Secondary Logo

Journal Logo

High-dose prescribed opioids are associated with increased risk of heroin use among United States military veterans

Banerjee, Geetanjolia; Edelman, E. Jenniferb,c; Barry, Declan T.d,e,f; Crystal, Stepheng; Gordon, Kirsha S.b,h; Gordon, Adam J.i; Gaither, Julie R.c,j; Green, Traci C.a,k,l; Kerns, Robert D.d,h; Manhapra, Ajayd,m; Moore, Brent A.d,h; Fiellin, David A.b,c; Marshall, Brandon D.L.a,*

doi: 10.1097/j.pain.0000000000001606
Research Paper
Editor's Choice
Global Year 2019

Despite evidence linking increased risk of opioid use disorder with specific opioid-prescribing patterns, the relationship between these patterns and heroin use is less understood. This study aimed to determine whether dose and duration of opioid prescriptions predict subsequent heroin use in United States veterans. We analyzed data from 2002 to 2012 from the Veterans Aging Cohort Study, a prospective cohort study. We used inverse probability of censoring weighted Cox regression to examine the relationship between self-reported past year heroin use and 2 primary predictors: (1) prior receipt of a high-dose opioid prescription (≥90 mg morphine equivalent daily dose), and (2) prior receipt of a long-term opioid prescription (≥90 days). Heroin use was ascertained using most recent value of time-updated self-reported past year heroin use. Models were adjusted for HIV and hepatitis C virus infection status, sociodemographics, pain interference, posttraumatic stress disorder, depression, and use of marijuana, cocaine, methamphetamines, and unhealthy alcohol use. In the final model, prior receipt of a high-dose opioid prescription was associated with past year heroin use (adjusted hazard ratio use = 2.54, 95% confidence interval: 1.26-5.10), whereas long-term opioid receipt was not (adjusted hazard ratio = 1.09, 95% confidence interval: 0.75-1.57). Patients receiving high-dose opioid prescriptions should be monitored for heroin use. These findings support current national guidelines recommending against prescribing high-dose opioids for treating pain.

In this study, veterans receiving high-dose prescription opioids (defined as ≥90 mg morphine equivalent daily dose) were at an increased risk of subsequent heroin use.

aDepartment of Epidemiology, Brown University School of Public Health, Providence, RI, United States

bDepartment of Internal Medicine, Yale School of Medicine, Yale University, New Haven, CT, United States

cCenter for Interdisciplinary Research on AIDS, Yale School of Public Health, Yale University, New Haven, CT, United States

dDepartment Psychiatry, Yale School of Medicine, Yale University, New Haven, CT, United States

eThe APT Foundation, Pain Treatment Services, New Haven, CT, United States

fChild Study Center at Yale School of Medicine, Yale University, New Haven, CT, United States

gCenter for Health Services Research, Institute for Health, Health Care Policy, and Aging Research, Rutgers University, New Brunswick, NJ, United States

hVA Connecticut Healthcare System, West Haven, CT, United States

iDepartment of Medicine, University of Utah School of Medicine, Salt Lake City, UT, United States

jDepartment of Pediatrics, Yale School of Medicine, Yale University, New Haven, CT, United States

kDepartment of Emergency Medicine, Boston University, Boston, MA, United States

lBoston Medical Center Injury Prevention Center, Boston, MA, United States

mAdvanced PACT Pain Clinic, Hampton VA Medical Center, Hampton, VA, United States

nVA New England MIRECC, West Haven, CT, United States

*Corresponding author. Address: Department of Epidemiology, Brown University, 121 South Main St, Room 208, Box G-S121-2, Providence, RI 02912, United States. Tel.: (401)-863-6427. E-mail address: (B.D.L. Marshall).

Sponsorships or competing interests that may be relevant to content are disclosed at the end of this article.

Supplemental digital content is available for this article. Direct URL citations appear in the printed text and are provided in the HTML and PDF versions of this article on the journal's Web site (

Back to Top | Article Outline

1. Introduction

The proliferation of prescription opioids for the treatment of chronic pain has been associated with significant increases in the nonmedical use of prescription opioids (NMUPO), opioid use disorder (OUD), and opioid overdose deaths.3,11,58 The prescription of high-dose opioids, typically defined as morphine equivalent daily doses (MEDD) of 90 or more milligrams, is specifically associated with adverse drug events,7,17,19,35,50 morbidity, and mortality.21 Long-term opioid therapy (LTOT) has also been associated with the development of OUD;21 the risk of incident OUD is greater with higher dose and longer duration of LTOT.7,19,45 Despite the substantial body of evidence linking an increased risk of OUD with specific prescription opioid use patterns,39 the association between receipt of prescription opioids and subsequent heroin use is less understood. The use of heroin increases the risk of opioid-attributable overdose, the exposure to potential contaminants such as illicitly manufactured fentanyl, and an increased likelihood of acquisition and transmission of blood-borne diseases such as HIV and hepatitis C virus.6,32,44

Understanding the relationship between prescription opioid use and initiation of heroin use among veterans is particularly important. Clinicians may prescribe opioids to men and women during military service or soon after discharge to address injuries sustained during combat, chronic pain, and/or psychological and physical ailments that can arise while readjusting to civilian life.25 Prior studies conducted by our research team have shown that hydrocodone, codeine, and oxycodone are the 3 most common opioids prescribed to US veterans engaged in care.18 A recent report from the Department of Veterans' Affairs (VA) notes that more than one million US veterans engaged in care in the VA health system receive prescription opioids for pain, and nearly half of them continue beyond 90 days.20,51 Appropriate prescribing of opioids to this population, with a higher underlying prevalence both of substance use disorders and of pain conditions than the general population, is particularly nuanced and challenging.1,3,26

The impact of prescription opioid receipt on subsequent heroin use is a phenomenon that has been described in the US mainstream media and among particular subpopulations,8,31,36,47,53 but the scope of this issue among veterans is still largely unexplored. In the current era, regular heroin use among the treatment-seeking general population is often preceded by NMUPO.12 We recently demonstrated that NMUPO was a strong and independent predictor of heroin initiation among veterans receiving care in the VA.2 However, it is less clear how receipt of prescription opioids from a VA provider is associated with later heroin use. A better understanding of the relationship between specific opioid prescribing practices and the subsequent use of heroin could assist in creating strategies that mitigate such risk in veterans.

Back to Top | Article Outline

2. Material and methods

2.1. Data

This study used longitudinal behavioral and pharmacy fill/refill data from the Veterans Aging Cohort Study (VACS); detailed data collection and survey methodology for VACS is previously described.33,34 In brief, the VACS is a prospective cohort study of military veterans living with HIV and uninfected veterans receiving medical care at 8 US VA sites. Since June of 2002, VACS has enrolled more than 7000 patients from the infectious disease or general medical clinics in Atlanta, Baltimore, Houston, Los Angeles, Pittsburgh, Washington DC, and multiple sites in New York City. This longitudinal study contains variables from 6 follow-up behavioral surveys, completed approximately every 18 months, from 2002 to 2012. The survey data are linked to VA electronic medical records (EMR) containing data on prescribed medications based on pharmacy dispensing, medical diagnoses, and laboratory results for each patient. The VACS was approved by the institutional review boards at each participating VA Medical Center and affiliated academic institutions.

Back to Top | Article Outline

2.2. Participant eligibility

Of the 7502 VACS participants who were potentially eligible for inclusion in our study because they had completed at least one survey, we excluded 689 (9.2%) of participants who reported any illicit opioid use (including heroin and NMUPO) in the year before the first survey the patient completed (baseline survey). We also excluded participants who were never prescribed opioids from the VA during the 10 years of follow-up, n = 2980 (39.7%), as well as participants who did not complete at least one additional follow-up survey, n = 263 (3.5%). The final analytic sample consisted of 3570 veterans. In all analyses, dropout was defined as a failure to complete any additional surveys after the baseline survey, while having been enrolled in the VACS study for greater than or equal to 2 years.

Back to Top | Article Outline

2.3. Primary exposures

Consistent with the methods outlined in previous VACS analyses using prescription opioid data from pharmacy records,18,22,58 we determined receipt of all outpatient opioids, including codeine, hydrocodone, oxycodone, oxycodone, morphine, fentanyl, hydromorphone, methadone for the treatment of pain, tramadol, and propoxyphene. We then determined: (1) the average MEDD supplied in the year before each follow-up survey, and (2) number of days supplied in the year before each follow-up survey, starting with the baseline survey. Average MEDD was dichotomized into high dose (≥90 mg) and low dose (<90 mg). This definition of high-dose prescription opioid receipt was based on VA and CDC opioid-prescribing guidelines.16,56 The number of days supplied in the year before was used to create a dichotomous indicator variable for LTOT, defined as ≥90 days of continuous use, allowing for a 30-day gap between fill/refill in a given year.42 We also categorized opioids into short-acting (ie, with a duration of action generally less than 12 hours) vs long-acting medications (including sustained- and extended-release products), according to Drug Enforcement Administration classifications.54

Back to Top | Article Outline

2.4. Primary outcome

Self-reported heroin use during the follow-up surveys was ascertained starting with a participant's first follow-up survey and was based on the time-updated survey item: “How often in the past year you used each drug—Heroin?” Past-year heroin use was defined as any change in respondent's answer to the previous survey question from “Never” or “No use in the past year,” to indicating any frequency of heroin use in the previous year. We used the date of the corresponding survey to approximate the date of past-year heroin use.

Back to Top | Article Outline

2.5. Confounders and other covariates

Selection of independent variables and potential confounders that were included in the analysis was based on a review of prior literature.15,30,38,41,43 Baseline VACS survey demographics included age, sex, race/ethnicity, marital status, employment status, and gross annual income. HIV status was identified using VA Immunology Case Registry, and hepatitis C virus status was determined using International Classification of Disease (ICD)-9 codes and laboratory data. Level of social isolation was ascertained from responses to the following baseline survey items: “How often do you see or hear from relatives or close friends? Would you say less than once a month, about once a month, a few times a month, a few times a week, every day?” and “How many close friends or family do you have with whom you feel at ease, can talk about private matters, or can call on for help?” We determined whether the participant had ever been homeless or stayed at a homeless shelter from the following questions on the baseline survey: “Have you ever been without a permanent address that you call home?” and “Have you ever stayed one or more nights in a shelter, on the street, in a park or an abandoned building?” In addition, we included time-updated survey variables for: other drug use (including marijuana, cocaine, and methamphetamines); unhealthy alcohol use; and diagnoses of posttraumatic stress disorder (PTSD) and major depression. Any past-year use of marijuana, cocaine, and methamphetamines was dichotomized as yes/no. Unhealthy alcohol use was characterized using the Alcohol Use Disorders Identification Test (AUDIT)-C; a score of 4 and above indicated unhealthy alcohol use for men, and a threshold of 3 for women.14,40 Diagnoses of PTSD, major depression, and OUD were ascertained from ICD-9 codes and updated at every survey wave. Pain interference in daily life was ascertained from responses to the following time-updated survey item: “During the past 4 weeks, how much did pain interfere with your normal work (including both work outside the home and housework)?” We dichotomized participants' responses into “some amount of pain interference in daily life” vs “no pain interference in daily life.”

Back to Top | Article Outline

2.6. Analyses

We analyzed patterns of opioid prescribing over time for each of the 6 follow-up waves using the Cochran–Armitage test for trend. We used the χ2 or the Fisher exact test to examine the baseline correlates of receipt of high-dose prescription opioids, as well as past-year heroin use at any time over the follow-up period. Unadjusted Kaplan–Meier analysis and the log-rank test were then used to calculate the incidence of past-year heroin use and compare time to first report of heroin use during follow-up, stratifying by the past-year receipt of high-dose opioid prescription at baseline and past-year receipt of LTOT, respectively. Next, we used Cox proportional hazards regression to estimate crude hazard ratios and 95% confidence intervals (CIs) related to factors associated with heroin use. We then constructed a separate multivariable Cox model, including all variables assessed in bivariable analyses, for the 2 main exposure variables: prior receipt of high-dose opioid prescription and prior receipt of LTOT. All variables were found to meet the proportional hazards assumption for the Cox regression models (ie, none exhibited significant deviance from this assumption at P < 0.05).10 Because employment status was highly correlated with education (P < 0.001) and income (P < 0.001), we did not include it in the fully adjusted final model.

We created inverse probability of censoring weights to account for potential biases arising from differential dropout. Weights were obtained through fitting a weighted pooled logistic regression model for dropout, including baseline and time-varying predictors of dropping out, defined as not completing all 6 follow-up surveys. We included a robust sandwich estimator in all Cox regression models to account for potential clustering heterogeneity by study site, and the estimation required to create the weights.27,29 Analyses were conducted using SAS version 9.4.

Back to Top | Article Outline

2.7. Sensitivity analyses

To address the potential for confounding and selection bias due to the inclusion of time-updated covariates affected by prior exposure using a standard regression approach, we conducted a sensitivity analysis using inverse probability of treatment and censoring weights.13 In brief, we used weighting to create a pseudopopulation in which exposure and dropout were independent of measured confounders. We used pooled logistic models to estimate the inverse of the probability of each participant's probability of receiving a high-dose prescription opioid and receiving LTOT, respectively. The numerator of the stabilized weights included baseline covariates that predicted the probability of the exposure, and the denominator included baseline and time-updated covariates that were included in the main analysis. We examined the weights for means, SDs, and extreme minimum and maximum values. Finally, we conducted a sensitivity analysis in which opioid dose was categorized as a 3-level variable in morphine milligram equivalents (MME): low (0-49 MME), medium (50-89 MME), and high (≥90 MME).

Back to Top | Article Outline

3. Results

3.1. Sample composition

The mean age of VACS members in the final analytic sample was 49.8 (SD = 9.0 years); 2302 (65.1%) participants were black, 324 (9.2%) were Hispanic, and 764 (21.6%) were white. The majority of participants were male (94.5%); approximately half of them were living with HIV and 32.9% were positive for hepatitis C virus infection. At baseline, 235 (6.6%) participants reported ever having engaged in injection drug use, and 38.6% reported having ever spending the night in a homeless shelter, street, or abandoned building. At baseline, 28.0% of participants reported that they were currently working for wages, 9.9% reported that were currently seeking employment and unemployed for less than 1 year, 33.4% reported that they were currently unable to work, 7.6% reported being self-employed, and 16.2% reported being retired. Of the 3570 participants who were in the analytic sample, 1160 (32.5%) were lost to follow-up over the study period.

Back to Top | Article Outline

3.2. Trends in prescription opioid receipt from the Veterans' Affairs

Of the 3570 eligible participants, the proportion of study participants who were receiving high-dose prescription opioids (≥90 mg MEDD) did not change over the follow-up period and ranged from 4.4% to 5.9% (the Cochran–Armitage test for trend P = 0.222). The proportion receiving LTOT (≥90 consecutive days) steadily increased from 23.3% in 2002 to 39.9% in 2012 (the Cochran–Armitage test for trend P < 0.001). A majority of participants received only short-acting opioids from the VA (Table 1).

Table 1

Table 1

Back to Top | Article Outline

3.3. Risk factors for past-year heroin use

Over the 10-year follow-up period, 174 (4.9%) participants reported past-year heroin use. At baseline, study participants who reported past-year heroin use were more likely to be black, report a lower gross annual income, report past-year cocaine and methamphetamine use, and have experienced homelessness and social isolation (Table 2).

Table 2

Table 2

Table 2-a

Table 2-a

The 10-year unadjusted incidence rate of past-year heroin use was 8.15 events per 1000 person-years. Time to past-year heroin use differed significantly between participants who had prior receipt of a high-dose prescription opioid at baseline and participants who had no prior receipt of high-dose opioids; the log-rank test was significant at P < 0.001 (Fig. 1).

Figure 1.

Figure 1.

By contrast, the rate of past-year heroin use did not differ significantly between participants who had long-term prior receipt of prescription opioids and participants who had short-term prior receipt of prescription opioids (P = 0.469, figure not shown). The rate of past-year heroin use did not vary significantly between participants receiving short-acting opioids, long-acting opioids, or both (data not shown).

Prior receipt of a high-dose opioid prescription opioid, having HCV infection, fewer years of education, diagnosis with PTSD, diagnosis with depression, and self-reported pain interference were all significantly associated with an increased risk of past-year heroin use in unadjusted Cox regression analyses (Table 3). In the first fully adjusted and weighted multivariable Cox regression analysis, prior receipt of high-dose prescription opioids from the VA was positively and significantly associated with past-year heroin use (adjusted hazard ratio [AHR] = 2.54, 95% CI = 1.26-5.10, Table 4). In the second fully adjusted Cox regression model, prior receipt of LTOT from the VA was not significantly associated with past-year heroin use (AHR = 1.09, 95% CI = 0.75-1.57). Being age 50 to 56 years, compared with participants younger than 42 years of age, having HCV infection, past-year cocaine use, and past-year stimulant use all significantly increased the hazard of past-year heroin use in both models. In addition, participants reporting pain interference in daily life had an increased risk of heroin use.

Table 3

Table 3

Table 4

Table 4

Back to Top | Article Outline

3.4. Sensitivity analyses

The Cox regression with prescription opioid dose categorized into low, medium, and high revealed that participants receiving a medium dose vs a low dose did not have a significantly increased crude hazard ratio of past-year heroin use. In the fully adjusted model, high dose compared to low dose was positively and significantly associated with past-year heroin use (AHR = 3.16, 95% CI = 1.75-5.68).

The stabilized inverse probability of exposure weights was statistically well behaved for both primary exposures of interest. The mean of the weights for receipt of high-dose prescription opioids was 0.99 (SD = 0.23) and they ranged from 0.25 to 4.51. The mean of the weights for receipt of a long-term prescription opioid was 1.00 (SD = 0.30) and they ranged from 0.18 to 3.61. In the sensitivity analysis with treatment and censoring weights, participants who received high-dose prescription opioids from the VA had 2.11 the hazard of past-year heroin use compared to participants who received prescription opioids in doses less than 90 mg MEDD (95% CI: 1.30-3.43). The receipt of LTOT from the VA was not significantly associated with past-year heroin use in the sensitivity analysis.

Back to Top | Article Outline

4. Discussion

In this study, we examined the relationship between specific opioid-prescribing patterns and subsequent heroin use in a population of US military veterans receiving medical care in the VA system. Our results indicate that among veterans who reported no past-year illicit opioid use at baseline and who received prescription opioids from the VA, prior receipt of high-dose prescription opioids increased the risk of subsequent heroin use over follow-up.

Our findings are broadly consistent with the literature, which shows that veterans who use heroin are also more likely to engage in polysubstance use, including use of nonopioid illicit drugs.48,49 The proportion of participants who reported past-year heroin use at each year of follow-up was slightly higher than the percent of individuals who reported past-year heroin use in the 2013 National Survey of Drug Use and Health.36 However, we suggest caution when comparing our study results to those of a national survey of the general population because, by design, all study participants received opioid analgesics at least once during the study period, and approximately half were persons living with HIV. Both of these eligibility criteria likely elevated the probability of heroin use over the study period. However, by identifying a population with a higher prevalence of risk factors for future heroin use, clinicians can more appropriately direct these patients to overdose education and naloxone distribution,46 in hopes of reducing the risk of potential heroin-attributable overdoses.

At baseline, study participants who reported past-year heroin use were more likely to be black. In addition, black participants were at an increased risk of past-year heroin use in the unadjusted Cox regression analyses. However, in multivariable models, the adjusted rate of past-year heroin use did not differ significantly between black participants and participants of other races. It is possible that variables serving as proxies for structural inequalities and included in the final model (eg, education and income) confounded the association between race and past-year heroin use. Because greater levels of education and a higher number of close friends/family were protective against past-year heroin use in our study, involvement with support groups and improved access to opioid safety and overdose prevention programs may mitigate the risk of future heroin use. In addition, peer engagement and outreach programs have proven to be an effective intervention in encouraging treatment seeking for OUDs, as well as for connecting veterans to evidence-based approaches to addiction treatment (eg, opioid agonist treatment with methadone or buprenorphine, and nonpharmacological treatments).5

Our findings suggest that diagnosis and treatment of psychiatric disorders are important considerations when prescribing opioid therapy, although existing guidelines concerning the use of opioids vary widely in their recommendations regarding patients with comorbid depression and other psychiatric disorders.52 Lower household wealth is associated with chronic pain that substantially limits participation in daily activities.28 Therefore, the fact that work-related pain interference in daily life remained associated with an increased risk of past-year heroin use in the fully adjusted model suggests that patients may be engaging in heroin use to alleviate pain that has not been adequately addressed by prescription opioid therapy and that efforts to reduce the burden of disabling chronic pain should prioritize socioeconomically vulnerable groups.

More than half of VA primary care patients report pain, with many reporting chronic pain; however, current VA/Department of Defense opioid therapy guidelines recommend use of long-acting opioids only for persistent pain, and prescription of opioid therapy for chronic pain only after other nonopioid analgesic pharmacotherapies and nonpharmacological pain treatments have insufficiently improved pain-related function.37 Although our results further substantiate current VA and CDC opioid-prescribing guidelines, which recommended prescribing opioids for short durations and recommend against prescribing opioids in doses greater than 90 mg MEDD,55 Gaither et al.22 found that from 1998 to 2010, the majority of VA patients initiating LTOT did not receive opioid therapy guideline-concordant care (although older guidelines were in effect when this study was published). Future research is needed to determine the extent to which opioid prescribing is consistent with current guidelines.

We posit that prior receipt of long-term opioid prescriptions was not associated with past-year heroin use because our data were collected before the enactment of recent CDC and VA/DOD guidelines regarding the tapering of high-dose prescription opioids, and it is unlikely that opioid tapering and/or abrupt discontinuation of opioid therapy drove patients to engage in heroin use during this period.16 Nonetheless, future studies should examine whether abrupt discontinuation or inappropriate tapering of opioid therapy is associated with an increased risk of transitioning to heroin use among veterans and high-risk populations. This is particularly important in the context of the VA's Opioid Safety Initiative, which strongly encourages opioid tapering, if not discontinuation, without explicit examination of unintended harms of this initiative.59 It is essential to continue addressing unmet pain management needs of veterans, as is also recommended by guidelines from the American College of Physicians, which specifically recommends increased use of nonopioid pharmacological and nonpharmacological approaches.9

Our study has a number of limitations. First, there is a potential for confounding by indication, whereby patients at higher risk of heroin use were also more likely to receive high-dose prescription opioids. We addressed this by using inverse probability weighting methods, and creating weights based on patients' probability of receiving high-dose opioids, conditioned on a comprehensive list of potential, time-updated confounders.58 Second, our exposure data were based on pharmacy fill/refill records from the VA, and we were not able to account for opioid prescriptions filled outside of the VA.23,24 One recent study demonstrated that 18% of veterans receive opioids both inside and outside of the VA.4 In that vein, receipt of a prescription opioid prescription may not have accurately captured the true amount of opioids actually being taken by VACS participants. There is also some likelihood of measurement error and bias in our outcome variable: past-year heroin use. Although this variable captures self-report of any heroin use in the 10-year follow-up period, we were not able to discern whether this is truly lifetime incident heroin use or represents reinitiation of heroin use. We only used a single domain (eg, work-related interference) to measure pain interference in daily life. The complexity of chronic pain and its interference in daily activities is likely not fully captured by this question. In addition, almost 50% of participants reported being retired or unable to work at baseline, which may have affected the validity and measurement of this item. Moreover, because we did not measure employment status over time, we are unable to examine whether daily pain interfered in employment status, or loss of employment related to chronic pain increased the risk of heroin use. Despite these limitations, our study has important implications for clinical care, policy, and future research. The finding that prior receipt of a high-dose prescription opioid may increase the risk of past-year heroin use among veterans receiving care at the VA supports the current CDC and VA guidelines that recommend prescribing opioid at lower doses. Our results indicate that further efforts are needed to develop effective screening strategies to identify heroin use among those prescribed with high-dose opioids.

Back to Top | Article Outline

Conflict of interest statement

The authors have no conflict of interest to declare.

Back to Top | Article Outline


This work was funded by a NIDA NIH Dissertation Grant, R36DA042877, as well as NIH grants U01 AA020790, U24 AA020794, and RO1 DA040471. BDLM is supported by NIH grant P20-GM125507.

Back to Top | Article Outline

Supplemental audio content

Audio accompanying this abstract is available online as supplemental digital content at

Back to Top | Article Outline


[1]. Back SE, Killeen TK, Teer AP, Hartwell EE, Federline A, Beylotte F, Cox E. Substance use disorders and PTSD: an exploratory study of treatment preferences among military veterans. Addict Behav 2014;39:369–73.
[2]. Banerjee G, Edelman EJ, Barry DT, Becker WC, Cerda M, Crystal S, Gaither JR, Gordon AJ, Gordon KS, Kerns RD, Martins SS, Fiellin DA, Marshall BD. Non-medical use of prescription opioids is associated with heroin initiation among US veterans: a prospective cohort study. Addiction 2016;111:2021–31.
[3]. Becker WC, Sullivan LE, Tetrault JM, Desai RA, Fiellin DA. Non-medical use, abuse and dependence on prescription opioids among U.S. adults: psychiatric, medical and substance use correlates. Drug Alcohol Depend 2008;94:38–47.
[4]. Becker WC, Fenton CT, Brandt CA, Doyle EL, Francis J, Goulet J, Moore BA, Torrise V, Kerns RD, Kreiner PW. Multiple sources of prescription payment and risky opioid therapy among veterans. Med Care 2017;7:S33–6.
[5]. Bennett AS, Elliott L, Golub A, Wolfson-Stofko B, Guarino H. Opioid-involved overdose among male Afghanistan/Iraq-era U.S. Military veterans: a multidimensional perspective. Subst Use Misuse 2017:1–11.
[6]. Bode AD, Singh M, Andrews J, Kapur G, Baez AA. Fentanyl laced heroin and its contribution to a spike in heroin overdose in Miami-Dade County. Am J Emerg Med 2017;35:1364–5.
[7]. Bohnert ASB, Valenstein M, Bair MJ, Ganoczy D, McCarthy JF, Ilgen MA, Blow FC. Association between opioid prescribing patterns and opioid overdose-related deaths. JAMA 2011;305:1315–21.
[8]. Boscarino JA, Rukstalis M, Hoffman SN, Han JJ, Erlich M, Gerhard GS, Stewart WF. Risk factors for drug dependence among out-patients on opioid therapy in a large US health-care system. Addiction 2010;105:1776–82.
[9]. Chou R, Deyo R, Friedly J, Skelly A, Hashimoto R, Weimer M, Fu R, Dana T, Kraegel P, Griffin J, Grusing S, Brodt ED. Nonpharmacologic therapies for low back pain: a systematic review for an American College of Physicians clinical practice guideline. Ann Intern Med 2017;166:493–505.
[10]. Christensen E. Multivariate survival analysis using Cox's regression model. Hepatology 1987;7:1346–58.
[11]. Cicero TJ, Surratt H, Inciardi JA, Munoz A. Relationship between therapeutic use and abuse of opioid analgesics in rural, suburban, and urban locations in the United States. Pharmacoepidemiol Drug Saf 2007;16:827–40.
[12]. Cicero TJ, Ellis MS, Surratt HL, Kurtz SP. The changing face of heroin use in the United States: a retrospective analysis of the past 50 years. JAMA Psychiatry 2014;71:821–6.
[13]. Cole SR, Hernan MA. Constructing inverse probability weights for marginal structural models. Am J Epidemiol 2008;168:656–64.
[14]. Dawson DA, Smith SM, Saha TD, Rubinsky AD, Grant BF. Comparative performance of the AUDIT-C in screening for DSM-IV and DSM-5 alcohol use disorders. Drug Alcohol Depend 2012;126:384–8.
[15]. Dobscha SK, Morasco BJ, Duckart JP, Macey T, Deyo RA. Correlates of prescription opioid initiation and long-term opioid use in veterans with persistent pain. Clin J Pain 2013;29:102–8.
[16]. Dowell D, Haegerich TM, Chou R. CDC guideline for prescribing opioids for chronic pain—United States, 2016. JAMA 2016;315:1624–5.
[17]. Dunn KM, Saunders KW, Rutter CM, Banta-Green CJ, Merrill JO, Sullivan M, Welsner CM, Silverberg MJ, Campbell CI, Psaty BM, Von Korff M. Opioid prescriptions for chronic pain and overdose. Ann Intern Med 2010;152:85–92.
[18]. Edelman EJ, Gordon K, Becker WC, Goulet JL, Skanderson M, Gaither JR, Brennan Braden J, Gordon AJ, Kerns RD, Justice AC, Fiellin DA. Receipt of opioid analgesics by HIV-infected and uninfected patients. J Gen Intern Med 2013;28:82–90.
[19]. Edlund MJ, Martin BC, Russo JE, DeVries A, Braden JB, Sullivan MD. The role of opioid prescription in the incident opioid abuse and dependence among individuals with chronic noncancer pain. Clin J Pain 2014;30:557–64.
[20]. Edlund MJ, Austen MA, Sullivan MD, Martin BC, Williams JS, Fortney JC, Hudson TJ. Patterns of opioid use for chronic noncancer pain in the Veterans Health Administration from 2009 to 2011. PAIN 2014;155:2337–43.
[21]. Franklin GM, Fulton-Kehoe D, Turner JA, Sullivan MD, Wickizer TM. Changes in opioid prescribing for chronic pain in Washington State. J Am Board Fam Med 2013;26:394–400.
[22]. Gaither JR, Goulet JL, Becker WC, Crystal S, Edelman EJ, Gordon K, Kerns RD, Rimland D, Skanderson M, Weisberg DF, Justice AC, Fiellin DA. Guideline-concordant management of opioid therapy among human immunodeficiency virus (HIV)-infected and uninfected veterans. J Pain 2014;15:1130–40.
[23]. Gellad WF, Zhao X, Thorpe CT, Thorpe JM, Sileanu FE, Cashy JP, Mor M, Hale JA, Radaomski T, Hausmann LR, Fine MJ, Good CB. Overlapping buprenorphine, opioid, and benzodiazepine prescriptions among veterans dually enrolled in Department of Veterans Affairs and Medicare Part D. Subst Abus 2017;38:22–5.
[24]. Gellad WF, Thorpe JM, Zhao X, Thorpe CT, Sileanu FE, Cashy JP, Hale JA. Impact of dual use of department of veterans Affairs and medicare Part D drug benefits on potentially unsafe opioid use. Am J Public Health 2018;108:248–55.
[25]. Golub A, Bennett AS. Prescription opioid initiation, correlates, and consequences among a sample of OEF/OIF military personnel. Subst Use Misuse 2013;48:811–20.
[26]. Gourlay DL, Heit HA, Almahrezi A. Universal precautions in pain medicine: a rational approach to the treatment of chronic pain. Pain Med 2005;6:107–12.
[27]. Graubard BI, Korn EL. Regression analysis with clustered data. Stat Med 1994;13:509–22.
[28]. Janevic MR, McLaughlin SJ, Heapy AA, Thacker C, Piette JD. Racial and socioeconomic disparities in disabling chronic pain: findings from the health and retirement study. J Pain 2017;18:1459–67.
[29]. Joffe MM, Have TRT, Feldman H, Kimmel SE. Model selection, confounder control, and marginal structural models. Am Statistician 2004;58:272–9.
[30]. Jones CM. Heroin use and heroin use risk behaviors among nonmedical users of prescription opioid pain relievers—United States, 2002–2004 and 2008–2010. Drug Alcohol Depend 2013;132:95–100.
[31]. Jones CM, Logan J, Gladden RM, Bohm MK. Demographic and substance use trends among heroin users–United States, 2002–2013. MMWR Morb Mortal Wkly Rep 2015;64:719–25.
[32]. Jordan AE, Des Jarlais DC, Hagan H. Prescription opioid misuse and its relation to injection drug use and hepatitis C virus infection: protocol for a systematic review and meta-analysis. Syst Rev 2014;3:1–5.
[33]. Justice AC, Lasky E, McGinnis KA, Skanderson M, Conigliaro J, Fultz SL, Crothers K, Rabeneck L, Rodriguez-Barradas M, Weissman SB, Bryant K. Medical disease and alcohol use among veterans with human immunodeficiency infection: a comparison of disease measurement strategies. Med Care 2006;44:S52–60.
[34]. Justice AC, Dombrowski E, Conigliaro J, Fultz SL, Gibson D, Madenwald T, Goulet J, Simberkoff M, Butt AA, Rimland D, Rodriguez-Barradas MC, Gibert CL, Oursler KA, Brown S, Leaf DA, Goetz MB, Bryant K. Veterans aging cohort study (VACS): overview and description. Med Care 2006;44(8 suppl 2):S13–24.
[35]. Krebs EE, Ramsey DC, Miloshoff JM, Blair MJ. Primary care monitoring of long-term opioid therapy among veterans with chronic pain. Pain Med 2011;12:740–6.
[36]. Lipari RN, Hughes A. The NSDUH report: trends in heroin use in the United States: 2002 to 2013. Rockville: Center for Behavioral Health Statistics and Quality Data Review 2015, Substance Abuse and Mental Health Services Administration (SAMHSA), 2015.
[37]. Lovejoy TI, Dobscha SK, Turk DC, Weimer MB, Morasco BJ. Correlates of prescription opioid therapy in Veterans with chronic pain and history of substance use disorder. J Rehabil Res Dev 2016;53:25–36.
[38]. Mars SG, Bourgois P, Karandinos G, Montero F, Ciccarone D. “Every 'never' I ever said came true”: transitions from opioid pills to heroin injecting. Int J Drug Pol 2014;25:257–66.
[39]. Martell BA, O'Connor PG, Kerns RD, Becker WC, Morales KH, Krosten TR, Fiellin DA. Systematic review: opioid treatment for chronic back pain: prevalence, efficacy, and association with addiction. Ann Intern Med 2007;146:116–27.
[40]. McGinnis KA, Tate JP, Williams EC, Skanderson M, Bryant KJ, Gordon AJ, Kraemer KL, Maisto SA, Crystal S, Fiellin DA, Justice AC. Comparison of AUDIT-C collected via electronic medical record and self-administered research survey in HIV infected and uninfected patients. Drug Alcohol Depend 2016;168:196–202.
[41]. Michna E, Ross EL, Hynes WL, Nedeljkovic SS, Soumekh S, Janfaza D, Palombi D, Jamison RN. Predicting aberrant drug behavior in patients treated for chronic pain: importance of abuse history. J Pain Symptom Manage 2004;28:250–8.
[42]. Miller M, Barber CW, Leatherman S, Fonda J, Hermos JA, Cho K, Gagnon DR. Prescription opioid duration of action and the risk of unintentional overdose among patients receiving opioid therapy. JAMA Intern Med 2015;175:608–15.
[43]. Morasco BJ, Duckart JP, Carr TP, Deyo RA, Dobscha SK. Clinical characteristics of veterans prescribed high doses of opioid medications for chronic non-cancer pain. PAIN 2010;151:625–32.
[44]. Neaigus A, Gyarmathy VA, Miller M, Frajzyngier VM, Friedman SR, Des Jarlais DC. Transitions to injecting drug use among noninjecting heroin users. J Acquir Immune Defic Syndr 2006;41:493–503.
[45]. Nuckols TK, Anderson L, Popescu I, Diamant AL, Doyle B, Di Capua P, Chou R, Opioid prescribing: a systematic review & critical appraisal of guidelines for chronic pain. Ann Intern Med, 2014;160:38–50.
[46]. Oliva EM, Nevedal A, Lewis ET, McCaa MD, Cochran MF, Konicki PE, Davis CS, Wilder C. Patient perspectives on an opioid overdose education and naloxone distribution program in the U.S. Department of Veterans Affairs. Subst Abuse 2016;37:118–26.
[47]. Pollini RA, Banta-Green CJ, Cuevas-Mota J, Metzner M, Teshale E, Garfein RS. Problematic use of prescription-type opioids prior to heroin use among young heroin injectors. Subst Abuse Rehabil 2011;2:173–80.
[48]. Robins LN, Helzer JE, Hesselbrock M, Wish E. Vietnam veterans three years after Vietnam: how our study changed our view of heroin. Am J Addict 2010;19:203–11.
[49]. Robins LN, Slobdyan S. Post-Vietnam heroin use and injection by returning US veterans: clues to preventing injection today. Addiction 2003;98:1053–60.
[50]. Saunders KW, Dunn KM, Merrill JO, Sullivan M, Weisner C, Branden JB, Psaty BM, Von Korff M. Relationship of opioid use and dosage levels to fractures in older chronic pain patients. J Gen Intern Med 2010;25:310–15.
[51]. Sullivan M. Half of veterans prescribed medical opioids continue to use them chronically. 2014. American Academy of Pain Medicine (AAPM). Available at: Accessed February 8, 2016.
[52]. Sullivan MD, Edlund MJ, Zhang L. Association between mental health disorders, problem drug use, and regular prescription opioid use. JAMA Intern Med 2006;166:2087–93.
[53]. Unick GJ, Rosenblum D, Mars S, Ciccarone D. Intertwined epidemics: national demographic trends in hospitalizations for heroin- and opioid-related overdoses, 1993–2009. PLoS One 2013;8:e54496.
[54]. U.S. Drug Enforcement Administration. Controlled substance schedules. 2010. Available at: Accessed July 5, 2018.
[55]. US Department of Veterans Affairs/Department of Defense. Clinical practice guidelines for management of opioid therapy for chronic pain. 2017. Available at:
[56]. VA/DoD clinical practice guideline: management of opioid therapy for chronic pain. Department of Veterans Affairs, Department of Defense. 2010. Available at: Accessed August 27, 2018.
[57]. Volkow ND, McLellan TA. Curtailing diversion and abuse of opioid analgesics without jeopardizing pain treatment. JAMA 2011;305:1346–7.
    [58]. Weisberg DF, Gordon KS, Barry DT, Becker WC, Crystal S, Edelman EJ, Gaither JR, Gordon AJ, Goulet JL, Kerns RD, Moore BA, Tate J, Justice AC, Fiellin DA. Long-term prescription of opioids and/or benzodiazepines and mortality among HIV-infected and uninfected patients. J Acquir Immune Defic Syndr 2015;69:223–33.
    [59]. Westanmo A, Marshall P, Jones E, Burns K, Krebs EE. Opioid dose reduction in VA health care system—implementation of primary care population-level initiative. Pain Med 2015;16:1019–26.

    Heroin; Opioids; Veterans; Longitudinal study

    Supplemental Digital Content

    Back to Top | Article Outline
    © 2019 International Association for the Study of Pain