Drug abuse, methadone treatment, and health services use among ...

DEPENDENCE Drug and Alcohol Dependence 60 (2000) 77-89

ELSEVIER

www.elsevier.com/locate/drugalcdep

Drug abuse, methadone treatment, and health services use among injection drug users with AIDS* Usha Sambamoorthi a Institute

a,*

,

Lynn A. Warner a, Stephen

Crystal a,b, James Walkup

a,c

forHealth, Health Care Policy, and Aging Research, Rutgers University, 30 College Avenue, New Brunswick, NJ 08901, USA b School of Social Work, Rutgers University, New Brunswick, NJ 08901, USA ’ Graduate School of Applied Psychology, Rutgers University, New Brunswick, NJ 08901, USA

Received 25 May 1999; received in revised form 3 November 1999; accepted 3 November 1999

Abstract

This paper compares health care use across subgroups of injection drug users (IDUs) with AIDS, as defined by current drug abuse status and participation in methadone maintenance treatment (MMT), using surveillance-identified IDU status and health care claims data. Merged Medicaid and AIDS surveillance data were analyzed using ordinary least squares regression, simple logistic regression and multinomial logistic regression. Consistent MMT was more likely among women, Whites and older subjects. Monthly total expenditures and inpatient expenditures were significantly lower for IDUs in MMT than for IDUs with claims indicative of current drug abuse. Consistent participation in MMT was associated with a higher probability of antiretroviral use and, among antiretroviral users, more consistent use of antiretrovirals. Merged administrative data sets can be an important data source that illuminate the relationships among drug abuse, drug treatment, and HIV-related health care. For AIDS-infected IDUs, consistent MMT may lower barriers to receipt of appropriate HIV-related health care and reinforce adherence to medical recommendations. 0 2000 Elsevier Science Ireland Ltd. All rights reserved. Keytcords:

HIV/AIDS; Drug abuse; Methadone

maintenance

treatment;

1. Introduction

Drug abuse, particularly injection drug use (IDU), is today responsible for a substantial proportion of AIDS cases in North America, Europe, and more recently in regions of the developing world, including South America and Southeast Asia (Selwyn, 1991). According to data from the Centers for Disease Control (CDC), more than one third (36%) of AIDS cases in the United States (US) were IDU-associated (Center for Disease Control 1999a). While national trends suggest the rate of HIV infection from IDU may be stable, it appears to be responsible for an increasing proportion of cases in certain areas, such as San Francisco and New York City (Lemp et al., 1997; Emlet and Gusz, 1998; Tortu et * The findings and opinions reported here are those of the authors and do not necessarily represent the views of any other individuals or organizations. * Corresponding author. Tel.: + l-732-932-8637; fax: + 1-732-9326872. E-mail address: [email protected] 0376~8716/00/$ - see front PII: SO376-8716(99)00142-S

matter

(U. Sambamoorthi)

0 2000 Elsevier

Science Ireland

Medicaid

al., 1998). Given the linkages among HIV infection, IDU, and a host of associated conditions in HIV including tuberculosis (Gollub et al., 1997) and hepatitis C viral infection (Bodsworth et al., 1996) these trends strongly influence costs and utilization patterns. Compared to non-IDUs with HIV, IDU patients have been shown in prior studies to be a high cost subgroup, less likely to use outpatient services (Solomon et al., 1991; Mor et al., 1992; Solomon et al., 1998) but likely to have more frequent and longer hospitalizations (Seage et al., 1993; Stein, 1994; Johnston et al., 1995). Additionally, IDUs with AIDS have been shown in some studies to be less likely than other persons with AIDS to accept antiretroviral (ARV) therapy (Broers et al., 1994), although there is also some evidence suggesting that there are no adherence differences between the two groups once ARV therapy begins (Broers et al., 1994; Jones et al., 1998). Research on utilization patterns of injection drug users (IDUs) with AIDS has been largely based on comparisons between IDUs with AIDS and non-IDUs

Ltd. All rights

reserved

78

U. Sambamoorthi et al. /Drug

and Alcohol Dependence 60 (2000) 77-89

with AIDS; most studies have paid little attention to the variation within the IDU population. A few studies that have focused on variation within the IDU population have tended to dichotomize their samples, either by comparing IDUs with AIDS who continue drug use to those who discontinue drug use, or by comparing IDUs who engage in MMT with those who do not. For example, studies have found that IDUs with AIDS who continue to use drugs are less likely than IDUs who discontinue drug use to participate in clinical trials (Stone et al., 1997), to attend health care appointments (Turner et al., 1995), and to receive ARV therapy (Celentano et al., 1998). Drug treatment has also been shown to be associated with decreased HIV risk-related behaviors (Ball et al., 1988; Rosenbaum et al., 1996; Shore et al., 1996; Marsch, 1998; Metzger et al., 1998) and increased receipt of ARV therapy (Broers et al., 1994; Strathdee et al., 1998). Studies on the impact of methadone maintenance treatment (MMT) have reported that receipt of MMT reduces hospitalization (Cacciola et al., 1998), improves access to health care (Turner et al., 1995) and compliance with tuberculosis treatment (Gourevitch et al., 1996), and increases the percentage of kept health care appointments (Brettle et al., 1994; Lafitte et al., 1998). In a study of ten-year outcomes for IDUs, time spent in institutions was significantly less among consistent MMT recipients than inconsistent recipients (Maddux and Desmond, 1992). Furthermore, availability of MMT at health care clinics has been shown to reduce hospitalization (Newschaffer et al., 1998) and to increase the rate of kept health care appointments for all drug users, whether or not they actually receive methadone themselves (Brettle et al., 1994). These studies demonstrate that continued drug abuse and participation in drug abuse treatment have important public health implications for access to care, as well as for cost of care. However, continued drug use and drug treatment is unlikely to be uniform across all demographic groups. For example, women who are IDUs have been found to have significantly better retention in substance abuse treatment (Schottenfeld et al., 1998) while men who are IDUs have been found to have higher rates of dropout from MMT (Anglin et al., 1987). Among HIV-infected persons, rates of continued drug and alcohol abuse appear to be higher for African Americans than other groups (Brown et al., 1993) and rates of receipt of MMT appear to be lower (Chaisson et al., 1989; Joe et al., 1991; Meandzija et al., 1994). Finally, strong associations between age and treatment adherence have been noted. For example, several studies report that consistent MMT is more likely among older IDUs than younger (Hser et al., 1990-l; MacGowan et al., 1996; Magura et al., 1998) and dropout is more likely for younger than for older persons with AIDS both in MMT programs (Grella et

al., 1994) and in HIV intervention trials (DiFrancesco et al., 1998). To determine what impact MMT may have on patterns and costs of health care among those with AIDS, we examined claims data from 2600 surveillance-identified IDUs with AIDS who received Medicaid benefits in NJ. We compared patterns across five subgroups of IDUs: (1) those who, during the observation period, never received MMT and had no drug abuse claims; (2) those who had consistent MMT without any drug abuse claims; (3) those who had inconsistent MMT without any drug abuse claims; (4) those who had either consistent or inconsistent MMT and have drug abuse claims; and (5) those who never received MMT and had drug abuse claims. Our study addresses the following research questions: 1. What is the rate of MMT among IDUs with AIDS, what proportion experience health care encounters indicative of continued drug abuse, and what are the correlates of drug abuse and MMT statuses? 2. Are there differences in the composition and level of monthly Medicaid expenditures between IDUs receiving MMT and IDUs who continue to abuse drugs and are not in treatment? 3. Adjusting for other factors that affect health care service utilization, are those in consistent MMT more consistent users of ARV therapy than IDUs who continue to abuse drugs and are not in treatment?

2. Data and methods 2.1. Data sources This study is based on adult Medicaid participants who were diagnosed with AIDS in NJ between January 1988 and December 1995. NJ is an important state in which to study this issue, as it is among the top ranking states in number of AIDS cases (Center for Disease Control, 1999b) and ranks first among states in the proportion of HIV-related disease associated with injection drug use (NJ DHHS, 1992). The present study was made possible because of file matching between Medicaid eligibility files and NJ’s AIDS Registry, conducted under a cooperative agreement between NJ State agencies. The link between the AIDS Registry and the Medicaid file was done by Department of Health (DOH) using identifying fields common to both files such as name, birth date, gender, and social security number fields. To perform the database link, DOH used a methodology suitable for the problem of determining database intersections where only partial overlaps are expected. Because the degree of overlap between the two databases is not known, it is difficult to establish exact error rates. A detailed description of


19


a similar match process has been reported by researchers in Maryland (Hidalgo, 1990), where it has been used successfully to report on trends in Medicaid enrollment for persons with AIDS (Bartnyska et al., 1995). Once the link was established, DOH provided the researchers at Rutgers University with a file in which the clients were identified only with numbers to ensure confidentiality. The researchers at Rutgers then combined three sources of data to derive client-level files: AIDS Registry data from the NJ DOH; paid Medicaid claims for medical care and prescription drugs from the NJ Department of Human Services, Division of Medical Assistance and Health Services (DMAHS); and the AIDS Community Care Alternatives Program (ACCAP) client profile file from DMAHS. The researchers also obtained human subjects approval from the Rutgers University Institutional Review Board at the time of designing the project and have since obtained annual approval for the study. 2.1.1. AIDS Registry Demographic characteristics (i.e. gender, race, county of residence at diagnosis), exposure category, date of AIDS diagnosis, vital status and date of death for decedents were obtained from the State’s AIDS Registry. Exposure category was based on drug use history as reported in the AIDS Registry, and patients were classified as either IDUs with AIDS, or non-IDUs with AIDS. 2.1.2. Medicaid paid claims NJ Medicaid covers inpatient and outpatient treatment encompassing 27 service categories including pharmacy, physician, inpatient hospital, outpatient hospital, independent clinic, home health agency, medical supplies and equipment, transportation, laboratory, optical appliance, durable medical equipment, psychology, podiatry, chiropractor and others. Covered substance abuse services include inpatient and outpatient detoxification, methadone maintenance, and support services. The Medicaid program also pays for home-based narcotic and drug abuse treatment through a waiver program available to persons with AIDS. A list of the different types of State plan services and waivered services is documented elsewhere (Merzel et al., 1992). Medicaid claims histories obtained by Rutgers researchers contained all processed claims for services and pharmacy prescriptions provided up to December 1996. To allow for time lags between receiving services, billing, payment, and appearance of paid claims in the computerized database, and because vital status information was available as of March 1996, services received through 291311996 were included in the analyses. The 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.

2.1.3. ACCAP$les Some of the NJ Medicaid population is enrolled in ACCAP, an HIV-specific Medicaid home and community-based care waiver program, which offers case management and private duty nursing, among other services. Participation in the waiver program was determined by matching the claims file against the centrally maintained administrative ACCAP client profile file. 2.2. Study population The study population was comprised of IDUs with AIDS who participated in NJ Medicaid between 1 January, 1988 and 29 March, 1996. Restriction of the analysis to those with AIDS narrows the variability in stage of illness and reduces bias associated with unmeasured differences in disease stage. Additional inclusion criteria were age 18 years or older at the time of AIDS diagnosis and receipt of Medicaid services for at least 90 days without participation in managed care or Medicare programs because claims data may not be complete for such individuals. We identified 2600 beneficiaries who met these criteria. 2.3. Measurement

of methadone

treatment

An outpatient encounter with either a Current Procedural Terminology (CPT) code or HCFA Common Procedure Coding System (HCPCS) code of 22006 was used to identify methadone maintenance visits. Service dates in MMT claims, ranging from a single day to a month of methadone use, were used to determine total number of days on MMT. For each individual, we divided the total number of days represented in the MMT service claim dates by duration of observation (the number of days enrolled in Medicaid after AIDS diagnosis) to determine the proportion of time on MMT. Beneficiaries with MMT claims representing more than 50% of the enrollment time were included in the consistent MMT group. 2.4. Measurement

of drug abuse

Current drug abuse was identified through claimsbased diagnosis codes conforming to the International ClassiJication of Diseases, 9th Revision, Clinical ModiJication (ICD-9-CM), including 304 for drug dependence, 305 for nondependent drug abuse, 292 for drug withdrawal syndrome, 573.3 for dirty needle (hepatitis), and 965 for poisoning by opiates and related derivatives. In addition to diagnosis codes, several procedure codes were included that are indicative of drug-related events, such as drug detoxification (i.e. 94.25, 94.65, 94.66, 94.67, 94.68, and 94.69), counseling for drug addiction (94.45) and referral for drug addiction rehabilitation (94.54). A similar approach was used to identify drug

80

II. Sambamoorthi et al. /Drug


users in New York State Medicaid claims data (Markson et al., 1994). Our procedure classified 32% of Registry-classified IDUs as having indications of current drug abuse. This procedure is probably not highly sensitive (some IDUs whose abuse is less severe may fail to be classified as current drug abusers), but it can be considered specific in that there are likely to be relatively few false positives. The procedure serves to classify the IDU population into groups with current drug involvement (i.e. drug involvement during the period for which Medicaid claims were available) and to provide a basis for examining statistical differences between these two groups. Using the MMT and drug abuse variables described above, each IDU with AIDS was classified into one of five mutually exclusive categories: (1) no MMT and no indication of current drug abuse (n = 1029); (2) consistent MMT and no indication of current drug abuse (n = 276); (3) inconsistent MMT and no indication of current drug abuse (n = 278); (4) any MMT with drug abuse claims (n = 184); and (5) no MMT with indications of current drug abuse (n = 833).

2.6. Measurement of other covariutes Geographical areas of NJ vary widely in HIV/AIDS prevalence and in the extent to which the health care system has responded to the demands of AIDS care. The highest-prevalence area of the state for HIV/AIDS is the five-county area nearest to New York City (NYC), comprising Essex, Hudson, Passaic, Bergen and Union Counties. In our analyses, we included proximity to NYC as a covariate. Access to the home and community based-services provided by ACCAP is associated with different patterns of service utilization given that waiver participants have access to various in-home care services that are not available to traditional Medicaid enrollees (Merzel et al., 1992). Therefore, we also used waiver status as a covariate in all our analyses. Racelethnicity was characterized as White, African American, and Latino. In multivariate analyses, White was used as the comparison group. Based on vital status as of March 1996, respondents were classified as decedents and non-decedents. Because the effects of age are likely to be non-linear, we included the following categories of age at AIDS diagnosis: 18-29 years (the reference group in multivariate models), 30-39 years, 40-49 years, 50 and older.

2.5. Measurement of health services use and costs 2.7. Analytic procedures Expenditure data were based on the actual amount reimbursed by the NJ Medicaid program and were reported in 1996 US dollars based on the national Consumer Price Index for medical care. The claims data on expenditures were aggregated by type of service to compute cumulative costs per beneficiary. Service categories included inpatient, pharmacy, and other non-inpatient services. Because the observation period varied among subjects depending on the date of enrollment, survival time, and the end of the follow-up period, we used average monthly expenditures per person, calculated by dividing the total cost per patient by the number of months of follow-up. Because of the importance of access to ARV therapy in retarding the progression of AIDS, intergroup differences in the proportion of time on ARVs were evaluated using pharmacy claims. National Drug Codes were used to identify ARV drugs in use during the period of observation, which included five nucleoside analogue reverse transcriptase inhibitors (NRTIs) dd1, ddC, 3TC, d4T, and zidovudine (ZDV or AZT). We classified respondents as users and non-users of ARV treatment, based on receipt of at least one prescription for an ARV drug. Among ARV users, we divided the total number of days represented by filled prescriptions by the length of the observation period to determine proportion of time on ARVs for each individual. Methods used to perform these calculations have been previously documented (Crystal et al., 1995).

In the bivariate analysis, IDUs with drug abuse and MMT were compared on demographic characteristics, and significance was tested with chi-square statistics. We employed multinomial logistic regression to examine factors associated with drug abuse and MMT. We tested for subgroup differences in the use of ARVs with chi-square statistics. The statistical significance of bivariate subgroup differences in average monthly cost of care and mean proportion of time on ARVs was evaluated with t-tests. Simple logistic regressions were estimated to predict the probability of ARV use. The results are presented as odds ratios (with 95% confidence intervals). To estimate the effects of drug abuse and MMT on costs, utilization, and proportion of time on ARVs, with controls for other characteristics such as gender and race, we employed ordinary least squares linear regression techniques. Monthly expenditures were transformed to a logarithmic scale to reduce skewness. Effect estimates for continuous independent variables on the log of monthly expenditures can be interpreted as percentage change for each unit of change in the independent variable. The effect of dummy variables in terms of percentage of expenditures can be estimated by exponentiating the regression coefficients of dummy variables and subtracting one (i.e. % change = eb - 1; Halvorsen and Palmquist, 1980). To better control for disease stage, dummy variables for year of AIDS diagnosis from the AIDS Registry were

U. Sambamoorthi

et al. /Drug

and Alcohol Dependence

included in the regression as control variables. Because the results were unaffected by the introduction of diagnosis year, we excluded year of diagnosis from the final regressions. To investigate the possibility that some of the differences in outcomes could be due to unobserved selection factors that bias estimates of differences between IDUs in MMT and IDUs who continue to abuse drugs and are not in MMT, we used a two-stage sample selection procedure (Heckman, 1979). In the first stage, we conducted a probit analysis of the probability of MMT versus current drug abuse without MMT, controlling for gender, race, age, and geographic location. The estimated probabilities were then used to calculate predieted rates of MMT for each sample subject, correcting for the non-random selection of subjects into the MMT program. In the second stage we predicted expenditures based on the MMT participation rate derived in the first stage (inverse Mills ratio), along with the other control variables. Findings from the models controlling for selection bias (available from the first author) were similar to the findings from the models without selectivity correction reported below. Table 1 Subject characteristics

by methadone Sample

‘%I

treatment No current abuse

and current drug

60 (2000) 77-89

81

3. Results

3.1. Variability among IDUs lvith AIDS Table 1 describes the study population and the five IDU subgroups. The majority of our study population were male (64%), and lived in the high-HIV prevalence area of the state near New York City (NYC; 70%). The racial distribution was 63”/0 African American, 20% White and 17% Latino. Median age at AIDS diagnosis was 37, about half falling between the ages of 30 and 39 (51 .l%). A majority of the population (73%) was alive as of March 1996. Of the total number of beneficiaries, 21% received MMT, either consistently or inconsistently and had no claims for drug abuse; 7% had claims indicative of both MMT and current drug abuse. Forty percent of the beneficiaries had no claims indicating either substance abuse or use of MMT, and somewhat fewer (32%) had claims indicative of current drug abuse with no use of MMT. There were differences in gender, racial mix, age at diagnosis, vital status, and geographic region by drug abuse and MMT categories. A signifi-

drug abuse”

Consistent

MMT

Inconsistent MMT

MMT abuse

and current

drug

Current abuse

All

100.0

39.6

10.6

10.7

7.1

32.0

Gender b Male Female

63.6 36.4

42.2 35.0

8.4 14.5

11.3 9.7

6.2 8.7

31.9 32.2

Race/ethnicityb White African American Latin0

19.9 63.4 16.7

38.6 39.6 40.8

17.8 8.7 9.4

9.7 9.6 15.9

6.8 7.2 7.1

27.2 35.0 26.7

diagnosisb 11.9 51.1 31.5 5.5

38.4 36.8 40.8 60.8

7.1 10.0 12.9 10.5

9.7 11.5 10.4 7.0

11.3 7.3 5.6 4.2

33.5 34.3 30.3 17.5

Age ut AIDS 18-29 years 30-39 years 4049 years 50 and over

County of residenceb Near New York City Elsewhere

69.5

36.9

11.5

10.7

8.1

30.3

30.5

45.8

8.6

10.7

4.7

32.8

Waiver participation Non-participants Participants

87.1 12.9

38.9 44.5

10.9 9.0

10.9 9.6

7.5 4.5

32.0 32.5

Vital status as of 29/3/96b Non-decedents 72.7 Decedents 27.3

40.4 37.5

11.3 8.7

11.4 8.9

1.9 4.9

29.0 40.0

1029

276

278

184

833

Number of observations

2600

a The study population is based on adult Medicaid least 90 days between 1 January 1988 and 29 March b Chi-square test; P10.05.

participants 1996.

with AIDS,

age 18 or older, whose utilization

drug

and costs were observed

for at

82


Table 2 Predictors

of current

drug abuse and methadone No current

Gender [Reference: Female

drug abuse

maintenance Consistent


treatmenta MMT

Inconsistent

MMT

MMT and drug abuse

OR

95% CI

OR

95% Cl

OR

95% CI

OR

95% CI

0.97

[0.79, I.191

1.95b

[1.46, 2.591

0.93

[0.69, 1.251

1.16

[0.82, 1.641

0.82 1.10

[0.64, 1.071 [0.79, 1.531

0.29b 0.53b

[0.20, 0.401 [0.34, 0.841

0.74 1.62’

IO.51, 1.091 [1.03, 2.531

0.70 1.06

[0.44, 1. lo] [0.60, I.871

[0.67, 1.221 [0.78, 1.501 [1.72, 5.021

1.75’ 2.94b 4.78b

[1.05, 2.921 [1.72, 5.031 [2.11, 10.81

1.23 1.30 1.61

[0.78, 1.941 [0.79,2.12] [0.69, 3.781

0.77 0.75 1.12

[0.48, 1.231 [0.44, 1.281 [0.41, 3.061

0.77b

[0.62, 0.951

1.51’

[1.08,2.10]

0.95

[0.70, 1.291

1.57c

[1.04, 2.381

1.22

[0.91, 1.621

0.79

[0.50, 1.241

0.96

(0.62, 1.481

0.67

[0.37, 1.211

Male]

Racejethnicity [Reference: White] African American Latin0

Age at AIDS diagnosis [Reference: 18-29 years] 3&39 years 0.90 40-49 years 1.08 50 and over 2.94” Counry of residence Near New York City [Reference: Elsewhere] Waiuer participation [Reference: Non-participants] Participants

Vital status as of 2913196 [Reference: . Non-decedents] Decedents 0.53b

[0.43,0.66]

0.60b

[0.43, 0.841

0.54b

[0.39,0.75]

0.65’

[0.43,0.98]

of follow up 0.97b

[0.96,0.97]

1.01

[l.OO, 1.021

0.99

[0.99, l.OO]

1.03b

[I .03, 1.041

Months

...

..

.

..

. .

a The study population is based on adult Medicaid participants with AIDS, age 18 or older, whose utilization and costs were observed for at least 90 days between 1 January 1988 and 29 March 1996. Odds ratio are estimated from a multinomial logistic regressions on current drug abuse and drug treatment among IDUs. The regressions also include an intercept term. The reference group for drug abuse and treatment categories is ‘current drug abuse’. b PSO.01. = 0.01