after the introduction of zidovudine.5 This finding also underscores the importance of studying variation within states in treatment patterns among individuals in specific payer systems such as Medicaid. Such an analysis would provide a complementary perspective to information from HCSUS, which involved a nationally representative sample. Focusing more closely on Medicaid beneficiaries within a single state, our study suggests that programs at the state level (e.g., New Jersey’s statewide HIV waiver) do contribute to variation in treatment patterns. Such information is important in determining targeted interventions at the state level. We also found higher treatment rates among dually eligible individuals even after adjusting for disease severity, suggesting that there are access advantages among AIDS patients who have both Medicare and Medicaid coverage relative to those who have only Medicaid coverage. These use differences may reflect socioeconomic characteristics, such as income and education, that tend to enhance access to care. They may also reflect better access to care for those with Medicare coverage through a greater choice of providers. This is particularly true in the case of office-based physicians, who receive higher reimbursements for services covered by Medicare than for those covered by Medicaid. For example, in New Jersey, at the time of this study, Medicare payments for office and hospital visits were nearly 4 times the fees paid by Medicaid.34 Some limitations of our data warrant consideration in interpreting the findings. Our study was based on a single payer (Medicaid) and may have missed some episodes of care


occurring in settings not reimbursed by Medicaid (e.g., privately financed or uncompensated care or services provided in a Department of Veterans Affairs hospital). However, we believe that most of the care received by this population, particularly pharmaceuticals, is reimbursed by Medicaid. The pattern of differences in PI/NNRTI use reported here may not be generalizable to the full population of HIV/AIDS patients because we did not include individuals outside the Medicaid system. However, in states such as New Jersey in which economic disadvantages are typical among persons with HIV, the majority of HIV-infected individuals enter the Medicaid system as the disease progresses.17,35 Also, because we excluded those with managed care encounters, we potentially underrepresented a segment of the AIDS population. However, penetration of managed care in the Medicaid disabled population of New Jersey was insignificant during the period of our study.36 Although we included some measures of disease stage and physical health status (e.g., year of diagnosis and vital status), this study lacked good independent measures of illness severity, such as CD4 counts, that are associated with treatment rates.10–12 However, it appears that the proportion of individuals in care for HIV disease who do not meet guidelines of the US Department of Health and Human Services (DHHS) for combination antiretroviral therapy (by virtue of CD4 count, viral load, or symptomatology) is relatively low. In fact, recent published analyses of data from HCSUS indicate that 99% of individuals in HIV care would meet DHHS criteria and be recommended for such treatment.37 Despite the limitations of our data, the present findings highlight the need for more research on nonfinancial barriers specific to traditionally disadvantaged subpopulations.

About the Authors The authors are with the Institute for Health, Health Care Policy, and Aging Research, Rutgers University, New Brunswick, NJ. Requests for reprints should be sent to Usha Sambamoorthi, PhD, Institute for Health, Health Care Policy, and Aging Research, Rutgers University, 30 College Ave, New Brunswick, NJ 08901 (e-mail: sambamoo@rci. rutgers.edu). This article was accepted October 11, 2000.


 RESEARCH 

Note. The findings and opinions reported here are those of the authors and do not necessarily represent the views of any other individuals or organizations.

Variations in the care of HIV-infected adults in the United States: results from the HIV Cost and Services Utilization Study. JAMA. 1999;281:2305–2315.

Contributors

11. Bing EG, Kilbourne AM, Brooks RA, et al. Protease inhibitor use among a community sample of people with HIV disease. J Acquir Immune Defic Syndr Hum Retrovirol. 1999;20:474–480.

U. Sambamoorthi and S. Crystal planned the study and selected the analytic approaches. U. Sambamoorthi carried out the analyses. U. Sambamoorthi and S. Crystal each drafted sections of the paper and reviewed and edited successive drafts. P. J. Moynihan conducted the literature review and contributed to writing the paper, particularly the introduction and discussion sections, as did E. McSpiritt. All authors contributed to interpretation of results and reviewed the manuscript at all stages.

Acknowledgments This research was supported by National Institute on Drug Abuse grants R01-DA11362-01 and R01DA11855-01 and by State/Territorial Minority HIV/ AIDS demonstration grant D92MP99003-01. We wish to acknowledge the programming assistance provided by Yang Lou and the research assistance provided by Michelle McGrath. We also gratefully acknowledge the cooperation and assistance of the New Jersey Department of Human Services and the New Jersey Department of Health and Senior Services in providing data for use in this study.

References 1. Moore RD, Stanton D, Gopalan R, Chaisson RE. Racial differences in the use of drug therapy for HIV disease in an urban community. N Engl J Med. 1994; 330:763–768. 2. Stein MD, Piette J, Mor V, et al. Differences in access to zidovudine (AZT) among symptomatic HIVinfected persons. J Gen Intern Med. 1991;6:35–40.

12. Sorvillo F, Kerndt P, Odem S, et al. Use of protease inhibitors among persons with AIDS in Los Angeles County. AIDS Care. 1999;11:147–155. 13. Keiser P, Kvanli MB, Turner D, et al. Protease inhibitor-based therapy is associated with decreased HIVrelated health care costs in men treated at a Veterans Administration hospital. J Acquir Immune Defic Syndr Hum Retrovirol. 1999;20:28–33. 14. Hellinger FJ. The use of health services by women with HIV infection. Health Serv Res. 1993;28:545–561. 15. Moore RD, Hidalgo J, Sugland BW, et al. Zidovudine and the natural history of the acquired immunodeficiency syndrome. N Engl J Med. 1991;20: 1412–1416. 16. Centers for Disease Control and Prevention. Basic statistics—ten states/territories and cities reporting highest number of AIDS cases. Available at: http://www. cdc.gov/nchstp/hiv_aids/stats/. Accessed October 13, 1999.

30. Rogers EM. The Diffusion of Innovations. New York, NY: Free Press; 1962. 31. Fairfield KM, Libman H, Davis RB, Eisenberg DM, Phillips RS. Delays in protease inhibitor use in clinical practice. J Gen Intern Med. 1999;14:395–401. 32. Kaplan JE, Parham DL, Soto-Torres L, et al. Adherence to guidelines for antiretroviral therapy and for preventing opportunistic infections in HIV-infected adults and adolescents in Ryan White-funded facilities in the United States. J Acquir Immune Defic Syndr Hum Retrovirol. 1999;21:228–235. 33. Department of Health and Senior Services, Division of AIDS Prevention and Control begins broad initiative to decrease perinatal transmission. AIDSline. 1999;11:1–2.

18. HIV/AIDS Treatment Directory. Vol 10. New York, NY: American Foundation for AIDS Research; 1999.

35. Flieshman JA. Transitions in insurance and employment among people with HIV infection. Inquiry. 1998;35:36–48.

19. Hogg RS, Heath KV, Yip B, et al. Improved survival among HIV-infected individuals following initiation of antiretroviral therapy. JAMA. 1998;279: 450–454.

4. Markson LE, Cosler LE, Turner BJ. Implications of generalists’ slow adoption of zidovudine in clinical practice. Arch Intern Med. 1994;154:1497–1504.

21. Miller NA. Medicaid 2176 home and communitybased care waivers: the first 10 years. Health Aff. 1992; 11:162–171.

5. Crystal S, Sambamoorthi U, Merzel C. The diffusion of innovation in AIDS treatment: zidovudine use in two New Jersey cohorts. Health Serv Res. 1995;30: 593–614.

22. Merzel C, Crystal S, Sambamoorthi U, et al. New Jersey’s Medicaid waiver for acquired immunodeficiency syndrome. Health Care Financing Rev. 1992;13: 27–44.

6. Panel on Clinical Practices for Treatment of HIV Infection. Guidelines for the use of antiretroviral agents in HIV-infected adults and adolescents. MMWR Morb Mortal Wkly Rep. 1998;47(RR-5):43–82.

23. Peterson ED, Wright SM, Daley J, et al. Racial variation in cardiac procedure use and survival following acute myocardial infarction in the Department of Veterans Affairs. JAMA. 1994;271:1175–1180.

7. Strathdee SA, Palepu A, Cornelisse PGA, et al. Barriers to use of free antiretroviral therapy in injection drug users. JAMA. 1998;280:547–549.

24. Whittle J, Conigliaro J, Good CB, et al. Racial differences in the use of invasive cardiovascular procedures in the Department of Veterans Affairs medical system. N Engl J Med. 1993;329:621–627.

10. Shapiro MF, Morton SC, McCaffrey DF, et al.

29. Stone VE, Clarke J, Lovell J, et al. HIV/AIDS patients’ perspectives on adhering to regimens containing protease inhibitors. J Gen Intern Med. 1998;13: 586–593.

34. Buchanan RJ, Colby D. Medicaid policies for the physicians’ services provided to Medicaid recipients with AIDS. AIDS Public Policy J. 1996;11:47–60.

3. Smith SR, Kirking DM. Access and use of medications in HIV disease. Health Serv Res. 1999;34: 123–144.

9. Palella J, Frank J, Delaney KM, et al. Declining morbidity and mortality among patients with advanced human immunodeficiency virus infection. N Engl J Med. 1998;338:853–860.

28. Diehr P, Yergan J, Chu J, et al. The influence of race on the use of surgical procedures for treatment of peripheral vascular disease of the lower extremities. Arch Surg. 1995;130:381–386.

17. Markson LE, McKee L, Mauskopf J, et al. Patterns of Medicaid expenditures after AIDS diagnosis. Health Care Financing Rev. 1994;15:43–59.

20. Fasciano NJ, Cherlow AL, Turner BJ, et al. Profile of Medicare beneficiaries with AIDS: application of an AIDS casefinding algorithm. Health Care Financing Rev. 1998;19:19–38.

8. Messeri P, Aidala A, Abramson D, Jessop DJ, Nelson K. Socioeconomic differences in the use of HIV combination antiretroviral therapy. Paper presented at: 16th Annual Meeting of the Association for Health Services Research: Research to Action—The Role of Health Services Research; June 1999; Chicago, Ill.

of early-stage breast cancer. Am J Public Health. 1992; 82:195–198.

36. Miller AD. Monitoring the quality of Medicaid managed care. N J Med. 1997;94:33–35. 37. Cunningham WE, Markson LE, Anderson RM, et al. Prevalence and predictors of highly active antiretroviral therapy use in persons with HIV infection in the US. J AIDS. 2001;25(2):115–123.

25. Carlisle DM, Leake BD, Shapiro MF. Racial and ethnic disparities in the use of cardiovascular procedures: associations with types of health insurance. Am J Public Health. 1997;87:263–267. 26. Gornick ME, Eggers PW, Reilly TW, et al. Effects of race and income on mortality and use of services among Medicare beneficiaries. N Engl J Med. 1996; 335:791–799. 27. Satariano ER, Swanson GM, Moll PP. Nonclinical factors associated with surgery received for treatment

