Drug and Alcohol Dependence 112 (2010) 99–106
Contents lists available at ScienceDirect
Drug and Alcohol Dependence
journal homepage: www.elsevier.com/locate/drugalcdep
Profiling multiple provider prescribing of opioids, benzodiazepines, stimulants,
and anorectics
B.L. Wilsey a,b,∗ , S.M. Fishman b , A.M. Gilson c , C. Casamalhuapa d , H. Baxi d , H. Zhang e , C.S. Li f
a
VA Northern California Health Care System, Sacramento, CA 95655, United States
Department of Anesthesiology and Pain Medicine, UC Davis Medical Center, Sacramento, CA 95817, United States
Pain & Policy Studies Group, Paul P. Carbone Comprehensive Cancer Center, University of Wisconsin School of Medicine and Public Health, Ann Arbor, MI 53711, United States
d
Clinical and Translational Science Center, UC Davis Health System, Sacramento, CA 95817, United States
e
Department of Statistics, University of California, Davis, CA 95616, United States
f
Department of Public Health Sciences, Division of Biostatistics, University of California, Davis, CA 95616, United States
b
c
a r t i c l e
i n f o
Article history:
Received 1 December 2009
Received in revised form 12 May 2010
Accepted 24 May 2010
Available online 20 June 2010
Keywords:
Drug prescriptions
Narcotics
Benzodiazepines
Central nervous system stimulants
Appetite depressants
a b s t r a c t
Background: The main objective of this study was to determine the prevalence of multiple providers for
different controlled substances using the largest electronic prescription monitoring program (PMP) in
the United States. A secondary objective was to explore patient and medication variables associated with
prescriptions involving multiple providers. PMPs monitor the final allocation of controlled substances
from pharmacist to patient. The primary purpose of this scrutiny is to diminish the utilization of multiple
providers for controlled substances.
Methods: This is a secondary data analysis of the California PMP, the Controlled Substance Utilization
Review and Evaluation System (CURES). The prevalence of multiple provider episodes was determined
using data collected during 2007. A series of binomial logistic regressions was used to predict the odds
ratio (OR) of multiple prescriber episodes for each generic type of controlled substance (i.e., opioid,
benzodiazepine, stimulant, or diet pill (anorectic) using demographic and prescription variables.
Results: Opioid prescriptions (12.8%) were most frequently involved in multiple provider episodes followed by benzodiazepines (4.2%), stimulants (1.4%), and anorectics (0.9%), respectively. The greatest
associations with multiple provider episodes were simultaneously receiving prescriptions for different
controlled substances.
Conclusions: Opioids were involved in multiple provider prescribing more frequently than other controlled substances. The likelihood of using multiple providers to obtain one class of medications was
substantially elevated as patients received additional categories of controlled substances from the same
provider or from multiple practitioners. Polypharmacy represents a signal that requires additional vigilance to detect the potential presence of doctor shopping.
Published by Elsevier Ireland Ltd.
1. Introduction
The international effort to control the quandary posed by drug
addiction and dependence had been considered a notable success
(Musto and Korsmeyer, 2002). Legislation was implemented in the
United States as the Controlled Substances Act (CSA) conforming
to international treaties (Gilson, 2010). These controls established
provisions for the identification of new medications with abuse
liability, labeling products with significant abuse potential, and
controlling the manufacture and distribution of such products in
∗ Corresponding author at: Pain Academic Office, UC Davis Medical Center, Ellison
Ambulatory Care Center, 4860 Y Street, Suite 3200, Sacramento, CA 95817, United
States. Tel.: +1 916 734 7836; fax: +1 916 734 6827.
E-mail address: blwilsey@ucdavis.edu (B.L. Wilsey).
0376-8716/$ – see front matter. Published by Elsevier Ireland Ltd.
doi:10.1016/j.drugalcdep.2010.05.007
medicinal use (Wright et al., 2008). However, newer methods of
control are being advocated as the non-medical use of prescription
medications has placed a considerable burden on society (Gilson
and Kreis, 2009). Many regulatory authorities are now attempting to minimize the abuse of controlled substances by establishing
prescription monitoring programs (PMPs). The most common PMP
goals involve education, delivering information, preventing diversion, and investigating abuse (Joranson et al., 2002; Kasprak, 2003).
PMPs originally operated by requiring healthcare practitioners
to transmit paper copies of prescriptions to law enforcement or
government health agencies. In the last decade, PMPs have utilized computerized monitoring called “electronic data transmission
(EDT) systems,” which is a more efficient type of PMP for controlling prescription medication diversion (Brushwood, 2003; Fishman
et al., 2004). Reports are provided, either solicited or unsolicited,
to medical providers alerting them of prescription acquisitions
100
B.L. Wilsey et al. / Drug and Alcohol Dependence 112 (2010) 99–106
potentially indicating medication misuse. Electronic surveillance
is intended to reduce the utilization of multiple providers, also
pejoratively known as “doctor shopping,” when individuals engage
prescribers who, unbeknownst to each other, write additional controlled substance prescriptions. The individual often visits several
pharmacies, receiving more of a medication than intended by any
single provider.
Despite the growing prevalence of prescription monitoring
systems, there is very little empirical data on the prevalence of multiple provider episodes for controlled substances. In a recent study,
PMP data were utilized to retrospectively analyze doctor shopping
for buprenorphine in France (Pradel et al., 2009). The authors found
that 6% of patients utilized multiple providers in 2005. Similarly,
a study in the United States found that over 93.1% of individuals
had no early refills for Schedule II opioids during 2006 using the
Massachusetts PMP database (Katz et al., 2009). To address this
question further, we analyzed data from the California Department
of Justice (DOJ) PMP database, the Controlled Substance Utilization
Review and Evaluation System (CURES). This program began collecting prescription information related to Schedule II and Schedule
III opioids in 1998 and 2005, respectively. Stimulant and benzodiazepine prescription data were added to the registry by legislative
mandate in 2007. This study also examined the extent that variables related to patient (i.e., age, gender, and geographic location)
and medication (e.g., amount and duration of prescribing, and
the co-prescribing of different controlled substances) characteristics were associated with the occurrence of multiple provider
episodes.
2. Methods
2.1. Administrative issues
A member of the investigative team was granted access to the CURES database
following a background check by the California DOJ. SQL ServerTM was used within
the DOJ to de-identify the database using a record linkage methodology (Newcombe,
1988) to permit identification of sequential prescriptions for each patient. Unique
computer-generated identifiers were devised for each provider and pharmacy. As
a result, no patient, provider, or pharmacy-level information was contained in this
report.
The Institutional Review Board of the University of California-Davis Medical Center and the VA Northern California Health Care System Research and Development
Committee granted approvals to conduct this research.
2.2. Study sample: CURES prescriptions
The data sets for the different controlled substances contained identical fields
corresponding to the information obtained from the CURES database. All entries
from January 1, 2007 through December 31, 2007 were used for the analyses. This
represented all prescriptions written in California for Schedule II, Schedule III, or
Schedule IV controlled substances as reported to the Department of Justice and
entered into the CURES database on a weekly basis by the dispensing pharmacy or
clinic (California Health and Safety Code Section 11165, 2010). Just under 4% of the
original prescriptions (1,117,088 out of 28,893,404) were excluded for the following
reasons: (a) missing or incomplete patient or provider identification, (b) implausible
prescriptions, such as prescription of the same formulation recorded for the same
time interval or transactions, (c) commercial transactions whereby the prescription
was written for more than 700 pills or 50 patches in a given 30-day period (personal
communication: Carla Watkins, Bureau of Narcotic Enforcement, Controlled Substance Utilization Review and Evaluation System, California Department of Justice)
and (d) use of medications not suggestive of standard delivery systems employed by
most patients (e.g., rectal suppositories, intravenous preparations, syrups, solutions,
etc.).
2.3. Study design
A case–control study design was utilized to identify all California residents
who used more than one provider for opioids, benzodiazepine, and amphetamine
prescriptions, and then randomly selected control cases who did not receive prescriptions from multiple providers for these same drug classes. The amphetamine
category was further subdivided into stimulants and anorectics (i.e., diet pills) to
determine if the results varied between these two classes. Non-amphetamine stimulants (e.g., methylphenidate and modafinal) and anorectics (e.g., phentermine and
phendimetrazine) were included for comparative purposes.
Table 1
Controlled substances in 2007 CURES database.
Rx count
Opioids
Hydrocodone
Codeine
IR Oxycodone
CR Oxycodone
Fentanyl Patch
CR Morphine
Methadone
Hydromorphone
IR Morphine
Fentanyl lozenges or tablets
Meperidine
Levorphanol
Benzodiazepines
Alprazolam
Lorazepam
Clonazepam
Diazepam
Temazepam
Triazolam
Flurazepam
Oxazepam
Chlordiazepoxide
Clorazepate & Comb.
Estazolam
Quazepam
11,888,536
1,908,222
1,090,311
543,190
517,618
455,026
328,917
210,708
80,286
32,562
13,837
2,800
2,344,024
2,299,623
1,404,641
1,168,350
1,056,685
178,543
109,164
73,830
55,625
28,916
10,858
497
Stimulants
Methylphenidate
Amphetamine & Comb.
Modafinil
Dexmethylphenidate
Dextroamphetamine
Lisdexamfetamine
Methamphetaminea
657,515
590,601
189,957
83,145
81,465
19,100
2,438
Anorectics
Phentermine
Diethylpropion
Phendimetrazine
Benzphetamine
297,135
32,029
17,356
2,803
a
Methamphetamine hydrochloride (Desoxyn), a Schedule II amphetamine for
ADHD.
2.3.1. Dependent variable: multiple provider episodes. The database was searched
for multiple provider episodes, which was defined when an individual received a
prescription for the same medication from two or more practitioners filled by two or
more pharmacies within a 30-day period. This recognized the potential for patients
to use multiple providers because they either (1) substituted clinicians, (2) obtained
medications from a practitioner covering for the patient’s customary provider, or
(3) received treatment from another practitioner that could be entirely appropriate
(dentist, emergency room doctor, etc.). This assumed that the pharmacy would act
as a “gatekeeper” and not knowingly allow multiple provider episodes.
2.3.2. Independent variables.
2.3.2.1. Numbers and types of controlled substances. In 2007, the CURES database
contained prescriptions for 17,070,418 opioids, 8,729,561 benzodiazepines,
1,624,156 stimulants, and 349,212 anorectics. The prescriptions represented generic
types of 12 opioids, 12 benzodiazepines, 6 stimulants, and 4 anorectics (Table 1).
2.3.2.2. Socio-demographic variables. Socio-demographic variables include age,
gender, and geographic location. Geographic location was stratified into six ordinal categories by separating California counties into groups, based upon published
population estimates (California State Association of Counties, 2007), so that rurality could be defined in terms of the Office of Management and Budget’s definition
of metropolitan and nonmetropolitan populations (Hart et al., 2005; Ricketts et al.,
1998). In this manner, we could determine the impact of varying geographic sizes on
the occurrence of multiple provider episodes for each class of controlled substances.
2.3.2.3. Amount and duration of prescribing. Differences in risk associated with the
amount of medication dispensed (i.e., amount of prescribed medication multiplied
by the number of pills or patches dispensed) were represented using a continuous
variable. Total doses for opioids, benzodiazepines, stimulants, and anorectics were
B.L. Wilsey et al. / Drug and Alcohol Dependence 112 (2010) 99–106
converted into morphine (Korff et al., 2008; Vieweg et al., 2005), diazepam (Giannini,
1997), dextroamphetamine (Bridges, 2007) or phentermine (WebMD, 2008) equivalents, respectively permitting an omnibus comparison of the effect of a unit change
in dose on the dependent variable.
Duration of prescribing (in days) was a continuous variable used to examine
the effect of longer treatment durations. This variable was calculated by finding
the difference between the date of the first controlled substance prescription and
the end date of the last prescription. The “last observation carried forward method”
(Mallinckrodt et al., 2001; Streiner, 2002) was used to impute the interval for the last
prescription because the second anchor (i.e., end date) was nonexistent. In instances
in which only one prescription was provided to a patient, the duration of treatment
was assumed to be 30 days. A special case was made when 30 or less pills of a shortacting opioid (e.g., codeine, or hydrocodone- or oxycodone-combination products)
were prescribed only once, where the duration of treatment was assumed to be
5 days. This assumption derives from qualitative research suggesting that opioids
are provided in emergency rooms as a bridge prescription pending referral to an
office-based physician (Wilsey et al., 2008), as well as the observation in the CURES
database that nearly all of the single prescriptions of controlled substances with 30
or less pills were for these short-acting opioids.
2.3.2.4. Simultaneous prescribing of different controlled substances. This study sought
to determine the impact of the simultaneous prescribing of different classes of controlled substances on the likelihood of multiple provider episodes for an individual
controlled substance class. In the case of opioids, the following categorization was
used:
• No simultaneous prescribing over a 30-day period of either benzodiazepines or
stimulants/anorectics,
• simultaneous prescribing of benzodiazepines only,
• simultaneous prescribing of stimulants/anorectics only, and
• simultaneous prescribing of both stimulants/anorectics and benzodiazepines.
Similar categorizations were performed for the other three classes of controlled
substances.
2.3.2.5. Simultaneous multiple provider prescribing for different controlled substances.
This study also examined the association between multiple provider prescribing for
one controlled substance class and the occurrence of multiple provider episodes
involving different classes of controlled substances within 30 days after a prescription is dispensed. Operationalizing this independent variable reflects a similar
multinomial categorization to the one described above for the simultaneous prescribing of other medications.
2.4. Statistical analyses
Control and multiple provider cases were drawn from the CURES database using
a random sampling procedure to reduce the size of the data used for statistical
analysis, to ensure a minimum of 20 cases in all categories of each independent
variable, and to maintain adequate power (Taylor, 1986). The random sample
that was selected for opioids was 101,651 prescriptions, which included 1651
records of multiple provider prescribing of both stimulants/anorecticss and benzodiazepines. The sample size selected for benzodiazepines was 101,445 prescriptions
and contained 1445 records of multiple provider prescribing of both opioids and
stimulants/anorectics. Each sample of amphetamines comprised 100,000 prescriptions.
A binomial logistic regression was used to predict the odds ratio (OR) of multiple prescriber episodes for each generic type of controlled substance (i.e., opioids,
benzodiazepines, and stimulant or anorectics). For each category of controlled substance, the estimated OR measures the odds by which 30 days of treatment with
that medication alters the risk of multiple provider episodes relative to a patient
who obtained the same medication but did not utilize multiple providers.
Statistical analyses were performed using the R statistical software package
(http://cran.us.r-project.org/). A p-value < 0.05 was considered statistically significant.
101
3.2. Regression analysis
3.2.1. Opioids (Table 2). With the exception of codeine, most of
the opioids studied had a higher involvement in multiple provider
episodes than hydrocodone. The opioids with the largest association with multiple provider prescribing were hydromorphone and
fentanyl lozenges or buccal tablets while codeine and methadone
demonstrated the smallest positive associations. Younger age was
associated with the use of multiple providers. Specifically, there
was approximately a 1% decrease in the odds of this behavior for
every year of age. No difference was found between genders. Relative to metropolitan areas with greater than 1 million people,
smaller populated counties were associated with a decreased likelihood of having multiple prescribers.
Although the medication amount in morphine equivalents and
the duration of prescribing were significantly associated with the
probability of a multiple provider episode, the point estimates of
these variables were negligible. Compared to individuals receiving only a single prescribed controlled substance, there was an
approximate doubling of the OR for multiple provider prescribing if another controlled substance class was being prescribed.
More striking was the 13-fold increase in the OR if both a stimulant/anorectic and benzodiazepine were prescribed along with
opioids during the same 30-day window of observation (compared
to patients with multiple providers who were prescribed opioids
and no additional controlled substance). Furthermore, individuals
who obtained stimulants/anorectics and/or benzodiazepines from
multiple providers were associated with an elevated likelihood of
being involved in this activity to procure opioids (a 10–21-fold
increase in the OR).
3.2.2. Benzodiazepines (Table 3). With the exception of chlordiazepoxide and estazalam, all of the benzodiazepines had less
association with multiple provider episodes than diazepam. Older
age was associated with a lower incidence of multiple providers,
with an almost 2% decrease in the odds of this behavior for every
year of age. Males were 10% more likely than females to obtain
prescriptions from two or more providers over a given 30-day
period. As with opioids, multiple provider episodes involving benzodiazepines were more prevalent in geographic areas with the
largest populations.
Both medication amount and duration of prescribing were associated with multiple provider episodes. Again, these independent
variables had an inconsequential effect on the odds of a multiple provider episode. Compared to those patients not receiving
simultaneous prescriptions of controlled substances, there was a
doubling to tripling of the OR for multiple provider prescribing
if another controlled substance class was being prescribed. More
striking was the 25-fold increase in multiple prescribing for benzodiazepines if stimulants/anorectics and opioids were prescribed to
the same patient during the same 30-day window of observation.
As with opioids, individuals involved in multiple provider prescribing of benzodiazepines were more likely to be involved in this
type of activity to attain controlled substances from other classes
(13–38-fold increase).
3. Results
3.1. Descriptive statistics
Overall, the incidence of doctor shopping, as defined in this
study, was 8.4%. Opioid prescriptions (12.8%) were most frequently
involved in multiple provider episodes followed by benzodiazepines (4.2%), stimulants (1.4%), and anorectics (0.9). These
percentages represent the proportion of prescriptions from each
controlled substance class written by two or more practitioners
filled by two or more pharmacies within a 30-day period.
3.2.3. Stimulants (Table 4). The largest association with multiple
provider prescribing was that of the reference category, dextroamphetamine. Dexmethylphenidate and modafinil were associated
with the largest negative association, representing the least multiple provider episodes. Older age was associated with an increased
likelihood of using multiple providers to obtain stimulants. Specifically, there was an approximate 1% increase in the OR of this
behavior for every year of age. Male patients were 23% less likely
than female patients to obtain a prescription for stimulants from
two or more providers over any 30-day period. When examining
102
B.L. Wilsey et al. / Drug and Alcohol Dependence 112 (2010) 99–106
Table 2
Opioid associations with independent correlates of multiple provider episodes (only statistically significant parameter estimates are shown).
Independent variables
Type of opioid (reference category is hydrocodone)
Methadone
CR Morphine
CR Oxycodone
Codeine
Fentanyl lozenges/buccal tablets
Hydromorphone
IR Morphine
IR Oxycodone
Estimate of
coefficient
0.150
0.183
0.361
−0.401
0.562
0.748
0.259
0.312
Patient demographics
Age
−0.011
Population (reference category is > 1,000,000 population)
<20,000
−0.570
20,000–50,000
−0.536
50,000–100,00
−0.367
100,000–500,000
−0.195
500,000–1,000,000
−0.154
Prescriptions
Medication amount
Duration of treatment
0.000021
0.006
SE
OR
95% CI for OR
0.064
0.055
0.048
0.034
0.178
0.067
0.126
0.035
1.162
1.201
1.436
0.670
1.755
2.113
1.295
1.366
(1.025, 1.319)
(1.079, 1.336)
(1.307, 1.578)
(0.626, 0.716)
(1.238, 2.486)
(1.854, 2.408)
(1.012, 1.657)
(1.275, 1.464)
0.001
0.989
(0.988, 0.990)
0.194
0.096
0.061
0.025
0.027
0.566
0.585
0.693
0.823
0.857
(0.386, 0.828)
(0.485, 0.706)
(0.615, 0.780)
(0.783, 0.865)
(0.813, 0.904)
0.000004
0.000092
1.00002
1.006
Simultaneous prescribing of different controlled substances (reference category is no simultaneous Rx)
Opioids and benzodiazepines (BZD)
0.847
0.022
Opioids and stimulants/anorectics
0.583
0.092
Opioids, BZD, and stimulants/anorectics
2.574
0.045
2.333
1.791
13.117
(1.00001, 1.00003)
(1.006, 1.006)
(2.236, 2.435)
(1.495, 2.146)
(12.008, 14.328)
Simultaneous multiple provider episodes of different controlled substances (reference category is no simultaneous multiple provider episodes)
Opioids and BZD
2.743
0.054
15.541
(13.982, 17.274)
Opioids and stimulants/anorectics
2.357
0.342
10.557
(5.403, 20.626)
Opioids, BZD, and stimulants/anorectics
3.063
0.057
21.400
(19.134, 23.933)
the effects of population by county, counties with 20,000–50,000
inhabitants had slightly more than one-half the OR for multiple
provider episodes as counties with greater than a million inhabitants.
The medication amount, expressed as dextroamphetamine
equivalents, was not associated with multiple provider episodes.
The duration of prescribing, although statistically significant, was
only minimally associated with the probability of a multiple
Table 3
Benzodiazepine associations with independent correlates of multiple provider episodes (only statistically significant parameter estimates are shown).
Independent variables
Estimate of coefficient
Type of benzodiazepine (reference category is diazepam)
Alprazolam
−0.161
Chlordiazepoxide
0.360
Clonazepam
−0.201
Clorazepate & Comb.
−1.156
Flurazepam
−0.479
Lorazepam
−0.313
Oxazepam
−0.512
Temazepam
−0.398
Triazolam
−0.355
SE
OR
95% CI for OR
0.046
0.148
0.051
0.383
0.155
0.047
0.187
0.059
0.117
0.851
1.433
0.818
0.315
0.619
0.731
0.600
0.672
0.701
(0.778, 0.931)
(1.072, 1.916)
(0.740, 0.905)
(0.148, 0.667)
(0.457, 0.839)
(0.666, 0.802)
(0.416, 0.864)
(0.598, 0.754)
(0.557, 0.882)
Patient demographics
Age
−0.018
Gender (reference category is females)
0.099
Population (reference category is > 1,000,000 population)
20,000–50,000
−0.553
50,000–100,00
−0.626
100,000–500,000
−0.166
500,000–1,000,000
−0.134
0.001
0.030
0.982
1.104
(0.980, 0.984)
(1.041, 1.172)
0.186
0.123
0.042
0.043
0.575
0.535
0.847
0.874
(0.400, 0.829)
(0.420, 0.680)
(0.780, 0.920)
(0.803, 0.952)
Prescriptions
Medication amount
Duration of treatment
0.00003
0.00016
1.0004
1.004
0.00039
0.004
Simultaneous prescribing of different controlled substances (reference category is no simultaneous Rx)
Benzodiazepines (BZD) and opioids
1.245
0.033
BZD and stimulants/anorectics
0.815
0.121
BZD, stimulants/anorectics, and opioids
3.223
0.051
3.472
2.259
25.103
(1.0004, 1.0005)
(1.004, 1.005)
(3.256, 3.704)
(1.782, 2.865)
(22.733, 27.721)
Simultaneous multiple provider episodes of different controlled substances (reference category is no simultaneous multiple provider episodes)
BZD and opioids
2.568
0.040
13.039
(12.065, 14.091)
BZD and stimulants/anorectics
3.025
0.312
20.601
(11.166, 38.007)
BZD, stimulants/anorectics, and opioids
3.643
0.058
38.219
(34.140, 42.785)
B.L. Wilsey et al. / Drug and Alcohol Dependence 112 (2010) 99–106
103
Table 4
Stimulant associations with independent correlates of multiple provider episodes (only statistically significant parameter estimates are shown).
Independent variables
Estimate of coefficient
SE
Type of stimulant (reference category is dextroamphetamine)
Dexmethylphenidate
−0.824
Methylphenidate
−0.536
Modafinil
−0.354
Patient demographics
Age
0.012
Gender (reference category is females)
−0.254
Population (reference category is > 1,000,000 population)
500,000–1,000,000
−0.246
Prescriptions
Duration of treatment
0.005
OR
95% CI for OR
0.170
0.105
0.121
0.439
0.585
0.702
(0.314, 0.612)
(0.476, 0.719)
(0.554, 0.890)
0.001
0.050
1.012
0.776
(1.010, 1.014)
(0.704, 0.855)
0.077
0.782
(0.672, 0.910)
0.00034
1.005
(1.005, 1.006)
2.659
2.758
4.000
(2.278, 3.103)
(2.359, 3.225)
(3.380, 4.732)
Simultaneous prescribing of different controlled substances (reference category is no simultaneous Rx)
Stimulants and benzodiazepines (BZD)
0.978
0.079
Stimulants and opioids
1.015
0.080
Stimulants, BZD, and opioids
1.386
0.086
Simultaneous multiple provider episodes of different controlled substances (reference category is no simultaneous multiple provider episodes)
Stimulants and BZD
2.977
0.136
19.623
(15.043, 25.597)
Stimulants and opioids
2.222
0.102
9.229
(7.550, 11.282)
Stimulants, BZD, and opioids
3.289
0.208
26.825
(17.861, 40.287)
provider episode. An approximate doubling of the OR for multiple provider prescribing was observed if another controlled
substance class was being prescribed, rather than when patients
were not simultaneously prescribed controlled substances. A quadrupling in the OR for multiple provider prescribing occurred if all
three classes of controlled substances (i.e., stimulants/anorectics,
benzodiazepines, and opioids) were prescribed during the same
timeframe. As with opioids and benzodiazepines, individuals
involved in multiple provider prescribing of stimulants were more
likely to be involved with obtaining controlled substances from
other classes (9–26-fold).
3.2.4. Anorectics (Table 5). Benzphetamine and phendimetrazine
had a higher involvement in multiple provider episodes than the
comparison medication, phentermine. Older age was more likely to
be associated with the use of multiple providers involving anorectics. Specifically, there was an approximate 1% increase in the OR
of this behavior for every year of age. The other demographic
variables, gender and population, did not achieve statistical significance.
Medication amount, expressed as phenteramine equivalents,
was not associated with multiple provider episodes. Although
the duration of prescribing achieved statistical significance, its
association with multiple provider prescribing of anorectics was
inconsequential. Compared to those patients not receiving simultaneous prescriptions for controlled substances, there was more
than a doubling of the likelihood for multiple provider prescribing
if another controlled substance class was being prescribed concurrently with an anorectic. A quadrupling in the OR occurred when all
three classes of controlled substances (i.e., stimulants/anorectics,
benzodiazepines, and opioids) were being prescribed during the
same 30-day timeframe. As with all other controlled substance
classes in the study, individuals involved with multiple provider
prescribing of anorectics were more likely to acquire controlled
substances from each of the other classes (10–27-fold increase).
4. Discussion
The present results suggest that opioids are more commonly
involved in multiple provider events than are other controlled sub-
Table 5
Anorectic associations with independent correlates of multiple provider episodes (only statistically significant parameter estimates are shown).
Independent variables
Estimate of
coefficient
Type of diet amphetamines (reference category is phentermine)
Benzphetamine
1.589
Phendimetrazine
0.652
SE
OR
95% CI for OR
0.190
0.121
4.900
1.919
(3.375, 7.111)
(1.514, 2.433)
Patient demographics
Age
0.009
Population (reference category is > 1,000,000 population)
<20,000
0.145
100,000–500,000
−0.208
0.003
1.009
(1.004, 1.014)
0.712
0.092
1.156
0.812
(0.286, 4.669)
(0.679, 0.972)
Prescriptions
Duration of treatment
0.000352
1.006
(1.005, 1.007)
2.322
2.811
4.024
(1.838, 2.933)
(2.359, 3.349)
(3.222, 5.026)
0.006
Simultaneous prescribing of different controlled substances (reference category is no simultaneous Rx)
Anorectics and benzodiazepines (BZD)
0.843
0.119
Anorectics and opioids
1.033
0.089
Anorectics, BZD, and opioids
1.392
0.113
Simultaneous multiple provider episodes of different controlled substances (reference category is no simultaneous multiple provider episodes)
Anorectics and BZD
2.297
0.247
9.945
(6.126, 16.144)
Anorectics and opioids
2.403
0.107
11.056
(8.960, 13.641)
Anorectics, BZD, and opioids
3.302
0.207
27.158
(18.093, 40.766)
104
B.L. Wilsey et al. / Drug and Alcohol Dependence 112 (2010) 99–106
stances. This was not entirely unanticipated, given the observation
that a recent National Survey on Drug Use and Health, measuring non-medical drug use, found that the misuse of prescription
opioids was second in prevalence only to the illicit use of marijuana (Office of Applied Studies, 2006). Somewhat unexpectedly,
hydrocodone, despite being the most commonly prescribed opioid
in California during 2007 (Table 1), had a lower association with
multiple prescriber episodes than any other opioid studied, with
the exception of codeine. A possible explanation for this finding
is that prescriptions for hydrocodone (as well as codeine) can be
refilled, unlike prescriptions for all other study opioids which are
contained in Schedule II. Having refillable prescriptions can reduce
the need for patients to seek multiple providers for the same medication. Alternatively, illicit users of prescription opioids who utilize
multiple providers may be more motivated to seek Schedule II opioids because they are perceived as more potent or desirable. In
this regard, two of the opioids with high associations with multiple provider episodes in the present study, hydromorphone and
sustained-release oxycodone, were favored by respondents when
asked to rate attractiveness for abuse of prescription opioids (Butler
et al., 2009).
The relatively low frequency in California of multiple providers
for opioids (12.8%) in 2007 was, in part, corroborated by PMP
data from another state, Massachusetts. During 1996–2006, most
patients in the Commonwealth of Massachusetts (87.5%) used 1–2
prescribers, 1–2 pharmacies, and had no early refills (Katz et al.,
2009). Taken together, the two studies indicate that doctor shopping for opioids is relatively infrequent from the perspective of
a prescriber searching for abusive behavior. These incidences are
not unlike the levels of aberrant drug-related behaviors in chronic
pain patients associated with opioid prescribing where the incidence of patients with aberrant behavior has been estimated to be
11.5% (Fishbain et al., 2008). This similarity is complemented by
the need for pharmacovigilance of both behaviors. Just as universal
precautions (i.e., using assessment tools to determine the risk for
addictive disorders, treatment agreements, and conducting ongoing monitoring of patients for the development of abuse related
problems) (Gourlay and Heit, 2006) has become the norm to reduce
prescription opioid abuse, PMPs have now emerged as a potentially
important tool for addressing prescription drug diversion (Katz et
al., 2007). Given the sheer magnitude of the problem (i.e., 12.8%
of 17.1 million equating to 2.2 million multiple provider prescriptions), PMP monitoring will be indispensable to reduce abuse and
diversion of prescription opioids.
Demographic characteristics associated with multiple provider
episodes were found to vary among the different classes of controlled substances (Tables 2–5). Younger age was associated with
the use of multiple providers for opioids and benzodiazepines.
Specifically, there was approximately a 1–2% decrease in the odds
of this behavior for every year of age. This is consistent with the
public health problem of a significant increase in the non-medical
use and abuse of prescription medications among adolescents and
young adults (Riggs, 2008). Contrariwise, older age was associated with an increased likelihood of using multiple providers to
obtain stimulants and anorectics. Specifically, there was an approximate 1% increase in the OR of this behavior for every year of age.
This might be an early manifestation of a demographic trend. As
the number of persons 65 years of age and older escalates with
the aging of the baby boomers, experts predict that prescription
drug abuse among the elderly, for certain medications, also will
rise significantly (Martin, 2008). Gender effects also varied with
the class of controlled substance. Males were more likely to use
multiple providers for benzodiazepines but less so for stimulants,
while gender showed no relationship between anorectics or opioids. The latter is somewhat contradictory to previous modeling
using PMP and managed care organization data in which males
were more likely to be involved in prescription opioid abuse (White
et al., 2009). Another discrepancy between our data and that of others involved the distinctly higher likelihood of multiple provider
episodes in large urban geographic areas compared to rural settings. Assuming that some of this activity represents illicit use,
this conflicts with reports suggesting that rural areas of Appalachia
experience more misuse than nearby urban centers (Havens et
al., 2007; Wunsch et al., 2009). One plausible explanation would
be that the relative lack of prescribers in rural California reduces
opportunities for finding multiple providers. Alternative reasons
might arise from cultural differences in Appalachian regions versus those in California and/or the sampling of probationers rather
than the general population (Havens et al., 2007; Wunsch et al.,
2007).
The duration of prescribing and the amount of medication
were generally statistically significant but clinically inconsequential determinants for using multiple providers. It is likely that these
variables had very small point estimates but achieved statistical
significance due to the large sample sizes. Patients neither appear
to seek multiple providers as time passes nor as they require higher
dosages. This finding may help reassure clinicians that these aberrant drug-related behaviors are not a likely consequence of chronic
therapy with controlled substances.
Of all the independent variables, simultaneously receiving prescriptions for different controlled substances and concurrent use of
multiple prescribers to obtain other controlled substances revealed
the greatest associations with multiple provider episodes. The
explanatory significance of acquiring numerous controlled substances from either the same or multiple providers suggests the
importance of polypharmacy in the occurrence of non-medical opioid use. Several years ago, an expert panel (Parente et al., 2004)
convened to consider factors that serve to identify patients with the
potential for controlled substance misuse or mismanagement, and
came to a similar conclusion. Controlled substance polypharmacy,
although not necessarily a direct measure of misuse, warrants
heightened vigilance.
Controlled substance polypharmacy is under scrutiny because
of its association with overdose deaths and other adverse events
(Wunsch et al., 2009). Recently, prescription drugs have replaced
heroin and cocaine as the leading drugs involved in unintentional
and undetermined fatal overdoses (Paulozzi and Xi, 2008). Between
1999 and 2002, the number of opioid analgesic poisonings on
death certificates increased 91.2%, while heroin and cocaine poisonings increased 12.4% and 22.8%, respectively (Paulozzi et al.,
2006). Although it is often unclear how people are obtaining the
substances involved in poisoning mortality, both pharmaceutical
diversion and multiple provider prescribing have been implicated
in this increased mortality from prescription opioids (Hall et al.,
2008).
As a means to address this public health issue, the Food and Drug
Administration (FDA) recently has begun to require pharmaceutical manufacturers to adopt broad risk evaluation and mitigation
strategies (REMS) for long-acting opioids to ensure that the benefits
of these drugs outweigh the risks (Food and Drug Administration,
2009). Unfortunately, an unduly restrictive REMS for a certain class
of opioids has the potential to deter clinicians from using these substances, even when they are medically warranted or, as has been
seen before, to even stimulate excessive use of other abusible drugs
not encompassed under the regulatory umbrella. Conversely, PMPs
have come to represent balanced methods to address some aspects
of prescription opioid abuse while ensuring access for patients who
need them and for which safe and effective use can be demonstrated. Although not investigated in the present study, curbing
multiple providers seems a reasonable objective for helping reduce
illicit use of controlled substances. As a result, it is logical that
maintaining and enhancing ongoing state PMP efforts, so that pre-
B.L. Wilsey et al. / Drug and Alcohol Dependence 112 (2010) 99–106
scribers have direct access to PMP data at the point of care, would
promote responsible prescribing and represent an integral part of
the solution to the societal abuse and diversion issues inspiring
the FDA’s conceptualization of a REMS approach (Food and Drug
Administration, 2009).
This study is characterized by a number of limitations that affect
the research questions that can be evaluated, as well as how the
results can be interpreted. For example, the CURES database was
conceptualized as a state drug control program designed to monitor provider prescribing to reduce the societal consequences of
the non-medical use of medications. It was never intended as a
research tool. Consequently, the data contain a limited amount of
information that can be used as explanatory factors for multivariate analyses. Although this study benefited from operationalizing
essential variable domains (e.g., patient and prescription characteristics, and the identification of multiple provider episodes), future
research investigating the phenomenon of multiple provider prescribing should attempt to elaborate the present model with other
relevant predictive characteristics.
Additionally, little empirical evidence exists to stratify multiple
provider prescribing in order to differentiate propensity for prescription drug abuse according to the extent that multiple providers
are used. An expert panel concluded that greater then five prescriptions for the same medication over a 1-year period was likely
indicative of misusing controlled substances (Parente et al., 2004)
but there has been no further attempt to determine the clinical
validity of this consensus conclusion. A related, but nevertheless
substantial, limitation is the inability to determine the association
between multiple provider prescribing and actual non-medical use
of these medications, as the CURES database contained no information documenting problems with diversion or abuse. At this time,
the explicit motivations for seeking multiple providers for controlled substances remain uncertain, as we are in the early stages of
this type of investigation and have yet to develop an understanding of this potentially illicit drug-seeking activity. Interviewing
subjects under anonymous and confidential circumstances may
enable further probing into this phenomenon. Although not feasible using a secondary data analysis containing de-identified PMP
data, similar studies have been performed involving the unlawful
channeling of regulated pharmaceuticals from legal sources to the
illicit marketplace by interviewing club drug users, street-based
illicit drug users, methadone maintenance patients, and HIV positive individuals who abuse and/or divert drugs (Inciardi et al.,
2007).
In addition to the limitations above, the definition of a multiple
provider episode utilized in this study was arbitrary and may have
produced spurious results in the case of a person who refilled the
prescription before the 30-day period using both a new provider
and a new pharmacy but who had a legitimate reason to do so.
Finally, there was also the issue of data entry into the CURES
database, which was limited to civilian practices thereby excluding Veteran Affairs, military, and prison facilities. In addition, the
prescriptions in the database had to be filled by pharmacists in California in order to be processed. We, therefore, could not detect
multiple provider episodes if they involved the crossing of state
(e.g., Oregon, Nevada, and Arizona) or international (e.g., Mexico or
via the internet) borders.
Although efforts are underway to utilize PMPs to improve health
care (Katz et al., 2008), there may be problems that undermine their
clinical effectiveness. For instance, a survey of providers indicated
some reluctance to utilize these programs (Barrett and Watson,
2005). Although a majority of the surveyed physicians believed
that these programs would be useful for monitoring prescription
histories and decreasing doctor shopping, very few (11%) actually requested information from the database (Barrett and Watson,
2005).
105
The incidence of multiple provider episodes was relatively low
for all controlled substances ranging from a high of 12.8% for opioids to a low of 0.9% for anorectics. Although only a small fraction
of prescriptions were involved, the consequences of the abuse and
diversion of controlled substances demand better ways of reducing
doctor shopping. Ideally, this must be accomplished without interfering with appropriate medical practice. There are indications that
PMPs offer a balanced approach. In a recent European study, the
diversion through doctor shopping for buprenorphine decreased
after a prescription monitoring program was instituted without
notable impact on the access to treatment (Pradel et al., 2009).
Although this finding awaits replication elsewhere, such results
portend the clinical and public health utility of PMPs, providing
an impetus for their widespread adoption.
Findings from the current study suggest that controlled substance polypharmacy requires special consideration in the clinical
setting. Notwithstanding its present shortcomings, PMP information is rapidly becoming an essential therapeutic tool to provide
appropriate medical care. Clinicians who provide prescription medications to patients with chronic pain may find that the availability
of PMP information at the point of care to identify multiple provider
episodes, as well as access to longitudinal records of PMP prescribing data, will help support more responsible prescribing of
controlled substances.
Role of funding source
Funding for this study was provided by the Robert Wood Johnson Foundation (RWJF). The RWJF had no further role in study
design; in the collection, analysis and interpretation of data; in the
writing of the report; or in the decision to submit the paper for
publication.
Contribution
Authors Wilsey, Fishman, and Gilson developed the study concept and design. The acquisition of data was conducted by authors
Casamalhuapa and Baxi. Analysis and interpretation of data was
conducted by authors Li, Zhang, Gilson, and Wilsey. Drafting of the
manuscript was performed by authors Wilsey, Fishman, and Gilson.
Critical revision of the manuscript for important intellectual content was performed by authors Wilsey, Fishman, Gilson, and Li. All
authors approved the final manuscript.
Conflict of interest
All authors declare that they have no conflicts of interest.
Acknowledgements
We gratefully acknowledge informatics and biostatistical support for this project from the UC Davis Clinical Translational
Science Center. This publication was made possible by Grant
Number UL1 RR024146 from the National Center for Research
Resources (NCRR), a component of the National Institutes of
Health (NIH), and NIH Roadmap for Medical Research. Its contents are solely the responsibility of the authors and do not
necessarily represent the official view of NCRR or NIH. Information on NCRR is available at http://www.ncrr.nih.gov/. Information
on Re-engineering the Clinical Research Enterprise can be obtained from http://nihroadmap.nih.gov/clinicalresearch/overviewtranslational.asp.
106
B.L. Wilsey et al. / Drug and Alcohol Dependence 112 (2010) 99–106
References
Barrett, K., Watson, A., 2005. Physician perspectives on a pilot prescription monitoring program. J. Pain Palliat. Care Pharmacother. 19, 5–13.
Bridges, S., 2007. In: Program, A.M.P. (Ed.), Memorandum: Clinical Edits and Revised
Dose Edits for ADD/ADHD Drugs. Arkansas Department of Health and Human
Services, Little Rock, AR, pp. 72203–81437.
Brushwood, D.B., 2003. Maximizing the value of electronic prescription monitoring
programs. J. Law Med. Ethics 31, 41–54.
Butler, S.F., Fernandez, K.C., Chang, A., Benoit, C., Morey, L.C., Black, R., Katz, N., 2009.
Measuring attractiveness for abuse of prescription opioids. Pain Med. 11, 67–80.
California Health and Safety Code Section 11165, 2010. http://law.onecle.com/
california/health/11165.html (accessed on 1.05.10).
California State Association of Counties, CA County Population, 2007. http://www.
csac.counties.org/default.asp?id=399 (accessed 17.03.08).
Food and Drug Administration, 2009. Opioid Drugs and Risk Evaluation
and Mitigation Strategies (REMS), FDA to Meet with Drug Companies
about REMS for Certain Opioid Drugs, http://www.fda.gov/Drugs/DrugSafety/
InformationbyDrugClass/ucm163647.htm (accessed on 19.06.09).
Fishbain, D.A., Cole, B., Lewis, J., Rosomoff, H.L., Rosomoff, R.S., 2008. What percentage of chronic nonmalignant pain patients exposed to chronic opioid analgesic
therapy develop abuse/addiction and/or aberrant drug-related behaviors? A
structured evidence-based review. Pain Med. 9, 444–459.
Fishman, S.M., Papazian, J.S., Gonzalez, S., Riches, P.S., Gilson, A., 2004. Regulating
opioid prescribing through prescription monitoring programs: balancing drug
diversion and treatment of pain. Pain Med. 5, 309–324.
Giannini, A., 1997. Drugs of Abuse, 2nd ed. Practice Management Information Corp.,
Los Angeles, pp. 121–125.
Gilson, A., 2010. In: Ballantyne, J., Rathmell, J., Fishman, S. (Eds.), Laws and Policies
Involving Pain Management, Bonica’s Management of Pain, 4th ed. Lippincott
Williams & Wilkins, New York, NY, pp. 166–183.
Gilson, A.M., Kreis, P.G., 2009. The burden of the nonmedical use of prescription
opioid analgesics. Pain Med. 10 (Suppl. 2), S89–S100.
Gourlay, D., Heit, H., 2006. Universal precautions: a matter of mutual trust and
responsibility. Pain Med. 7, 210–211 (author reply 212).
Hall, A.J., Logan, J.E., Toblin, R.L., Kaplan, J.A., Kraner, J.C., Bixler, D., Crosby, A.E.,
Paulozzi, L.J., 2008. Patterns of abuse among unintentional pharmaceutical overdose fatalities. JAMA 300, 2613–2620.
Hart, L.G., Larson, E.H., Lishner, D.M., 2005. Rural definitions for health policy and
research. Am. J. Public Health 95, 1149–1155.
Havens, J.R., Oser, C.B., Leukefeld, C.G., Webster, J.M., Martin, S.S., O’Connell, D.J.,
Surratt, H.L., Inciardi, J.A., 2007. Differences in prevalence of prescription opiate misuse among rural and urban probationers. Am. J. Drug Alcohol Abuse 33,
309–317.
Inciardi, J.A., Surratt, H.L., Kurtz, S.P., Cicero, T.J., 2007. Mechanisms of prescription
drug diversion among drug-involved club- and street-based populations. Pain
Med. 8, 171–183.
Joranson, D.E., Carrow, G.M., Ryan, K.M., Schaefer, L., Gilson, A.M., Good, P., Eadie, J.,
Peine, S., Dahl, J.L., 2002. Pain management and prescription monitoring. J. Pain
Symptom Manage. 23, 231–238.
Kasprak, J., 2003. State programs for electronic monitoring of controlled substance
prescriptions. In: Connecticut General Assembly OLR Research Report, Hartford,
CT.
Katz, N., Houle, B., Fernandez, K.C., Kreiner, P., Thomas, C.P., Kim, M., Carrow, G.M.,
Audet, A., Brushwood, D., 2008. Update on prescription monitoring in clinical
practice: a survey study of prescription monitoring program administrators.
Pain Med. 9, 587–594.
Katz, N., Panas, L., Kim, M., Audet, A.D., Bilansky, A., Eadie, J., Kreiner, P., Paillard, F.C.,
Thomas, C., Carrow, G., 2009. Usefulness of prescription monitoring programs for
surveillance-analysis of Schedule II opioid prescription data in Massachusetts,
1996–2006. Pharmacoepidemiol. Drug Saf. 19, 115–123.
Katz, N.P., Adams, E.H., Benneyan, J.C., Birnbaum, H.G., Budman, S.H., Buzzeo, R.W.,
Carr, D.B., Cicero, T.J., Gourlay, D., Inciardi, J.A., Joranson, D.E., Kesslick, J., Lande,
S.D., 2007. Foundations of opioid risk management. Clin. J. Pain 23, 103–118.
Korff, M.V., Saunders, K., Thomas Ray, G., Boudreau, D., Campbell, C., Merrill, J., Sullivan, M.D., Rutter, C.M., Silverberg, M.J., Banta-Green, C., Weisner, C., 2008.
De facto long-term opioid therapy for noncancer pain. Clin. J. Pain 24, 521–
527.
Mallinckrodt, C.H., Clark, W.S., David, S.R., 2001. Accounting for dropout bias using
mixed-effects models. J. Biopharm. Stat. 11, 9–21.
Martin, C.M., 2008. Prescription drug abuse in the elderly. Consult. Pharm. 23,
930–934, 936, 941-932.
Musto, D., Korsmeyer, P., 2002. The Quest for Drug Control: Politics and Federal
Policy in a Period of Increasing Substance Abuse, 1963–1981. Yale University
Press, New Haven, CT.
Newcombe, H.B., 1988. Handbook of Record Linkage: Methods for Health and Statistical Studies, Administration, and Business. Oxford University Press, Oxford,
UK.
Office of Applied Studies, 2006. Results from the 2005 National Survey on Drug
Use and Health: National Findings. NSDUH Series H-30, DHHS Publication No
SMA 06-4195. Substance Abuse and Mental Health Services Administration,
Rockville, MD.
Parente, S.T., Kim, S.S., Finch, M.D., Schloff, L.A., Rector, T.S., Seifeldin, R., Haddox, J.D.,
2004. Identifying controlled substance patterns of utilization requiring evaluation using administrative claims data. Am. J. Manag. Care 10, 783–790.
Paulozzi, L.J., Budnitz, D.S., Xi, Y., 2006. Increasing deaths from opioid analgesics in
the United States. Pharmacoepidemiol. Drug Saf. 15, 618–627.
Paulozzi, L.J., Xi, Y., 2008. Recent changes in drug poisoning mortality in the United
States by urban–rural status and by drug type. Pharmacoepidemiol. Drug Saf.
17, 997–1005.
Pradel, V., Frauger, E., Thirion, X., Ronfle, E., Lapierre, V., Masut, A., Coudert, C., Blin,
O., Micallef, J., 2009. Impact of a prescription monitoring program on doctorshopping for high dosage buprenorphine. Pharmacoepidemiol. Drug Saf. 18,
36–43.
Ricketts, T., Johnson-Webb, K., Taylor, P., June 1998. Definitions of Rural: A Handbook
for Health Policy Makers and Researchers. A Technical Issues Paper prepared
for the Federal Office of Rural Health Policy. Health Resources and Services
Administration, US DHHS.
Riggs, P., 2008. Non-medical use and abuse of commonly prescribed medications.
Curr. Med. Res. Opin. 24, 869–877.
Streiner, D.L., 2002. The case of the missing data: methods of dealing with dropouts
and other research vagaries. Can. J. Psychiatry 47, 68–75.
Taylor, J.M., 1986. Choosing the number of controls in a matched case–control study,
some sample size, power and efficiency considerations. Stat. Med. 5, 29–36.
Vieweg, W.V., Lipps, W.F., Fernandez, A., 2005. Opioids and methadone equivalents
for clinicians. Prim. Care Companion J. Clin. Psychiatry 7, 86–88.
WebMD, 2008. http://www.rxlist.com/cgi/generic/phenterm ids.htm (accessed on
19.06.09).
White, A.G., Birnbaum, H.G., Schiller, M., Tang, J., Katz, N.P., 2009. Analytic models
to identify patients at risk for prescription opioid abuse. Am. J. Manag. Care 15,
897–906.
Wilsey, B.L., Fishman, S.M., Crandall, M., Casamalhuapa, C., Bertakis, K.D., 2008. A
qualitative study of the barriers to chronic pain management in the ED. Am. J.
Emerg. Med. 26, 255–263.
Wright, C.T., Schnoll, S., Bernstein, D., 2008. Risk evaluation and mitigation strategies for drugs with abuse liability: public interest, special interest, conflicts of
interest, and the industry perspective. Ann. N.Y. Acad. Sci. 1141, 284–303.
Wunsch, M.J., Nakamoto, K., Behonick, G., Massello, W., 2009. Opioid deaths in rural
Virginia: a description of the high prevalence of accidental fatalities involving
prescribed medications. Am. J. Addict. 18, 5–14.
Wunsch, M.J., Nakamoto, K., Goswami, A., Schnoll, S.H., 2007. Prescription drug abuse
among prisoners in rural Southwestern Virginia. J. Addict. Dis. 26, 15–22.