Nonmedical Prescription Opioid Use and Use Disorders Among Adults Aged 18 Through 64 Years in the United States, 2003-2013

First, we examined data of adults who participated in the 2003-2013 National Surveys on Drug Use and Health (NSDUH), which are conducted by the Substance Abuse and Mental Health Services Administration (SAMHSA). They provided nationally representative data on nonmedical use of prescription opioids and use disorders among the civilian, noninstitutionalized population aged 12 years or older in the United States. Excluded from NSDUH were persons without a household address (eg, homeless persons not living in shelters), active-duty military personnel, and institutional residents. The data collection protocol of the NSDUH was approved by the institutional review board at the Research Triangle Institute (RTI) International. Written informed consent was obtained from each study participant. The annual mean weighted response rate of the 2003-2013 NSDUH was 66.3% (range, 60.2%-70.2%),¹² according to the definition of response rate 2 for in-person household surveys by the American Association for Public Opinion Research.¹³ Details regarding NSDUH data collection are provided elsewhere.¹²

Second, we examined data from the 2003-2013 Multiple Cause of Death Files from the National Vital Statistics System.¹⁴ Drug overdose deaths were those assigned an underlying cause of death using the International Classification of Diseases, Tenth Revision (ICD-10) codes X40-X44 (unintentional), X60-X64 (suicide), X85 (homicide), and Y10-Y14 (undetermined intent). Prescription opioid overdose deaths were those assigned specific ICD-10 codes T40.2-T40.4. Similar to NSDUH analyses. We limited analyses to drug overdose deaths among adults aged 18 through 64 years in the United States.

The NSDUH defined 12-month nonmedical use of prescription opioids as use in the prior 12 months without a prescription or as use with a prescription simply for the experience or feeling caused by opioids.¹²If respondents reported using prescription opioids nonmedically in the past year, they were asked to state the number of days they used them nonmedically. Consistent with previous work,^11,15,16 we definedfrequent users as those with 100 days or more of nonmedical opioid use in the past year and highly frequent users as those with 200 days or more. For persons reporting nonmedical prescription opioid use, NSDUH collected sources of the medications, including given by a friend or a relative for free, prescribed by 1 or more physicians, stolen from a friend or a relative, bought from a friend or a relative, and bought from a drug dealer or stranger. Age at first nonmedical use of prescription opioids was also assessed (<18, 18-29, or ≥30 years).

The NSDUH estimated past 12-month major depressive episodes (MDE) and each specific substance use disorder (dependence on or abuse of alcohol, marijuana, cocaine, hallucinogens, heroin, inhalants, or nonmedical use of prescription pain relievers, sedatives, or stimulants) based on assessments of individual diagnostic criteria from the DSM-IV.¹⁷ Nicotine dependence among cigarette smokers was assessed using the nicotine dependence syndrome scale.¹⁸

Sociodemographic characteristics from NSDUH data included age, sex, race/ethnicity, educational attainment, employment status, marital status, health insurance, metropolitan statistical area, census region, and year. Race/ethnicity was NSDUH respondent’s self-classification of racial and ethnic origin and identification based on the classifications developed by the US Census Bureau. This study assessed race/ethnicity because prior research indicates racial/ethnic differences in prescription opioid–related mortality.⁵

All analyses were conducted for persons aged 18 through 64 years. Between 2003 and 2013 for each examined year, descriptive analyses were conducted to estimate the 12-month prevalence of nonmedical use of prescription opioids and prescription opioid use disorders, the proportion of 12-month nonmedical opioid use initiates, the prevalence of frequent users and highly frequent users, and the mean number of days of nonmedical use among adults and among nonmedical users. Similarly, we estimated trends in the rates of overdose deaths involving prescription opioids between 2003 and 2013. For percentage estimates, bivariable logistic regression models were applied to assess the unadjusted annual prevalence, to test for differences between the 2003 estimate and the estimate in each year in 2004-2013 (using PREDMARG and PRED_EFF statements in SUDAAN,^19,20 2-sided t test with a significance level of .05), and to test P values of β coefficients of the year variable. For mean numbers of days of nonmedical use, linear regression models were applied to examine differences between the 2003 estimate and the estimate in each year spanning 2004 through 2013 (using the REFLEVEL statement in SUDAAN,¹⁹ 2-sided t test with a significance level of .05) and to test P values of β coefficients of the year variable. For drug overdose death rates, 2-tailed ztests (with a significance level of .05)²¹ were used to test differences between the 2003 rate and the rate in each year spanning 2004 through 2013. A Bonferroni correction was used to reduce chances of type I errors of testing overall trends. For each outcome, if the overall trend did not meet the statistical significance threshold after the Bonferroni correction, differences between the 2003 estimate and the estimates in other years were not reported.

Bivariable and multivariable logistic regression modeling were applied to assess unadjusted and model-adjusted relative risk (using PREDMARG and PRED_EFF statements in SUDAAN^19,20) for prescription opioid use disorders among nonmedical opioid users. Because major depressive episodes were unavailable in the 2003 NSDUH and because the source of opioids was unavailable in 2003 and 2004 NSDUHs, separate multivariable models were calculated for 2005-2013 with these additional variables included and for 2003-2013 without them. This study used SUDAAN software to account for the complex sample design and sampling weights of NSDUH data. Moreover, bivariable and multivariable zero-truncated negative binomial regression models were applied to examine factors associated with the number of days of nonmedical use of prescription opioids among nonmedical users. Stata statistical software^22,23 was used for a zero-truncated negative binomial regression function that accounts for the complex sample design and sample weights of NSDUH. Multicollinearity (using variance inflation factors) and potential interaction effects between examined factors were assessed and were not identified in final multivariable models. Furthermore, for the prevalence of prescription opioid use disorders among nonmedical opioid users, we conducted tests for interaction effects for specific subgroup comparisons (eg, between non-Hispanic whites and non-Hispanic blacks, between non-Hispanic whites and Hispanics, etc). To test whether 2010 might be a pivotal year for the trends, we repeated these analyses by using the year 2010 as the reference group and examined whether each intensity outcome differed in each year spanning 2011 through 2013.

Based on 472 200 sampled persons, we found that among adults aged 18 through 64 years, the 12-month prevalence of nonmedical use of prescription opioids decreased from 5.4% (95% CI, 5.08%-5.70%) in 2003 to 4.9% (95% CI, 4.58%-5.22%) in 2013 (absolute difference, −0.5%; 95% CI, −0.11% to −0.89%), and the 12-month initiation rate decreased from 1.0% (95% CI, 0.86%-1.11%) in 2003 to 0.6% (95% CI, 0.52%-0.71%) in 2013 (absolute difference, −0.4%; 95% CI, −0.52% to −0.22%; Table 1). In contrast, the 12-month prevalence of prescription opioid use disorders increased from 0.6% (95% CI, 0.54%-0.76%) in 2003 to 0.9% (95% CI, 0.75%-1.01%) in 2013 (absolute difference, 0.3%; 95% CI, 0.03%-0.43%) (Table 1and eFigure 1 in the Supplement). The 12-month prevalence of high-frequency use (≥200 days) also increased from 0.3% (95% CI, 0.19%-0.35%) in 2003 to 0.4% (95% CI, 0.31%-0.48%) in 2013 (absolute difference, 0.1%; 95% CI, 0.01%-0.29%). The drug overdose death rates involving prescription opioids increased from 4.5 per 100 000 (95% CI, 4.42-4.61) in 2003 to 7.8 per 100 000 (95% CI, 7.64-7.89) in 2013 (absolute difference 3.3, 95% CI, 3.09-3.41) (Table 1 and eFigure 2 in the Supplement).

Consistent with Table 2 and based on the results of testing for interactions for subgroup comparisons, compared with non-Hispanic white users, both non-Hispanic blacks (P = .001) and Hispanics (P = .006) had lower prevalence of prescription opioid use disorders. The prevalence was lower among users with a college degree or beyond than among users without a high school diploma (P < .001). Compared with users with full-time employment, users who were disabled for work had higher prevalence of the disorders (P < .001). The prevalence was higher among users without health insurance (P = .008) and among users with Medicaid (P < .001) than among users with private health insurance only. Nonmedical users with nicotine dependence had higher prevalence of prescription opioid use disorders than users without (P < .001). The prevalences among those with alcohol, marijuana, cocaine, hallucinogen, heroin, stimulant, and sedative use disorders were higher than users without the corresponding disorders (with each subgroup comparison tested separately for interaction, P < .001). With data from 2005-2013 NSDUHs, users with major depressive episode had a higher prevalence of prescription opioid use disorders than users without (P < .001); and compared with users who obtained prescription opioids from friends or relatives for free, users who stole or bought them from friends or relatives, who were prescribed by physicians, or who bought them from drug dealers or strangers had higher prevalence of prescription opioid use disorders (with each subgroup comparison tested separately for interaction, P < .001).

Among users, the mean number of days of nonmedical use of prescription opioids was higher in every year spanning 2008 through 2013 than in 2003 (Table 3). The results of the zero-truncated negative binominal regression model show consistent results even after controlling for potential confounding factors (adjusted incidence rate ratios [IRRs], 1.19-1.29). Users in 2010 had similar number of days of nonmedical use of prescription opioids compared with users in each year spanning 2011 through 2013. Non-Hispanic black users took prescription opioids nonmedically more often than non-Hispanic white users (IRR, 1.22; 95% CI, 1.08-1.39). Users with high school education or higher used prescription opioids nonmedically less often than users without a high school diploma (IRRs, 0.46-0.80). Users who were disabled for work used prescription opioids nonmedically more often than users with full-time employment (IRR, 1.33; 95% CI, 1.14-1.55). Users without health insurance or users with Medicaid used prescription opioids nonmedically more often than users with private insurance only (IRRs, 1.26-1.36).

Users with nicotine dependence used prescription opioids nonmedically more often than users without (IRR, 1.54, 95% CI, 1.43-1.65). Users with cocaine, heroin, stimulant, and sedative use disorders used prescription opioids nonmedically more often than users without the corresponding disorders (IRRs, 1.46-1.96). The multivariable model based on 2005-2013 NSDUH data show that users who stole opioids from friends or relatives, who were prescribed by physicians, and who bought them from friends or relatives or from drug dealers or strangers used prescription opioids nonmedically more often than users who obtained prescription opioids from friends or relatives for free (IRRs, 1.28-2.81).

Using nationally representative data, we found during years 2003 through 2013, among adults aged 18 through 64 years, the percentage of nonmedical use of prescription opioids decreased. However, we found several measures of high-risk nonmedical prescription opioid use that either increased or persisted through 2013. Specifically, we found significant increases in high-intensity prescription opioid use (measured by both prevalence of prescription opioid use disorders and high frequency of use) among nonmedical prescription opioid users in each year from 2010 through 2013 compared with those in 2003, even after controlling for potential confounding factors. Furthermore, when compared with those in 2010, additional analyses at bivariable and multivariable levels consistently showed that both intensity outcomes did not decrease in any year from 2011 through 2013. These national increases also paralleled the increases in prescription opioid-related overdose death rates in the United States identified in our study using a separate national data set. These results are consistent with persistent high-risk nonmedical use among adults through at least 2013. The significant increases in high-intensity (and therefore especially high-risk) prescription opioid use contrast a slight downward trend in the overall percentage of nonmedical use.

A recent study, using indicators from the Researched Abuse, Diversion, and Addiction-Related Surveillance (RADARS) System, found that the abuse and diversion of prescription opioids increased between 2002 and 2010 and plateaued or decreased between 2011 and 2013.²⁴ RADARS provides proprietary, nonnationally representative data but does not aim to estimate prevalence.^24,25 For example, not all parts of the United States have prescription drug diversion agencies, and operational details and emphases of diversion agencies vary according to the needs of the community and local drug activity.^24,25 Since the completeness and accuracy of RADARS is difficult to ascertain, the nationally representative results in our study may be especially important in providing an accurate picture of the current status of the epidemic.

The epidemic has evolved over time. The increases in the prevalence of adults with prescription opioid use disorders and high-frequency use identified in our study are a major public health concern. Furthermore, the increases in high-intensity prescription opioid use occurred in the context of increased prescribing of prescription opioids.¹⁵ Previous studies have shown a strong relationship between inappropriate opioid prescribing and negative health outcomes.²⁶ The percentage of prescription opioid users who used an opioid stronger than morphine or morphine-equivalent increased from 57.6% in the years 1999 through 2002 to 80.0% in the years 2011 through 2012.²⁷ Also, usage of higher amounts of prescription opioids is a significant risk factor for overdose death.²⁸

Despite these limitations, we found a significangt decrease in the precentage of nonmedical use of prescription opioids, as well as significant increases in the prevalence of prescription opioid use disorders, high-frequency use, and related mortality among adults aged 18 through 64 years in the United States over the past decade. Furthermore, the increases identified in this study occurred in the context of increasing heroin use and heroin-related overdose deaths in the United States,³⁹ supporting a need to address nonmedical use of prescription opioid and heroin abuse in a coordinated and comprehensive manner.^31,33,40

During the 2003-2013 years, among adults aged 18 through 64 years, the percentage of nonmedical use of prescription opioids decreased. In contrast, the prevalence of prescription opioid use disorders, frequency of use, and related mortality increased.

 read more at JAMA