Represents a cumulative percentage in each group over time after program
entry. Major opioid users with dual diagnosis had an increased risk of relapse
compared with nonopioid users with or without a dual diagnosis (P<.001). Without dual diagnosis, major opioid users and nonopioid
users did not have a significantly different relapse risk.
Domino KB, Hornbein TF, Polissar NL, Renner G, Johnson J, Alberti S, Hankes L. Risk Factors for Relapse in Health Care Professionals With Substance Use Disorders. JAMA. 2005;293(12):1453-1460. doi:10.1001/jama.293.12.1453
Author Affiliations: Departments of Anesthesiology
(Drs Domino and Hornbein) and Psychiatry (Dr Hankes), University of Washington,
Seattle; Mountain-Whisper-Light Statistical Consulting, Seattle (Dr Polissar);
and Washington Physicians Health Program, Seattle (Dr Hankes and Mss Renner
and Johnson and Mr Alberti).
Context Substance use disorders among physicians are important and persistent
problems. Considerable debate exists over whether use of major opioids, especially
among anesthesiologists, is associated with a higher relapse rate compared
with alcohol and nonopioids. Moreover, the risk factors for relapse with current
treatment and monitoring strategies are unknown.
Objective To test the hypothesis that chemically dependent health care professionals
using a major opioid (eg, fentanyl, sufentanil, morphine, meperidine) as drug
of choice are at higher risk of relapse.
Design, Setting, and Participants Retrospective cohort study of 292 health care professionals enrolled
in the Washington Physicians Health Program, an independent posttreatment
monitoring program, followed up between January 1, 1991, and December 31,
Main Outcome Measure Factors associated with relapse, defined as the resumption of substance
use after initial diagnosis and completion of primary treatment for chemical
Results Twenty-five percent (74 of 292 individuals) had at least 1 relapse.
A family history of a substance use disorder increased the risk of relapse
(hazard ratio [HR], 2.29; 95% confidence interval [CI], 1.44-3.64). The use
of a major opioid increased the risk of relapse significantly in the presence
of a coexisting psychiatric disorder (HR, 5.79; 95% CI, 2.89-11.42) but not
in the absence of a coexisting psychiatric disorder (HR, 0.85; 95% CI, 0.33-2.17).
The presence of all 3 factors—major opioid use, dual diagnosis, and
family history—markedly increased the risk of relapse (HR, 13.25; 95%
CI, 5.22-33.59). The risk of subsequent relapses increased after the first
relapse (HR, 1.69; 95% CI, 1.13-2.53).
Conclusions The risk of relapse with substance use was increased in health care
professionals who used a major opioid or had a coexisting psychiatric illness
or a family history of a substance use disorder. The presence of more than
1 of these risk factors and previous relapse further increased the likelihood
of relapse. These observations should be considered in monitoring the recovery
of health care professionals.
The prevalence of chemical dependency (excluding nicotine) among physicians
has been estimated to be 10% to 15%,1 similar
to that in the general population.1- 3 Following
completion of primary treatment, recovery is best achieved through continuing
group therapy and regular attendance at mutual help groups.4 Because
of the proclivity to relapse, ongoing monitoring can help ensure sustained
remission of individuals occupying safety-sensitive positions. Monitoring
methods have changed over the past decade and now include frequent contact
for behavioral assessment, random urine testing with observed micturition,
and workplace surveillance.5,6 Treatment
programs estimate that up to 70% of health care professionals successfully
return to medical practice.7- 12
Data on the incidence of relapse and risk factors contributing to the
likelihood of relapse after initial treatment for substance use are lacking.
Virtually every study of chemical dependency among health care professionals
has had relatively short follow-ups, limitations in statistical methods or
analyses, and variable intensity of monitoring.
Among health care professionals, anesthesiologists appear to be at somewhat
higher risk. They are overrepresented in drug treatment programs relative
to their proportion among medical specialties.10,13,14 With
their hands-on access to many potent addicting drugs,15 they
have a predilection for parenteral opioids16 and
nearly 3 times the risk of drug-related death than general internal medicine
We examined 11 years of outcome data from the Washington Physicians
Health Program (WPHP), a posttreatment program monitoring health care professionals
with substance use disorders. We sought in particular to identify factors
that might predispose individuals to relapse. In addition, we examined whether
those who self-administered potent opioids might be more at risk of relapse
(and perhaps death) than users of other drugs.
The study was approved by the University of Washington Human Subjects
Review Committee. Outcome data from the WPHP were analyzed using a retrospective
cohort design. The WPHP monitors physicians (MD, DO), veterinarians (DVM),
dentists (DDS, DMD), podiatrists (DPM), registered pharmacists (RPh), and
physician assistants (PA). The cohort we selected consisted of those entering
the program for monitoring of a substance use disorder between January 1,
1991, and December 1, 2001, and followed up through December 31, 2001. Excluded
from the study were those who entered the program after relapse, those for
whom date of enrollment was missing from the database, and those for whom
outcome was not known.
Information from the WPHP database included the following: date of program
enrollment, age at entry, sex, type of medical professional training status,
family history of substance use disorder, current smoking status, diagnosis
of a coexisting psychiatric disorder (dual diagnosis), drug of choice, and
route of drug administration. Multiple-drug use was classified by the treatment
program according to the predominant substance used. All major opioids were
considered predominant over alcohol and other classes of drugs. Drug of choice
was initially categorized into 6 groups: fentanyl (includes fentanyl citrate
and sufentanil citrate), other major opioids (morphine, meperidine hydrochloride,
methadone hydrochloride, heroin, controlled-release oxycodone hydrochloride),
minor opioids (butorphanol, codeine, hydrocodone, nalbuphine hydrochloride,
oxycodone, pentazocine, propoxyphene, and tramadol hydrochloride), alcohol,
cocaine, and others (including benzodiazepines).
Family history of substance use disorder and presence of current smoking
were ascertained from 1 or all of 3 sources: (1) at initial contact by WPHP
staff’s structured interview, (2) by reviewing the discharge summary
from the treatment center, and (3) by reviewing the standard intake form completed
by the individual at the beginning of the monitoring program.
The diagnoses of psychiatric disorders were made by a board-certified
psychiatrist doing clinical evaluation while individuals were in inpatient
treatment. Dual diagnosis was ascertained by the WPHP staff reviewing the
individual’s discharge summary from the treatment center. The diagnoses
for those individuals who had a coexisting psychiatric disorder were taken
from their Diagnostic and Statistical Manual of Mental Disorders,
Fourth Edition (DSM-IV) discharge diagnoses.
Using the DSM-IV discharge diagnoses following several
months of inpatient treatment and multiple negative urine toxicology results
diminishes the likelihood of substance-induced psychiatric disorders.
Specialties were categorized as medical (family practice, internal medicine
and medical specialties, pediatrics, psychiatry, radiology, other), surgical
(emergency medicine, obstetrics and gynecology, surgery and surgical specialties),
anesthesiology, and nonphysicians (veterinary medicine, dental surgery and
dental medicine, physician assistants, podiatric medicine, pharmacy).
Relapse was defined as the resumption of substance use after initial
diagnosis and completion of primary treatment for chemical dependency. The
method of detection of the relapse (self-report, behavioral monitoring, chemical
monitoring, workplace monitoring, regulatory board reports, or other) was
noted. The date of relapse, date of death, return to medical practice, and
specialty change were also recorded.
Due to differences in time of follow-up and covariates that may affect
outcome, the effects of factors potentially influencing the risk of relapse
and the time to first relapse were analyzed using methods of survival analysis,
including the Cox proportional hazards regression model and the log-rank test.
These methods provide unbiased estimates of the cumulative relapse rate in
the population from which this cohort was drawn for any specified duration
after enrollment and an unbiased way to compare the risk of relapse among
the various drugs of choice and between categories of other variables.
To determine whether the risk of relapse varies over time after enrollment,
the relapse rate per 1000 person-years in each period in the program was calculated
(<2, 2 to <5, and ≥5 years); these groupings are based on monitoring
stages in the program. Using the χ2 test, the observed number
relapsing in each period was compared with the expected number relapsing based
on the assumption that the risk of relapse does not change after enrollment.
Differences in the distribution of drug of choice and other characteristics
were compared in early (1991-1996) and late (1997-2001) periods by the χ2 test.
To ascertain variables potentially confounding the effect of drug choice
on relapse and therefore to be included in a multivariate analysis, cross-tabulation
with χ2 tests, t tests, or the Mann-Whitney
test were used to detect covariates associated with the drug of choice. Furthermore,
the association of each covariate with risk of relapse was also determined.
From these initial analyses, appropriate variables were included in a multivariate
analysis. Because drug of choice was highly associated with route of administration
and specialty, only drug of choice was analyzed in the multivariate model.
Estimated 5-year relapse rates (based on the Kaplan-Meier method) with 95%
confidence intervals (CIs), based on the log-transformation method,18 were calculated.
The Cox proportional hazards model (as used in survival analysis) was
used to compare the risk of a second or later relapse with the risk of a first
relapse. For this analysis, we defined the following treatment periods: stage
0 consisted of all individuals; stage 1 were those with 1 relapse, stage 2
were those with 2 relapses, and stage 3 were those with 3 or more relapses.
Calculations of standard errors and statistical significance in the Cox models
were adjusted for the statistical dependence of multiple observations (stages)
per patient using the robust sandwich estimator of the variance, which adjusts
the estimates of variance to take into account the correlation of observations
within a person.19 Analyses were performed
using R version 2.0.1 (R Foundation for Statistical Computing, Vienna, Austria); P<.05 was used to determine statistical significance.
Of 300 individuals who met inclusion criteria, 8 were excluded for enrollment
in the program after relapse (n = 2), missing date of enrollment
(n = 5), and uncertain outcome status (n = 1). Of the
292 individuals remaining, 84% were men, 72% were 40 years or older, and 72%
had a family history of substance use disorder (86% involving first-degree
relatives; Table 1). Thirty-seven percent
had a dual diagnosis and of those, 93% had a DSM-IV Axis
I diagnosis only (Table 1). The drug
of choice was alcohol in more than half of the individuals (n = 164);
fentanyl (n = 27) and other major opioids (n = 15) represented
the drug of choice in 14% (Table 1).
There were no cases involving use of heroin, methadone, or time-released oxycodone.
Most fentanyl users were anesthesiologists (22 of 27). There were no significant
differences in most individual characteristics and in the drug of choice for
the 161 individuals who entered from 1991 through 1996 compared with the 131
individuals entered from 1997 through 2001. However, an increased proportion
of individuals in the later period presented with a dual diagnosis (50% in
1997-2001 vs 26% in 1991-1996, P<.001).
Seventy-four (25%) of 292 individuals had at least 1 relapse. Fourteen
(5%) had exactly 2 relapses and 10 (3%) had 3 or more relapses. The drug of
relapse was the initial drug of choice in 85% (n = 63) of the cases
(Table 2). Most relapses were detected
by chemical or workplace monitoring and 58% occurred within the first 2 years
in the program (Table 2). The risk of
relapse decreased (P<.001) with increasing duration
in the program, from 91 per 1000 person-years at 0 to 2 years in the program,
to 58 per 1000 person-years at 2 to 5 years, and to 32 per 1000 person-years
after 5 years. Ten individuals (13%) had first relapses after 5 years. Of
the 51 individuals who had a relapse and were followed up for 5 years or more,
61% (n = 31) successfully returned to the practice of medicine.
In contrast, all individuals followed up for 5 years or more (n = 110)
without a relapse successfully returned to the practice of medicine (P<.001). Only 5 anesthesiologists of 22 with fentanyl
as drug of choice returned to anesthesia practice without known relapse. Four
of these anesthesiologists had a family history, 1 had dual diagnosis, and
none had both of these characteristics.
Hazard ratios (HRs) for relapse for the various classes of drug of choice
are shown in Table 3. Fentanyl had a
slightly but nonsignificantly lower risk of relapse than other major opioids
(morphine, meperidine), and was combined with other major opioids for analysis.
Pairwise comparisons demonstrated that there were no significant differences
in relapse risk between cocaine, alcohol, minor opioids, and other drugs.
Therefore, for subsequent analyses, the drug of choice was categorized as
major opioid vs all other drugs. Compared with all other drugs, the risk of
relapse was almost doubled when a major opioid was the drug of choice (HR,
1.80; 95% CI, 1.03-3.13; P = .04; Table 4). The estimated cumulative relapse at
5 years was 35% for major opioid use compared with 25% for all other drugs
(Table 4). The risk of relapse was also
approximately doubled by having a dual diagnosis (HR, 2.12; 95% CI, 1.33-3.36; P = .002) or family history of substance use
disorder (HR, 2.14; 95% CI, 1.18-3.90; P = .01; Table 4). Specialty affected relapse risk, with
an increased risk in nonphysicians compared with physicians in medical specialties
(Table 4). Age, sex, current smoking,
and training status did not influence the risk of relapse (Table 4).
Those anesthesiologists who returned to the practice of anesthesiology
had an increased risk of relapse compared with those who did not return to
anesthesiology practice (HR, 8.54; 95% CI, 1.08-67.56; P = .04; Table 4).
Comparing anesthesiologists who returned to anesthesiology vs those who did
not, there were no substantial differences in drug of choice or any of the
characteristics, except for a greater proportion of residents in those who
did not return to anesthesiology (P = .07).
Multivariate analyses could not meaningfully be performed on specialty change
from anesthesiology due to the small sample size.
Family history (HR, 2.29; 95% CI, 1.44-3.64), dual diagnosis (HR, 2.25;
95% CI, 1.23-4.11), and use of a major opioid (HR, 1.78; 95% CI, 1.02-3.09)
were statistically significant predictors of relapse, even when controlling
for the other 2 factors in a model with just these 3 factors. However, there
was an important interaction in the risk of relapse with use of a major opioid
and dual diagnosis (Table 5; Figure). Major opioid users with a coexisting
psychiatric illness had a significantly increased risk of relapse compared
with nonopioid users with (HR, 3.36; 95% CI, 1.64-6.87; data not shown) or
without (HR, 5.79; 95% CI, 2.89-11.42) a dual diagnosis (Table 5). In contrast, major opioid users without a coexisting psychiatric
illness had a relapse risk similar to that of nonopioid users (HR, 0.85; 95%
CI, 0.33-2.17; Table 5; Figure). A major opioid user with both a coexisting psychiatric
illness and family history of substance use had a markedly increased risk
of relapse (HR, 13.25; 95% CI, 5.22-33.59; Table
Risk factors for relapse in the subset of individuals with alcohol as
the drug of choice (Table 6) show similar
HRs and estimated cumulative relapse rates as the entire cohort. Family history
(HR, 2.31; 95% CI, 1.01-5.26; P=.05) and dual diagnosis
(HR, 2.41; 95% CI, 1.26-4.61; P = .008)
were significant predictors of relapse in the multivariate model (Table 6).
The likelihood of relapse increased with each relapse. The estimated
5-year rate of first relapse for all individuals in stage 0 was 26% (95% CI,
20%-31%). For those who relapsed once (stage 1), the probability of 2 or more
relapses increased to 43% (HR, 1.69 vs stage 0; 95% CI, 1.13-2.53; P = .02). We could discern no significant differences in
characteristics for individuals who had multiple relapses compared with those
who had only 1 relapse.
Three individuals died of unintentional overdose in relapse; 2 deaths
occurred while still in the monitoring program and 1 death after completion
of the program. The drug of choice was meperidine in all 3 individuals; 2
had a family history and 1 had dual diagnosis. None was an anesthesiologist
and none was a resident. None of the 27 individuals whose drug of choice was
fentanyl died. Two additional deaths in relapse were reported to the WPHP
after the end of the study follow-up period; in both cases the drug of choice
The risk of relapse with substance use was markedly increased in health
care professionals who used a major opioid, had a coexisting psychiatric illness,
or had a family history of a substance use disorder. The presence of more
than 1 of these risk factors and previous relapse further increased the likelihood
of relapse. Major opioid users without a coexisting psychiatric illness did
not have a significantly elevated risk of relapse.
The study used a retrospective cohort design, and, as is true for any
observational study, risk factors represent associations not causation. Several
of the covariates were highly correlated with each other, such as major opioid
with parenteral route of administration and with specialty as an anesthesiologist.
These associations prevented us from isolating route of administration and
medical profession or specialty independent from the drug of choice.
Although drug of choice and most individual characteristics did not
differ over the 11-year study period, having a coexisting psychiatric disorder,
a factor strongly associated with risk of relapse, was increased in the later
compared with the earlier period. Implementation of more comprehensive psychological
assessment may have contributed to this increase. During the study period,
treatment centers became more sophisticated in diagnosing and treating other
mental health conditions due to a heightened awareness among chemical dependency
counselors, more involvement of psychiatrists, and use of psychometric testing
as a standard procedure. In addition, medical professional clients spent more
time in treatment, and this longer period of observation revealed mental health
symptoms that were not necessarily attributable to chemical toxicity or withdrawal.
Although rehabilitation with return to successful medical practice may
be ultimately achieved despite recurrent substance use, in our study population,
all health care professionals who did not successfully return to medical practice
or who died were in the relapse group. Thus even 1 relapse has poor prognostic
significance. Because the number of individuals with more than 1 relapse was
small, we chose the first relapse as the major outcome variable. The incidence
of relapse after the 5-year monitoring program had ended may have been underestimated
because of loss of follow-up by the program. Although most relapses were observed
in the first 2 years, 13% of relapses occurred after the 5-year monitoring
period, with some identified even 9 years after initial diagnosis.
Because data were derived from a single monitoring program, one must
extrapolate to other programs with caution. The number of individuals was
small, particularly within subgroups, causing us to combine drugs of disparate
classes and addictive potential (eg, cocaine with alcohol).
Major opioid use increased the risk of relapse significantly in the
presence of a coexisting psychiatric disorder but not in the absence of a
coexisting psychiatric disorder (Table 5).
This finding, while supported by early literature,14,20 is
in contrast to some recent studies.21,22 The
absence of a detected increased opioid risk in earlier studies may result
from small sample sizes,10,22 a
shorter follow-up period,21 or variation in
important covariates such as coexisting psychiatric diagnosis and family history.10,21,22
The importance of a coexisting psychiatric disorder in increasing the
risk of relapse is consistent with other reports in the literature.10,21 Analysis of 101 physicians treated
at the Rush Behavioral Health program between 1984 and 1991 found higher proportions
of coexisting psychiatric diagnoses, especially personality disorder, in relapsing
compared with nonrelapsing physicians.21 A
variety of other psychological factors described as contributing to relapse,
such as persistent denial, failure to accept the disease, dishonesty, stress,
overconfidence, and withdrawal, could reflect the presence of coexisting psychiatric
Anesthesiologists represent a larger proportion of physicians in substance
abuse programs than predicted by their relative numbers,13,14,21 perhaps
reflecting their ease of access to highly addictive major opioids13,15 or the heightened awareness of the
behavioral indicators of chemical dependency by anesthesiology departments.23 In addition, Alexander et al17 found
that anesthesiologists had nearly 3 times the risk of drug-related death than
general internal medicine physicians. Due to the confounding of use of major
opioids and specialty, we did not have sufficient power to investigate the
independent effects of the specialty of anesthesiology with risk of relapse.
However, recent studies from the Medical Society of New Jersey’s Physician
Health Program22 and the California Physicians
Diversion Program9 found no higher risk for
anesthesiologists compared with other specialists. These findings differ from
those reported by Menk et al20 more than a
decade ago. That retrospective survey of anesthesiology residency program
directors described a poor success rate of reentry into anesthesiology residency,
with a high death rate for residents using parenteral opioids. Treatment protocols
and recommendations for return to specialty have evolved in the last decade,
with increased emphasis on aggressive follow-up and monitoring5,6 and
an increasing tendency to encourage opioid-addicted anesthesia personnel to
change to a different specialty. The report by Menk et al20 poses
the possibility that trainees may be at higher risk for relapse although our
study did not observe a statistically significant influence of age or training
status on relapse rates.
We did not find the incidence of relapse to be greater for persons using
fentanyl or sufentanil than for those using other major opioids such as morphine
and meperidine. Although the numbers are small, the risk of relapse appeared
to be increased for those anesthesiologists who returned to the practice of
anesthesiology compared with those who did not. Five of the 22 anesthesiologists
using fentanyl or sufentanil did return to the practice of their specialty
with no evidence of relapse. All 5 possessed 1 additional risk factor; 4,
a family history; and 1, a dual diagnosis. Multivariate analysis regarding
how these additional factors influenced relapse was impractical because of
the small sample size.
Should anesthesiologists using major opioids return to the practice
of anesthesiology? From experience with only 22 individuals, we are not comfortable
making a definitive recommendation, yet certain of our observations may shed
helpful light on this question. First, because the risk of relapse for major
opioid users without other risk factors is no higher than that for users of
other drugs with no other risk factors (Table
5), perhaps anesthesiologists who have used fentanyl or other major
opioids but who have no other risk factors and no history of relapse might
be reasonable candidates for return to their specialty. Second, a coexisting
psychiatric disorder and family history of substance use increase the likelihood
of relapse, as does each relapse; the combination of more than 1 of these
conditions appears to further compound the risk. Thus additional risk factors
and relapse make return to anesthesia practice more problematic. Whatever
the decision on this question, more intensive and more prolonged monitoring
and treatment might enhance the odds for successful recovery. To better explore
the question of the advisability of returning to the practice of anesthesiology,
aggregating the experience from other physician health programs would be highly
In health care professionals with a substance use disorder, the presence
of a coexisting psychiatric illness or a family history of substance use disorder
significantly increased the likelihood of relapse, as did the presence of
prior relapse. Use of major opioids also increased risk of relapse in the
presence of family history and even more dramatically in those with a dual
diagnosis, and the combination of all 3 risk factors further magnified the
likelihood of relapse. State physician health programs might wish to consider
managing substance-using professionals who have 1 or more of these 3 risk
factors and those with prior relapse with more intensive and more prolonged
Corresponding Author: Karen B. Domino, MD,
MPH, Department of Anesthesiology, University of Washington, Box 356540, Seattle,
WA 98195-6540 (firstname.lastname@example.org).
Author Contributions: Dr Domino had full access
to all of the data in the study and takes responsibility for the integrity
of the data and the accuracy of the data analysis.
Study concept and design: Domino, Hornbein,
Renner, Johnson, Alberti, Hankes.
Acquisition of data: Renner, Johnson, Alberti,
Analysis and interpretation of data: Domino,
Hornbein, Polissar, Hankes.
Drafting of the manuscript: Domino, Hornbein,
Critical revision of the manuscript for important
intellectual content: Domino, Hornbein, Polissar, Renner, Johnson,
Statistical analysis: Domino, Polissar.
Administrative, technical, or material support:
Domino, Polissar, Renner, Hankes.
Study supervision: Domino, Polissar, Hankes.
Financial Disclosures: None reported.
Acknowledgment: We would like to acknowledge
Blazej Neradilek, MS, for data analysis and Lynn Akerlund for manuscript preparation.