Association of common mental disorders in 1998 with regular prescription opioid use in 2001: unadjusted odds ratios with 95% confidence intervals.
Sullivan MD, Edlund MJ, Zhang L, Unützer J, Wells KB. Association Between Mental Health Disorders, Problem Drug Use, and Regular Prescription Opioid Use. Arch Intern Med. 2006;166(19):2087-2093. doi:10.1001/archinte.166.19.2087
Copyright 2006 American Medical Association. All Rights Reserved. Applicable FARS/DFARS Restrictions Apply to Government Use.2006
Use of opioids for chronic noncancer pain is increasing, but standards of care for this practice are poorly defined. Psychiatric disorders are associated with increased physical symptoms such as pain and may be associated with opioid use, but no prospective population-based studies have addressed this issue.
Analysis of longitudinal data from 6439 participants in the 1998 and 2001 waves of Healthcare for Communities, a nationally representative telephone community survey.
Two hundred thirty-seven subjects (3.6%) reported regular prescription opioid use in 2001. In unadjusted logistic regression models, respondents with a common mental health disorder in 1998 (1165 [12.6%]; major depression, dysthymia, generalized anxiety disorder, or panic disorder) were more likely to report opioid use in 2001 than those without any of these disorders (odds ratio [OR], 4.43; 95% confidence interval [CI], 3.64-5.38; P<.001). Risk was increased for initiation (OR, 3.26; 95% CI, 2.44-4.34; P<.001) and continuation (OR, 2.30; 95% CI, 1.02-5.17; P = .04) of opioids. Respondents reporting problem drug use (136 [2.0%]; OR, 3.57; 95% CI, 2.32-5.50; P<.001) but not problem alcohol use (401 [6.5%]; OR, 0.73; 95% CI, 0.43-1.24; P = .25) reported higher rates of prescribed opioid use than those without problem use. In multivariate logistic regression models controlling for 1998 demographic and clinical variables, common mental health disorder (OR, 1.96; 95% CI, 1.47-2.62; P<.001) and problem drug use (OR, 2.98; 95% CI, 1.68-5.30; P<.001) remained significant predictors of opioid use in 2001.
Common mental health disorders and problem drug use are associated with initiation and use of prescribed opioids in the general population. Attention to psychiatric disorders is important when considering opioid therapy.
In primary care settings around the world, 5% to 35% of patients have pain not associated with cancer that persists for at least 6 months and is associated with disability or seeking health care.1 Between 1980 and 2000, the rate of prescribing opioids at US outpatient visits for chronic musculoskeletal pain doubled from 8% to 16% of visits.2 This increase has been interpreted by some specialists as evidence of better treatment of unrelieved pain,3 whereas others have expressed concern about the safety and effectiveness of long-term opioid therapy for chronic noncancer pain.4
Patients with mental disorders generally have been excluded from randomized trials to test the effectiveness of opioids for chronic noncancer pain, but results from observational studies suggest an association between mental disorders and increased opioid use for chronic noncancer pain.5- 8 It has not been possible to determine from these small or cross-sectional studies whether this association is found outside of specialty pain clinic settings. It also has not been possible to determine whether mental disorders increase the risk of opioid use or whether opioid use increases the risk of mental disorders. The present study investigates this association by using data from the 1998 and 2001 waves of a large, nationally representative, population-based survey, Healthcare for Communities (HCC). We hypothesized that, controlling for pain-related interference and other covariates, individuals with common mental health and substance abuse disorders in 1998 would be more likely to report regular use (including initiation and continuation) of prescribed opioids in 2001 than those without these disorders.
Data are from HCC, a part of the Robert Wood Johnson Foundation's Health Tracking Initiative. The HCC was a nationwide telephone survey designed to track the effects of the changing health care system on individuals at risk for alcohol abuse, drug abuse, and mental health disorders. Study methods have been described previously.9 The HCC respondents were a stratified probability sample of participants in the Community Tracking Study, a nationally representative survey of the US civilian population from 1996 through 1997.10 The Community Tracking Study includes a sample clustered within 60 randomly selected US communities and a national sample. The first wave (HCC1) was conducted in 1997 and 1998, and the second wave (HCC2) was conducted in 2000 and 2001. From the 14 985 respondents selected for HCC1, 9585 complete interviews were obtained, for a response rate of 64.0%. In HCC2, all HCC1 respondents were followed up with a response rate of 69.5%, yielding 6659 completed interviews. We used sample weights designed by HCC statisticians to adjust for the probability of selection, nonresponse to the 2001 HCC2 survey, and the number of households in the Community Tracking Study survey that did not have a telephone. We excluded respondents reporting cancer diagnoses other than skin cancer within the past 3 years (n = 132) to isolate noncancer pain. Those missing data for key variables (n = 88; most often presence of arthritis [n = 48] or health insurance plan [n = 33]) were also excluded, resulting in a sample of 6439. Informed consent was obtained orally before the interview, and the study was approved by the institutional review boards at University of California at Los Angeles and RAND.
The HCC respondents were asked about all prescribed medications that they had taken “at least several times a week for a month or more” in the past 12 months. Respondents were asked to read the names of all their medications directly from their pill bottles to the interviewers. These medications were later grouped into therapeutic categories by study staff. From these data, we constructed a dichotomous measure of regular prescription opioid use of at least 1 month. A total of 350 individuals reported such regular opioid use at 1 or more time points. We sorted respondents into 4 groups on the basis of their HCC1 and HCC2 regular prescription opioid use: no use (n = 6089), 1998 use only (n = 113), 2001 use only (n = 159), or both 1998 and 2001 use (n = 78). Of the 237 subjects reporting opioid use in 2001, 183 (77.2%) confirmed use for at least 2 months. The strength of the association between common mental disorders and opioid use was similar for this group confirming more long-term opioid use but was somewhat larger. Although we do not have data on the reasons these opioids were prescribed, pain is the primary indication for opioids, and only 10 respondents (4.2%) using opioids in 2001 did not report a chronic pain condition or pain interference at that time.
Our model is based on the work of Andersen11 who posited that access to medical treatment is a function of patient need, predisposing factors, and enabling factors. We used several complementary measures of physical health and chronic physical conditions as measures of the need for opioid treatment. These measures included the Physical Component Summary–12, which is an aggregate measure of physical health functioning, and a measure of self-reported health (excellent, very good, good, fair, or poor), both from the Medical Outcomes Study Short Form–12.12 The HCC contains information about 17 self-reported chronic medical conditions. Our final models included the number of chronic conditions and those individual conditions associated with opioid use (P<.10): arthritis, chronic low back pain, chronic headaches, and other chronic pain conditions. We also included a dichotomized measure of pain interference with respondents' normal work (including work outside the home and housework) derived from the Medical Outcomes Study Short Form–12 pain item (extremely/a lot vs moderately/a little bit/not at all). All independent variables were from the 1998 wave.
The presence during the preceding 12 months of the mental disorders most commonly diagnosed and treated in primary care (major depression, dysthymia, generalized anxiety disorder, and panic disorder) was assessed by using short-form versions13 of the World Health Organization's Composite International Diagnostic Interview (CIDI),14 which is based on the Diagnostic and Statistical Manual of Mental Disorders, Revised Third Edition.15 The CIDI is a fully structured interview designed to be administered by lay interviewers.
Problem alcohol and drug use were assessed with common assessment tools. These scales do not exactly replicate Diagnostic and Statistical Manual of Mental Disorders, Revised Third Edition diagnoses of substance abuse or dependence (eg, they lack a clear determination of loss of control over use), so we use the terms problem alcohol use and problem drug use. Alcohol problems were assessed by using the Alcohol Use Disorders Identification Test, with a score of 8 or more representing problem alcohol use.16 Problem drug use was designated on the basis of answering yes to either of 2 items adapted from the CIDI—indications of tolerance to the medication and/or emotional or psychological problems from drug use (such as depression, anhedonia, paranoia, delusions, or hallucinations). Problem use of prescription drugs was defined for the respondent as use “on your own, meaning without a doctor's prescription, or in larger amounts than prescribed, or for a longer period than prescribed.” These questions thus captured a broader group than do classic substance abuse criteria. Significant problems have been noted with the sensitivity and specificity of the Diagnostic and Statistical Manual of Mental Disorders, Revised Third Edition criteria for opioid abuse and dependence when applied to patients using prescribed opioids long term.17,18 The abuse questions were asked in a section of the survey separate from questions about prescription medication use. Other covariates included sex, income, marital status, insurance status, work disability, race, age, education, and region of the country.
In separate logistic regressions, we regressed opioid use in 2001 on the presence of any 1 of the 4 common mental disorders in 1998 (major depression, dysthymia, generalized anxiety disorder, or panic disorder) and each disorder separately. In separate logistic regressions, we focused on initiation of opioids between 1998 and 2001 (by excluding those using opioids in 1998) and on the continuation of the use of opioids (by excluding those not using opioids in 1998).
Using multiple logistic regressions, we then investigated the effects of having any of the 4 common mental disorders and problem substance use in 1998 on regular opioid use in 2001, first adjusting for sociodemographic factors, and then sociodemographic and clinical factors. There were no significant interactions between common mental disorders and any of the other independent variables. We repeated these analyses in subsets focusing on opioid initiation (those not using opioids in 1998) and continuation (those using opioids in 1998). All analyses were conducted by using SUDAAN 9.019 with sample weights and taking into account the clustered survey design for estimates of standard errors. Results are reported in raw numbers of respondents, but percentages are weighted to be interpretable with respect to the general population.
Regular prescription opioid use (at least several times a week for a month or more) during at least 1 assessment period was reported by 350 subjects (5.4%). Demographic and clinical characteristics of the HCC total sample and of the opioid use groups are shown in Table 1. A total of 6202 (96.3%) respondents reported no regular opioid use in 2001, whereas 237 (3.6%) reported regular opioid use; 1165 (12.6%) met criteria for at least 1 of 4 common mental disorders (major depression, dysthymia, generalized anxiety disorder, or panic disorder) in 1998. In unadjusted logistic regression models, respondents with any of these disorders in 1998 had higher rates of prescription opioid use in 2001 than those without (odds ratio [OR], 4.43; 95% confidence interval [CI], 3.64-5.38; P<.001) (Figure). Each individual 1998 disorder was also significantly associated with opioid use (P<.001) and the ORs were large (2.56-6.51). The OR for generalized anxiety disorder was significantly lower than those for panic (P = .02) or dysthymia (P = .02). Individuals with problem drug use diagnosed in 1998 (n = 136) also had higher rates of prescribed opioid use in 2001 (OR, 3.57; 95% CI, 2.32-5.50; P<.001) than those without problem drug use (n = 6303). Drugs most frequently used by those noting problem drug use were marijuana (n = 74), analgesics (n = 61), and sedatives (n = 42). Individuals with problem alcohol use in 1998 (n = 401) were not more likely to use prescription opioids (OR, 0.73; 95% CI, 0.43-1.24; P = .25) than those without (n = 6038).
In separate logistic regressions, mental disorder in 1998 was associated with increased risk for initiation of regular opioid use between 1998 and 2001 (OR, 3.26; 95% CI, 2.44-4.34; P<.001). Problem drug use in 1998 also was associated with increased risk of opioid initiation (OR, 4.03; 95% CI, 2.63-6.17; P<.001) but problem alcohol use in 1998 was not (OR, 0.68; 95% CI, 0.29-1.61; P = .38). Common mental disorder also was associated with increased risk of continuation of opioids from 1998 through 2001 (OR, 2.30; 95% CI, 1.02-5.17; P = .04). Neither problem drug use in 1998 (OR, 0.59; 95% CI, 0.12-2.86; P = .51) nor problem alcohol use in 1998 (OR, 0.29; 95% CI, 0.07-1.22; P = .09) was associated with opioid continuation.
Patients receiving opioids were 2 to 3 times more likely to report a perceived need for mental health treatment (10% vs 19%-32%) (Table 1). Patients receiving opioids at any time point did not report a greater perceived need for substance abuse treatment. Among the 237 regular users of prescribed opioids in 2001, 21 (8.9%) reported opioid use beyond physician's prescription and 8 (38.1%) of these 21 reported having physical or psychological problems because of misusing drugs.
After we adjusted for sociodemographic and 1998 clinical factors (Table 2), including the presence of chronic pain conditions and level of pain interference, common mental disorder in 1998 (OR, 1.96; 95% CI, 1.47-2.62; P<.001) and problem drug use (OR, 2.98; 95% CI, 1.68-5.30; P<.001) but not problem alcohol use (OR, 0.63; 95% CI, 0.35-1.15; P = .13) remained significant predictors of opioid use in 2001. If we adjusted for 2001 rather than 1998 pain interference, common mental disorder in 1998 (OR, 1.77; 95% CI, 1.29-2.43; P<.001) and problem drug use (OR, 2.37; 95% CI, 1.25-4.48; P<.01) remained significant predictors of opioid use in 2001. In the adjusted models, “other” race, education levels of less than high school and college or higher, location in the southern United States, arthritis, chronic back pain, and high pain interference remained significant predictors of opioid use.
In the 1998 adjusted models, any common mental disorder (OR, 1.80; 95% CI, 1.26-2.58; P<.01) and problem drug use (OR, 3.29; 95% CI, 1.79-6.05; P<.001) remained significant predictors of opioid initiation. In these adjusted models, any common mental disorder (OR, 1.75; 95% CI, 0.67-4.57; P = .24) and problem drug use (OR, 1.13; 95% CI, 0.19-6.63; P = .88) were no longer significant predictors of opioid continuation.
This study provides the first prospective, population-based data about the association of common mental and substance use disorders with the regular use of prescribed opioid medication. In adjusted models, respondents with a mental health disorder in 1998 were twice as likely to use prescribed opioids in 2001. Respondents with a mental health disorder who were not using opioids in 1998 were more likely to initiate opioid use during the 3-year follow-up. Although those with problem drug use in 1998 were 3 times as likely to use prescribed opioids in 2001, this group is far smaller than the number of patients with common mental health disorders. Individuals receiving opioids long term also reported greater perceived need for mental health care but not substance abuse care. This association of mental disorders with increased opioid use is consistent with results from our previous cross-sectional analysis of a larger 1998 sample.8
Why are individuals with psychiatric disorders more likely to use opioids long term? It is well known that depression and anxiety disorders are associated with increased rates of psychological and physical symptoms. Pain arising from chronic medical disorders also is rated as more severe in the presence of major depression.20 One intriguing possibility is that opioids may be used to treat a poorly differentiated state of mental and physical pain. We believe effects of mental disorder may be stronger for initiation than continuation of opioids because initiation may be related more directly to patient distress, whereas continuation also is determined by response to opioids, which may be inhibited in those with depression.21
Our findings are consistent with population-based evidence that unmet need for mental health care is associated with higher rates of substance use.22 There is also a centuries' long history of opioids being used by physicians to treat depression and other mental disorders.23 Exogenous opioids elevate mood and endogenous opioids modulate basal emotional state.24 Recent study results suggest that treatment-resistant depression25 and obsessive-compulsive disorder26 may respond to opioid treatment. Use and abuse of opioids appears to be common in individuals with posttraumatic stress disorder, borderline personality disorder, and other dissociative disorders.27,28 However, evidence suggests the condition of these individuals may deteriorate with opioid use,29,30 and they may actually improve psychiatrically by using the opioid antagonist naltrexone hydrochloride.31 Although substance abuse and major depression increase the likelihood of using prescribed opioids, few data support this use because patients with substance abuse and major depression have been excluded from most controlled trials of opioids for chronic pain.32
Our data suggest that diagnosis and treatment of psychiatric disorders are important steps when considering patients for opioid therapy. However, existing guidelines and consensus statements concerning the use of opioids for chronic noncancer pain vary widely in their recommendations about depression and other psychiatric disorders. Some guidelines33,34 mention the importance of the assessment of psychiatric disorders, but other guidelines35,36 and reviews37,38 make no mention of psychiatric disorders. However, adequate care for depression may improve chronic pain, a finding suggested by results from a recent trial of collaborative care for depression in older adults that demonstrated that arthritis pain intensity and pain interference were reduced by improved treatment for depression.39 Such depression treatment effects might reduce the need for long-term analgesic therapy or permit shifting patients to less addictive pain treatments than opioids.
The strengths of our study include the use of prospective data from a large nationally representative sample of the general population, which avoids the selection bias common in the clinical samples in which opioid use has been studied to date and allows interpretation of our results in terms of the national population. The HCC also uses the systematic and well-validated World Health Organization CIDI approach to psychiatric diagnosis. Finally, the HCC provides extensive data about demographic and clinical covariates that could confound the relationship between psychiatric disorders and opioid use.
Our study has a number of limitations. First, the response rates to the telephone survey were 64.0% in 1998 and 69.5% in 2001, which is similar to other general population telephone surveys performed at that time. For example, the Behavioral Risk Factor Surveillance System40 had a 63.5% response rate after determining household eligibility in 1998. Compared with responders to the 2001 survey, nonresponders were more likely to be younger than 30 years; to be of race other than white; to be black or Hispanic; to be single; to have less than a high school education; to have an annual income less than $19 000; to not be insured; to have fair to poor general health; and to meet criteria for dysthymia, generalized anxiety, or panic disorder. Because these are many of the significant predictors of opioid use in 2001, their loss to follow-up introduces a conservative bias to our findings. Second, substance abuse disorders were assessed with abbreviated versions of the World Health Organization CIDI interview that did not fully document all the criteria needed for a drug abuse diagnosis. Third, there was no direct clinical observation of the study subjects. All assessments were conducted by telephone and rely on self-report. Physical health status was documented through a count of chronic conditions and the physical component score of the Medical Outcomes Study Short Form–12. Fourth, subjects read the names of their opioids from their pill bottles and verified that they were taking them regularly, but we do not know why they were prescribed opioids or if they were abusing them. Finally, this is an observational study from which it is not possible to derive unbiased data about the effects of opioid use on clinical outcomes.
In summary, our study results demonstrate that common psychiatric disorders (depression, anxiety, and drug abuse) predict initiation and ongoing regular use of opioids in patients with chronic pain. Although much attention has been given to the possibility of substance abuse in regular users of prescribed opioids, our data suggest that common depressive or anxiety disorders also are associated strongly with opioid use and may pose a more significant clinical problem in this population because they are much more common. Greater attention to these common psychiatric disorders among candidates for opioid therapy is needed if we are to provide patients with chronic pain the best possible health care. This may include screening for psychiatric disorders, ensuring that psychiatric treatment is adequate before initiating opioid therapy, and closer follow-up during opioid therapy for those with psychiatric disorders to prevent negative outcomes such as declining functional status or opioid misuse.
Correspondence: Mark D. Sullivan, MD, PhD, Department of Psychiatry and Behavioral Sciences, PO Box 356560, University of Washington, 1959 NE Pacific St, Seattle, WA 98195-6560 (firstname.lastname@example.org).
Accepted for Publication: July 13, 2006.
Author Contributions: Dr Sullivan 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: Sullivan, Edlund, Unutzer, and Wells. Acquisition of data: Wells. Analysis and interpretation of data: Sullivan, Edlund, Zhang, Unutzer, and Wells. Drafting of the manuscript: Sullivan and Edlund. Critical revision of the manuscript for important intellectual content: Edlund, Zhang, Unutzer, and Wells. Statistical analysis: Edlund and Zhang. Obtained funding: Wells. Administrative, technical, and material support: Wells. Study supervision: Sullivan.
Financial Disclosure: None reported.
Funding/Support: Dr Sullivan was supported partially by a grant from the Greenwall Foundation. Dr Edlund is supported by a Career Development Award from the Veterans Affairs Health Services Research and Development Service. Dr Unützer is supported as a Paul Beeson Faculty Scholar. Healthcare for Communities was funded by the Robert Wood Johnson Foundation (Kenneth Wells, primary investigator).
Role of the Sponsor: The Robert Wood Johnson Foundation had no role in the design and conduct of this study; collection, management, analysis, and interpretation of the data; and preparation, review, or approval of the manuscript.
Acknowledgment: We thank Wayne Katon, MD, and Michael Von Korff, ScD, for helpful comments on an earlier draft of the manuscript and Alison Sattler, BA, for secretarial assistance.