D'Aunno T, Pollack HA. Changes in Methadone Treatment PracticesResults From a National Panel Study, 1988-2000. JAMA. 2002;288(7):850-856. doi:10.1001/jama.288.7.850
Author Affiliations: School of Social Service Administration and Department of Health Studies, University of Chicago (Dr D'Aunno), and Department of Health Management and Policy, School of Public Health (Dr Pollack), University of Michigan, Ann Arbor. Dr D'Aunno is now with INSEAD, Fontainebleau, France.
Context Results from several studies conducted in the early 1990s showed that
the majority of US methadone maintenance programs did not use treatment practices
that met established standards for the care of heroin users. Effective treatment
for heroin users is critical given the upsurge in heroin use and the continued
role of injection drug use in the human immunodeficiency virus and hepatitis
Objectives To examine the extent to which US methadone maintenance treatment programs
have made changes in the past 12 years to provide adequate methadone doses
and to identify factors associated with variation in program performance.
Design, Setting, and Participants Program directors and clinical supervisors of nationally representative
methadone treatment programs that varied by ownership (for-profit, public,
or private not-for-profit) and setting (eg, free-standing, hospital-based)
were surveyed in 1988 (n = 172), 1990 (n = 140), 1995 (n = 116), and 2000
(n = 150).
Main Outcome Measures Percentage of patients in each treatment program receiving methadone
dosages of less than 40, 60, and 80 mg/d.
Results The percentage of patients receiving methadone dosage levels less than
the recommended 60 mg/d has decreased from 79.5% in 1988 to 35.5% in 2000.
Results also show that programs with a greater percentage of African American
patients are especially likely to dispense low dosages, while programs with
Joint Commission on Accreditation of Healthcare Organizations accreditation
are more likely to provide adequate methadone doses.
Conclusions Efforts to improve methadone treatment practices appear to be making
progress, but many patients are still receiving substandard care.
How well do the nation's methadone maintenance treatment facilities
meet established standards for best practice care?
This is a critical question for public health policy and, more broadly,
for social policy.1 Changes in methadone treatment
practices are important in responding to the changing prevalence and modalities
of heroin use, to the continued role of injection drug use in the human immunodeficiency
virus (HIV) and hepatitis C epidemics, and to changes in the financing and
regulation of methadone treatment. Assessing methadone treatment practices
is also important in light of recent research results about methadone treatment
An upsurge in heroin use has increased the need for effective treatment
services for opioid dependence. Admissions for opioid dependence have surpassed
admissions for cocaine abuse in the nation's drug abuse treatment programs.2 Population-based epidemiological studies also show
a marked increase in heroin use.3 The Office
of National Drug Control Policy now estimates that the number of heroin users
in the United States has increased by more than half in less than a decade,
from 600 000 in the early 1990s to 980 000 in 2001.4
Analysts agree that increased heroin use is influenced by declining
heroin prices and by increased purity. Increased purity creates additional
concern because this may increase the addictive potential of street heroin.5 Average pure heroin content contained in a $100 purchase
has increased 3-fold between 1988 and 1995 in cities across the United States.5 High purity also facilitates noninjecting forms of
heroin use, such as smoking and snorting, which may attract new users who
are deterred by injection-related risks. Newly available modalities of heroin
use may thus have attracted new users, most of whom are younger than 26 years
of age.2 The long-term risk behaviors of noninjection
heroin users are currently unknown. Public health consequences would be significant
if initial noninjectors frequently convert to needle use.
Injection heroin use also remains central to HIV transmission and to
the transmission of other infectious agents.6
Many studies indicate that methadone treatment is associated with reduced
HIV seroconversion among methadone patients and is also associated with reduced
needle-sharing and other risk behaviors.7,8
A recent Institute of Medicine report concludes, as have previous expert reports,1,8 that effective methadone treatment
for opioid dependence is critical to prevent HIV transmission. The majority
of injection drug users have now contracted hepatitis C. Effective methadone
treatment may be important to slowing this epidemic as well.9,10
Moreover, important changes are occurring in financing and regulation
that may carry significant implications for methadone treatment effectiveness.
One prominent change is the introduction of managed care as a payment and
administrative mechanism for methadone treatment. Although a recent study
indicates that methadone treatment programs are less likely to be involved
in managed care than are their nonmethadone counterparts,11
some 30% of all methadone treatment programs participate in some form of managed
care.11 For those programs and the tens of
thousands of patients they serve, managed care can potentially improve treatment
practices. At the same time, managed care might hinder treatment effectiveness
through practices such as limiting treatment duration, a central factor in
A second prominent change is the pending move to regulate methadone
treatment facilities through the Center for Substance Abuse Treatment (CSAT)
rather than through the Drug Enforcement Agency and the Food and Drug Administration
(FDA). This substantial shift was prompted by concerns about quality of care
in methadone treatment programs.14,15
The CSAT, which is part of the federal Substance Abuse and Mental Health Services
Administration, brings extensive experience working with treatment programs
to improve quality. This regulatory shift may therefore create opportunities
to improve patient care.
Finally, results from prior studies of methadone treatment practices16 and treatment effectiveness17
support the need for a current assessment of treatment practices. Data from
the few national studies of methadone dose levels published in the early 1990s
indicate that patients typically received suboptimal doses.18- 20
Data collected in 1988 from a nationally representative sample of 172 methadone
maintenance facilities showed that the average dosage was 45 mg/d; 1990 data
from the same panel of facilities showed virtually identical average dosages,
46 mg/d.18 Data from a random sample of patient
discharge abstracts (N = 261) at 26 randomly selected treatment facilities
in 1990 showed that the average methadone dosage per day was 50 mg.19 Yet another study of 24 methadone treatment facilities
conducted in 1990 by the General Accounting Office concluded that, on average,
dose levels were too low.20
In response to these study results, several initiatives were launched
to improve treatment practices. The CSAT developed methadone treatment guidelines
and distributed them to state substance abuse agencies and to treatment programs
across the nation.21,22 The Institute
of Medicine convened a study panel that recommended several changes in treatment
practices and their regulation.14 Cooper23 published an editorial encouraging physicians to
use effective dose levels. The National Institute on Drug Abuse funded a study
to examine the development of a quality assurance program for methadone treatment.15 More recently, a National Institutes of Health consensus
panel (1997) produced guidelines for effective methadone treatment practices.1
Subsequent surveys indicate at least partial successes of such efforts.
National panel data indicate that methadone dosages increased significantly
from an average of 45 mg/d in 1988 to 59 mg/d in 1995.16
This increase raised average dosages to the minimum level recommended by national
consensus panels (60 mg/d).1 Other significant
improvements in treatment practices were documented over the same period.
Average time in treatment increased from 20 months to 21 months; the average
upper dosage limit increased from 79 mg/d to 93 mg/d; the percentage of patients
receiving progressively decreasing methadone doses decreased from 34% to 22%.
Programs also waited longer before encouraging patients to detoxify from methadone.
In 1990, only 27% of programs waited more than 1 year to encourage detoxification,
but, by 1995, 55% of programs exceeded the same threshold.
Although these trends reflect important progress, recent studies suggest
that methadone dosages between 80 and 100 mg/d are more effective than are
dosages in the range of 60 to 80 mg/d.17,24
Strain et al17 randomly assigned patients at
an urban treatment program to a moderate (40-50 mg/d) or high (80-100 mg/d)
methadone dose group. After 30 weeks, both patient self-report and urinalysis
indicated significantly higher prevalence of recent opioid use within the
moderate-dose group. There were no differences between the 2 groups in self-reported
adverse effects from methadone use. Thus, documented increases in methadone
dose levels may be insufficient to yield effective treatment outcomes. Given
the current price and purity of heroin, even higher methadone doses may be
It is important to note, however, the long-standing resistance to methadone
maintenance treatment in general, and in particular to high-dose, long-term
treatment.25,26 Some citizens,
drug users, and drug treatment professionals and staff have expressed concern
that methadone treatment "substitutes one addiction for another." This view
is particularly likely to be held by individuals who support an abstinence
model of treatment and recovery. Vaughn27 showed
that the more that clinical supervisors in methadone treatment programs supported
an abstinence model of treatment, the more likely their programs were to provide
low methadone doses.
In sum, the current study adds to prior research16,28
by examining changes in methadone treatment practices using 1988-2000 data
from a nationally representative panel of methadone maintenance programs.
We also examine how important characteristics of these programs (eg, Joint
Commission on Accreditation of Healthcare Organizations [JCAHO] accreditation)
are related to differences in average methadone dose levels.
This study uses data from a panel survey conducted in 1988, 1990, 1995,
and 2000 by the Institute for Social Research at the University of Michigan
and by the National Opinion Research Center at the University of Chicago.29
We define a methadone treatment unit as a physical
facility with resources dedicated specifically to treating opiate dependence
through methadone. Methadone programs with multiple treatment units or multiple
sites were identified, and units were sampled randomly from such programs.
The initial sampling frame consisted of the FDA's 1988 list of the nation's
methadone maintenance treatment units (N = 587). Because the population of
methadone treatment units has changed in the past several years as a result
of the founding of new treatment units, we augmented the 1988 sampling frame
with the FDA's 1998 list of methadone units (N = 871). Because the FDA licenses
all methadone providers, one can identify the entire US population of methadone
treatment units with a relatively high degree of certainty.
Of the 210 units selected for the 1988 study, 172 participated, a response
rate of 82%. Participating units did not differ significantly from nonrespondents
in 1988.18 Thus, in 1990, we contacted only
the participating units from 1988. Of these 172 units, 11 were no longer providing
methadone treatment and thus 161 units were still eligible for the study.
Of the 161 eligible units, 140 (87%) participated in the 1990 study. In 1995,
we contacted the 140 participating units that provided methadone in 1990.
Thirteen of these units were no longer eligible because they had stopped providing
methadone treatment by 1995. Four units that had begun providing methadone
treatment in the interim were added to the 1995 sample. Of the combined 131
eligible units, 116 participated and 15 declined to do so, a response rate
of 89%. Analyses showed no evidence of nonresponse bias due to units that
dropped out from the study from 1988 to 1990 or from 1990 to 1995.16
For the 2000 data collection, we contacted the 116 participating units
from 1995, and of these, 112 agreed to participate, a 97% response rate. In
addition, we added 47 new methadone units to the 2000 sample. These units
were selected at random from among methadone programs that had begun since
1988. The purpose of adding this subsample to the panel units was to ensure
that the 2000 cross-section as a whole was representative of the population
of methadone treatment units; the panel sample alone would not have represented
these newer units. The response rate among these new sample units was 81%
(n = 38). To determine if these new sample units had a significantly different
relationship with the study's dependent variables than the panel units, we
entered a control variable into the regression analyses. This control for
new sample units was not significant in any equations, and we dropped it in
the results reported herein. Thus, the total sample for 2000 was 150 units,
with a combined response rate of 92%.
The unit director and supervisor of clinical services of each participating
unit completed telephone surveys. The survey team followed established methodologies
shown in previous research to maximize data reliability and validity in telephone
surveys.30 Respondents received a worksheet
in advance of our call that informed them of the requested data. This enabled
respondents to consult financial and administrative records prior to the call.
The survey team conducted dozens of internal consistency checks as soon as
the data were collected.
Methadone Treatment Practices Using data from clinical supervisors, we calculated the percentage of
patients in each treatment unit who received dosages that were below 40, 60,
or 80 mg/d. These measures were calculated only for patients who had been
receiving the same methadone dose for at least 2 weeks. They therefore measure
the dose level that units dispense for patients whose dose levels have stabilized.
Patient Characteristics Clinical supervisors reported 4 important characteristics of patient
mix that could be correlated with methadone treatment practices: employment
status (percentage of patients who are currently unemployed), race/ethnicity
(percentage of African American and Hispanic/Latino patients), sex (percentage
of male patients), and average patient age. All of these measures are characteristics
of the mix of patients at the unit level of analysis; we do not have data
from individual patients.
Unit Characteristics We measured several organizational characteristics that could relate
to variation in treatment practices.16,18
Clinical supervisors reported the percentage of staff members who are ex-addicts.
Directors reported unit ownership (public, private for-profit, or private
not-for-profit; we used private not-for-profit as the referent category).
Finally, given recent efforts to improve the quality of methadone treatment
through accreditation, we used data from the unit director to measure JCAHO
accreditation (1 = yes; 0 = no).
Geographic Location and Time Effects Previous research suggests important geographic variation in methadone
treatment practices.16,18 We measured
geographic location using a census division scale (Northeast, Midwest, South,
and West) with the Northeast region serving as the referent category. To capture
time effects, we created dummy variables corresponding to the year in which
survey data were collected, with 1988 serving as the referent year.
This study uses a longitudinal, panel design. This design is most effective
for examining changes in treatment practices over time. Panel data analysis
is complicated, however, by unobserved unit heterogeneity and by potential
nonresponse bias.31- 33
Methadone doses within the same treatment units are likely to be correlated
over time due to unobserved characteristics of these same units. Standard
linear and logistic regression models, which assume independence among observations,
must be modified for panel data analyses. Nonrandom attrition may also bias
To address these concerns, we analyzed the data using a random-effects
specification that accounts for repeated measures over time.16,31
For each dependent variable, a random-effects model uses all data from the
4 waves simultaneously, but allows for the possibility of random unit effects.
Explanatory variables included patient characteristics, unit characteristics,
environmental factors, and 3 dummy variables to model changes in the outcome
variables across 1988, 1990, 1995, and 2000.
Random-effects models are also helpful to mitigate bias due to nonrandom
attrition over time.32,33 In the
presence of attrition, naive analysis using only the data from units that
participated in all 4 surveys can produce biased results.33
All available data from the 4 surveys were used, including data from those
units that did not participate in 1990, 1995, or 2000.
A related advantage of random-effects models is that they help to address
effects due to unobserved unit heterogeneity. Though we measure unit characteristics
in our multivariate model, we did not measure some important characteristics,
such as features of local drug markets and the attitudes of specific payers,
managers, patients, and treatment staff. Because these unmeasured characteristics
are rather stable over time, and because we wish to examine the impact of
observed unit characteristics such as patient mix, our baseline model is the
random-effects regression specification estimated by D'Aunno et al.16
Finally, because this random-effects specification relies on important
statistical assumptions, we also estimated a more restrictive fixed-effects
model and compared the 2 models using a Hausman test. This procedure compared
our random-effects results with a fixed-effects model that allowed for the
possibility that unit characteristics that we did not measure (eg, staff training)
are correlated with the observed covariates. When the random-effects model
is correctly specified, the fixed-effects estimator yields consistent (though
inefficient) point estimates for all time-varying covariates. If point estimates
from the 2 models yield statistically significant differences, unobserved
heterogeneity may bias the random-effects coefficients. Analyses were performed
using Stata 6.0 (Stata Corp, College Station, Tex) and P<.05 was the level of significance.
Table 1 shows mean methadone
dosages by year within the study sample. A substantial decline was observed
between 1995 and 2000 in the prevalence of low doses. By 2000, 64.5% of methadone
patients received dosages that exceeded 60 mg/d, the minimum recommended by
the Institute of Medicine and by consensus panels. Among those patients receiving
dosages below this threshold, most received doses between 40 and 60 mg/d.
However, 13.2% of all methadone patients continued to receive less than 40
mg/d. Only 7.9% of methadone units reported maximum methadone dose levels
below Institute of Medicine guidelines. Although these trends indicate increasing
provider adherence to the dose levels recommended by consensus panels, few
methadone patients appear to receive optimal doses as indicated by recent
research. Only 32.4% of patients receive more than 80 mg/d, the range marked
as optimal by Strain et al.17
Table 2 shows the results
of our random-effects regressions. Controlling for unit characteristics, our
regressions confirm the trends indicated by Table 1. Compared with the 1988 baseline, the percentage of patients
receiving dosages below 40 mg/d declined by 30.20 percentage points by the
year 2000. The percentage of patients below 60 mg/d dropped by more than 42
percentage points. The percentage of patients receiving dosages below 80 mg/d
declined more slowly, but still declined by more than 25 percentage points.
All of these trend results were significant (P<.001).
As in previous analyses,16 units with
a greater percentage of African American patients were especially likely to
dispense low doses. Each percentage point increase in the proportion of African
American patients was associated with a 0.27 percentage point increase in
the proportion of patients receiving doses below the 60 mg recommended minimum.
No statistically or clinically significant differences were observed by patients'
Other unit characteristics were also correlated with methadone dose.
Units with a smaller number of methadone patients were more likely to dispense
low doses. Facilities with JCAHO accreditation were less likely to provide
No differences were observed by for-profit status or by public ownership
status. Units with high percentages of unemployed patients were less likely
than other units to provide low dosages. The percentage of ex-addicts among
unit staffs was not associated with significant differences in methadone dosage.
Although our statistical model includes several variables to capture
treatment practices and case mix of specific units, the variables included
in our model may be correlated with unit characteristics that we did not measure,
such as staff training, that may influence methadone dose levels. We examined
this possibility by estimating a fixed-effects model. When we compare the
2 models, we find nearly identical results for yearly trends and for the estimated
effects of number of methadone patients.
The effect of African American case mix is also smaller (though less
precisely measured) in the fixed-effects specification. These results suggest
that units that treat a large proportion of African American patients have
characteristics associated with lower doses. For example, perhaps counselors
in units that treat a high percentage of African American patients are not
trained adequately about proper dose levels. Or, perhaps there is more turnover
of staff members in such units that prevents them from establishing relationships
with patients that would support higher doses.
A Hausman test formally supports these possible explanations. This procedure
rejects the similarity of the 2 models (P<.004),
indicating that unit characteristics that we did not measure are correlated
with observed covariates in our model. Comparing specific point estimates,
we find quite similar year effects in the fixed-effects and random-effects
specifications. In our fixed-effects specification, we found that private
for-profit status was associated with lower doses. We also found weaker effects
for JCAHO accreditation, supporting the interpretation that units with more
effective practices have self-selected into the JCAHO-accredited group.
Because current variations in methadone dosage practices are most important
for public policy, we also examined the 2000 cross-section (Table 3). Within this 2000 cross-section, JCAHO accreditation displayed
the largest correlation with methadone dose. Patients in JCAHO-accredited
units were 10 percentage points more likely to receive doses above the minimum
recommended guidelines and were also significantly more likely to receive
dosages exceeding 80 mg/d.
In additional analyses (data not shown), we identified several other
variables correlated with methadone dose. Units that predominantly serve drug
users younger than 30 years of age or that rapidly seek to "detoxify" patients
from methadone were more likely than other units to dispense lower dosages.
In addition, units whose directors believed "strongly" or "very strongly"
in 12-step programs were more likely to give dosages below 40 mg/d.
Results from this panel study indicate considerable progress in meeting
consensus guidelines for methadone treatment. Two thirds of all patients now
receive dosages above 60 mg/d. Treatment practices in the year 2000 contrast
strongly with 1988, when 80% of patients received dosages below this level.
Despite these trends, only 32.4% of patients are receiving dosages above
80 mg/d, the level indicated as optimal in recent randomized trials. Increasing
dosage levels to 80 mg/d may be the most pressing current issue, in light
of recent clinical evidence and in light of increased heroin purity of street
drugs. In fact, the increase in dose levels we have observed may reflect increased
purity: treatment providers may have increased dosage in response to more
severe abuse or dependence among their patients. Treatment providers may remain
one step behind if they do not continue to increase dosages to the 80 mg/d
level. At a minimum, individuals who have responsibility and authority for
determining dose levels should conduct a review of policies and practices
to ensure that dose levels are tailored to meet the needs of individual patients.
We expect, however, that increasing dosage levels to 80 mg/d will be
difficult in many programs, especially those that hold strong abstinence orientation
or those whose patients or staff are otherwise ambivalent regarding the clinical
or values trade-offs inherent to methadone use. It is also true that some
patients can do well on lower doses of methadone; "more" is not always better.
Because this study did not collect outcome data from patients, we are not
able to examine this possibility.
Further, some drug treatment professionals and staff (and some patients)
do not fully support the use of methadone to treat heroin use disorders. Others
do not support the principle that methadone is an appropriate long-term solution
to heroin dependence or abuse. Perhaps these programs should consider alternative
therapies such as buprenorphrine, shown in clinical trials to be effective
when properly dispensed.34Alternative medications
do not carry the stigma associated with methadone, nor do they carry the symbolic
importance of methadone within the drug treatment community and broader society.
Another critical result is that units that treat predominately African
American patients continue to provide low doses. Within-unit changes in the
percentage of African American patients had smaller association with dosage
than did comparisons across treatment units with different proportions of
African American patients. This pattern of results—indicated formally
through comparison of fixed-effects and random-effects specifications—indicates
that these low doses reflect characteristics of treatment programs that are
widely used by African American patients. It is possible that African American
patients are receiving treatment at programs that have underlying weaknesses
in resources, including funds and staff, that make them less likely to keep
up with best practices. These programs are located disproportionately in urban
settings that have difficulty attracting and retaining well-educated and well-motivated
staff members. Further, these programs are not likely to have funds to spend
on training or educational sessions that could improve their practices.
These patterns are of special concern, since adherence to methadone
dose recommendations may be correlated with other aspects of the quality of
care. Further research on quality of treatment available to African Americans
is warranted, especially in light of the high HIV incidence and prevalence
among African American injection drug users.
Finally, the results indicate that JCAHO accreditation seems to play
an important role in adherence to consensus guidelines. Because JCAHO does
not set explicit standards for methadone dose levels, we suspect that the
higher doses that accredited units give to patients reflect some of their
underlying characteristics. Specifically, one possible reason that units with
JCAHO accreditation are more likely to provide higher methadone doses is that
such units have more adequate resources overall, including staff and funds
Additional research is needed to understand how JCAHO accreditation
affects conformity to consensus guidelines. But the relationship between accreditation
and adequate dosage practices may suggest directions for the CSAT to pursue
as it begins its regulatory control of the nation's methadone treatment programs.