In the United States, prescribing and coprescribing of opioid analgesics and benzodiazepines is increasing.1–7 Although some patients benefit from long-term (≥90 days) opioid therapy (eg, for chronic pain), many do not; furthermore, side effects are common, and serious adverse events are associated with higher doses.8,9 Similarly, the role of long-term benzodiazepines in treating anxiety10–13 and chronic insomnia14 is limited given the relative safety and efficacy of alternative treatments.15 Trends in increased prescription of opioid and benzodiazepine medications, individually and in combination, warrant assessment of the associated risks, especially in vulnerable populations such as HIV-infected individuals. Although analyses of accidental overdose deaths in the United States have raised concern over concurrent opioid and benzodiazepine use, overdose may not be the exclusive driver of increased mortality associated with opioid or benzodiazepine use. For example, in observational studies, opioids and/or benzodiazepines have been associated with increased risk of admission to hospital because of falls, automobile accidents, and developing more frequent and severe pneumonia.16–22 Moreover, the mortality risk of combined long-term opioid and benzodiazepine receipt has not been previously evaluated.
Long-term exposure to opioids and/or benzodiazepines may elicit safety concerns because of the populations to which they are prescribed and the individual properties of these medications. Opioids and benzodiazepines are often prescribed to individuals with mental health and substance use disorders.23,24 In addition, the sedative properties and addiction potential of these medications may increase in combination.25,26 HIV-infected patients may be at an increased risk for unsafe use of these medications because of their higher prevalence of polypharmacy, diminished capacity for drug metabolism and elimination,27 and higher rates of mental health and substance use disorders.28,29
We have previously demonstrated an association between polypharmacy and mortality among HIV-infected and uninfected patients.30 The harms of long-term opioid and/or benzodiazepine receipt beyond increasing medication count are unknown. This study was designed to quantify and compare the mortality risk associated with long-term opioid and/or benzodiazepine receipt among HIV-infected patients on antiretroviral therapy (ART) and uninfected patients. Our analysis both considers the impact of opioid dose on mortality and uses propensity score matching to address confounding by indication.31
We extracted data from the Veterans Aging Cohort Study-Virtual Cohort (VACS-VC) for fiscal year (FY) 2009 (October 1, 2008 through September 30, 2009). This period was chosen to ensure that we examined current prescribing practices and had sufficient follow-up time to test for an association with mortality. The VACS-VC has been described in detail elsewhere.32–35 Briefly, the VACS-VC is a prospective observational cohort consisting of HIV-infected patients matched 1:2 by age, sex, race/ethnicity, and site of care to uninfected patients identified from the US Veterans Health Administration (VHA) administrative database. The data compiled from this cohort originate from the Immunology Case Registry, a registry of HIV-infected patients, the VHA paperless electronic medical record, and the decision support system.36,37 Before analyses, the data are extensively cleaned and validated after established protocols. The Institutional Review Boards at Yale University and the VA Connecticut Healthcare System approved the conduct of the analyses described herein.
We excluded individuals who (1) did not have at least 1 inpatient or outpatient clinical encounter in 2009, (2) did not have active pharmacy data, (3) had an ambiguous HIV status, and (4) had any cancer diagnosis other than minor skin cancers (nonepithelial). In addition, HIV-infected patients not on ART were excluded because of their small number and the profound mortality benefit of ART such that including patients not on ART would introduce excessive heterogeneity of mortality risk to the sample.
Opioid and Benzodiazepine Receipt
Opioid receipt was determined by pharmacy data indicating all outpatient oral and transdermal opioids as previously described.38 Medications prescribed for the treatment of opioid dependence (methadone and buprenorphine) were excluded. Long-term opioid receipt, based on prescription information and the assumption that medications were taken as directed, was defined as ≥90 consecutive days of opioid therapy, allowing for a 30-day refill window.39 Patients meeting these criteria at any point during FY 2009 were considered to have long-term opioid receipt. Average morphine-equivalent daily dosages (MEDD) were estimated using standard conversion factors described in detail elsewhere.38,39 Benzodiazepine receipt was determined by pharmacy data indicating receipt of the following medications: alprazolam, chlordiazepoxide, clonazepam, clorazepam, diazepam, estazolam, flurazepam, lorazepam, midazolam, oxazepam, temazepam, and triazolam. Long-term benzodiazepine receipt was defined using the same criteria as for opioids. Long-term opioid receipt and benzodiazepine receipt were defined as ≥90 days of receipt of both medications at any point during 2009 whether or not the time frames overlapped.
Data for all-cause mortality were obtained from the VHA vital status file, which includes data from the VHA through the Beneficiary Identification Records Locator Subsystem, and social security and Medicare data. The reliability and validity of the vital status file has been established with the National Death Registry.40,41 We conducted surveillance for death in FY 2010, and time to death was calculated from the start of FY 2010.
Sociodemographic variables, including age, gender, and race/ethnicity were collected from hospital administrative data. Clinical variables, which included serious mental illness (major depression, bipolar disorder, posttraumatic stress disorder, and schizophrenia), substance use disorders (drug and alcohol abuse or dependence disorders), smoking, acute and chronic pain diagnoses,38 were based on International Classification of Diseases, Ninth Revision (ICD-9) (see appendix) diagnostic and procedure codes. Respiratory and renal diagnoses, as well as factors such as hospitalization, which are thought to be associated with increased risk for benzodiazepine-related mortality,15 were identified. Hepatitis C status was based on ICD-9 codes and laboratory data.
Average long-term outpatient medication count (all medications excluding opioids, benzodiazepines, and ART) was calculated based on pharmacy data. ART was excluded from medication count because of its known mortality benefit and to facilitate appropriate comparisons between HIV-infected and uninfected patients. For HIV-infected patients, CD4 count and viral load averaged over FY2009 were collected from laboratory data. A score greater than or equal to 4 on the Alcohol Use Disorders Identification Test-Consumption,42 collected as part of routine clinical care in the VHA, was used to define unhealthy alcohol use.43
We used the VACS index as a measure of overall severity of illness. The VACS index is a validated prognostic measure that includes age, CD4 count, viral load, hemoglobin, FIB-4, eGFR, and hepatitis C infection and has been shown to predict mortality in patients with HIV on ART.44 The VACS index predicts 30-day mortality after medical intensive care unit admission in HIV-infected and uninfected patients45 and is correlated with functional performance,46 fragility fractures,47 and markers of inflammation (IL-6, D-Dimer, soluble CD14)48 in HIV-infected patients. We assumed “normal” CD4 count (>500 cells/mm3) and an undetectable viral load (<20 copies/mL) in calculating the VACS index for uninfected patients.
Descriptive statistics were performed and stratified by HIV status. We used t-tests for continuous variables, or a nonparametric counterpart for non-normally distributed variables, and χ2 for categorical variables, considering P < 0.05 to be statistically significant.
Bivariate analyses and Cochran–Mantel–Haenszel statistics established associations with the outcome of all-cause mortality. We tested for an interaction between long-term opioid receipt and long-term benzodiazepine receipt and their association with death, as well as an interaction between HIV status and long-term opioid and/or benzodiazepine receipt and their association with death.
Long-term opioid and/or benzodiazepine exposure was represented by 1 variable with 4 mutually exclusive levels: long-term opioid receipt, long-term benzodiazepine receipt, both long-term opioid and benzodiazepine receipt, and neither. Risk of death was determined by multivariable Cox proportional hazards regression adjusting for race/ethnicity, major depression, bipolar disorder, posttraumatic stress disorder, schizophrenia, drug use disorder, alcohol use disorder, smoking status, medication count, acute pain and chronic pain diagnoses, and VACS index score.
Propensity Score Analysis
Observational studies, especially those focusing on medication receipt and mortality, are susceptible to confounding by indication, whereby patients with higher mortality risks are more likely to receive certain medications.49 We conducted propensity score analyses to address this concern.50,51 A propensity score was generated based on patients' probability to receive long-term opioids and/or benzodiazepines using logistic regression, which included demographics, HIV and Hepatitis C status, smoking, pain diagnoses, among 43 medical and psychiatric conditions. Variables were selected based on bivariate analyses and the consensus of the research team because those variables were considered to have potential associations with exposure and/or outcome.52 We evaluated model discrimination with c-statistics and model estimates. We used the model based on exposure to long-term opioids and/or benzodiazepines because results of separate logistic regressions for receipt of long-term opioid, benzodiazepine, opioid or benzodiazepine, and both opioid and benzodiazepine did not differ meaningfully. Exposed and unexposed individuals were 1:1 matched by propensity score using a greedy algorithm.53 The matched sample was tested for balance based on clinical and demographic variables.
Hazard ratios (HRs) for long-term opioid receipt, long-term benzodiazepine receipt, and long-term opioid and benzodiazepine receipt were generated using unadjusted Cox regression. In sensitivity analyses, separate propensity score and matching models were evaluated by HIV status. Different combinations of variables in the logistic regressions that generated the propensity scores for the HIV-infected and uninfected samples did not improve model discrimination, and the final propensity score models used the same variables as for the overall model. Additional sensitivity analyses were performed wherein we restricted ART exposure to ≥3, ≥6, and ≥12 months.
To assess the impact of opioid dose on mortality risk, we subcategorized opioid receipt into mutually exclusive dose categories of <20, 20 to <50, 50 to <100, and ≥100 mg MEDD, based on previous analyses evaluating opioid overdose,54,55 in the overall, HIV-infected, and uninfected propensity-matched samples. All statistical analyses were performed using SAS version 9.2 (SAS Institute Inc, Cary, NC).
Our sample consisted of 64,602 individuals (16,989 HIV-infected and 47,613 uninfected). HIV-infected patients were more likely to be male, younger, and white compared with uninfected patients (Table 1). Long-term opioid receipt, benzodiazepine receipt, and long-term opioid and benzodiazepine receipt were lower in HIV-infected individuals compared with uninfected individuals. Average days with long-term opioid or benzodiazepine receipt did not differ by HIV status.
Long-term Opioid and/or Benzodiazepine Receipt and Mortality
There were 1570 deaths: 539 (3.2%) among HIV-infected patients and 1031 (2.2%) among uninfected patients, P < 0.001. Among those who died, the median time to death was 180.1 days (25th percentile, 92.0 days; 75th percentile, 265 days).
Table 2 demonstrates the results of unadjusted and adjusted Cox proportional hazards models in the overall sample and stratified by HIV status.
The adjusted analyses show incremental increasing harm for long-term opioid receipt [HR 1.39, 95% confidence interval (CI: 1.21 to 1.60)], long-term benzodiazepine receipt (HR 1.33, 95% CI: 1.10 to 1.62), and long-term opioid and benzodiazepine receipt (HR 1.51, 95% CI: 1.22 to 1.87). The interaction between HIV status and long-term opioid receipt regarding mortality was of borderline significance (P = 0.06), and other tested interactions were not statistically significant. The risk of death was higher among HIV-infected patients with long-term opioid receipt (1.54, 95% CI: 1.21 to 1.96) compared with uninfected patients (HR 1.35, 95% CI: 1.14 to 1.61). Among HIV-infected patients, the adjusted risk of death for long-term benzodiazepine receipt was not statistically significant (P = 0.45).
In addition, the VACS index and several other factors were associated with mortality. Black and Hispanic individuals had lower mortality compared with white individuals. In the overall sample, each additional long-term medication count (excluding ART, opioid, benzodiazepine medications) was associated with a 5% increased relative risk of mortality (HR 1.05, 95% CI: 1.04 to 1.07). Alcohol use disorder was associated with an increased risk of mortality (HR 1.63, 95% CI: 1.39 to 1.90).
Patient Characteristics in Propensity-Matched Sample
Logistic regression had good discrimination (c-statistic = 0.77). From 64,602 eligible individuals, with 15,911 long-term opioid and/or benzodiazepine recipients, 13,564 pairs were included in the matched sample. Table 3 shows the baseline characteristics of the matched sample. The sample was well balanced on variables included in the logistic regression that generated the propensity score (eg, serious mental illnesses). VACS index score was higher among patients who had not received opioids and/or benzodiazepines.
Separate propensity score models for HIV-infected and uninfected patients also showed good fit (c-statistic = 0.77 for HIV-infected and 0.76 for uninfected), and the matched samples likewise showed good balance on clinical and demographic variables.
Long-term Opioid and/or Benzodiazepine Receipt and Mortality in Propensity-Matched Patients
Risk of death associated with long-term opioid receipt, long-term benzodiazepine receipt, and long-term opioid and benzodiazepine receipt in the propensity-matched sample yielded similar trends, with attenuated effect sizes, compared with the unmatched analysis (Table 4). Long-term opioid receipt alone and long-term benzodiazepine receipt alone were associated with an increased risk for mortality (HR 1.40, 95% CI: 1.22 to 1.61 for long-term opioid receipt; HR 1.26, 95% CI: 1.08 to 1.48, for long-term benzodiazepine receipt); patients with both long-term opioid and benzodiazepine receipt had an increased risk for mortality above opioid or benzodiazepine receipt alone (HR 1.56, 95% CI: 1.26 to 1.92).
The interaction between HIV status and long-term opioid receipt was significant (P = 0.01), whereas other tested interactions were not statistically significant. Separate analyses for propensity-matched samples of HIV-infected patients (n = 6076) and uninfected patients (n = 20,980) are presented alongside the overall matched sample in Table 4. Long-term opioid receipt was associated with higher risk of mortality among HIV-infected patients compared with uninfected patients, HR 1.46 (95% CI: 1.15 to 1.87) and HR 1.25 (95% CI: 1.05 to 1.49), respectively.
Sensitivity analyses that restricted ART exposure to ≥3, ≥6, and ≥12 months yielded results that were similar to any ART exposure.
Opioid Dose, Benzodiazepine Receipt, and Mortality in Propensity-Matched Patients
Unadjusted Cox proportional hazards models in propensity-matched patients showed a dose-dependent association with death (Fig. 1). Among patients receiving long-term opioids, those who received doses <50 mg MEDD did not have an increased risk of death, whereas patients who received doses from 50 mg to <100 mg of morphine/d and doses ≥100 mg MEDD were at significantly greater risk of death, with HR of 2.00 (95% CI: 1.53 to 2.62) and 2.34 (95% CI: 1.91 to 2.86), respectively. Long-term opioid and benzodiazepine receipt was associated with increased risk of death for opioid dosages ≥20 mg MEDD in the overall sample (P = 0.03) and the HIV-infected sample (P = 0.06).
Using propensity score matching to control for confounding by indication, long-term opioid receipt was associated with a 40% increased risk of all-cause mortality; this risk was increased among those with long-term benzodiazepine receipt and among HIV-infected patients. In addition, long-term benzodiazepine receipt was associated with increased risk of death — 26% increased risk — regardless of long-term opioid receipt. These trends were similar to the results from the multivariable adjusted Cox regression but attenuated potentially because of confounding by indication that was unaccounted for in the multivariable model. When long-term opioid receipt was subcategorized into dose categories, there was a dose-dependent association with mortality, and a threshold of 50 mg MEDD was associated with increased risk of death. Among patients with receipt of long-term opioid and benzodiazepine receipt, the threshold for increased risk of death was 20 mg MEDD.
This study is the first to establish the risk for all-cause mortality among HIV-infected and uninfected patients who have received long-term opioids and/or benzodiazepines while controlling for confounding by indication. In addition, this study adds to the literature in establishing a risk of mortality for long-term benzodiazepine receipt alone and in the setting of long-term opioid receipt.13
The use of long-term opioids for chronic pain is the subject of scientific debate and public scrutiny because of limited evidence for their long-term efficacy and safety,8,56 their potential for abuse,57 their contribution to addiction,58,59 and increasing trends in fatal and nonfatal prescription drug overdose.54,55,60,61 Of the many potential reasons for our findings of increased mortality in individuals with long-term opioid and benzodiazepine receipt, one is the risk of overdose conferred by these medications. Opioids and benzodiazepines have independent and synergistic effects that can lead to overdose.25,62 Studies have shown that receipt of greater than 50 mg MEDD is associated with an increased risk of overdose death.54,55,61 However, these studies did not use propensity score matching, or consider benzodiazepine receipt or HIV status.
In comparison with opioids, benzodiazepines, used alone, have a wider therapeutic index. However, in combination with opioids, benzodiazepines are more prone to pharmacodynamic and pharmacokinetic interactions that can lead to respiratory depression. Alprazolam, midazolam, triazolam, codeine, fentanyl, hydrocodone, oxycodone, and methadone all undergo metabolism by CYP3A4.62 Of particular relevance to HIV-infected individuals, protease inhibitors, some macrolides, and some azoles may also interact with both benzodiazepines and opioids through CYP3A4 and CYP2D6 inhibition leading to increased blood levels of opioids and benzodiazepines among patients coprescribed these medications.63,64 These pharmacokinetic interactions may contribute to our finding of increased risk of death among HIV-infected patients who received long-term opioids and benzodiazepines compared with uninfected patients, and the lower opioid dose threshold associated with mortality among HIV-infected patients. HIV-infected patients have higher rates of overdose than uninfected patients.65
Overdose is not the only potential cause of mortality for patients who have received long-term opioids and/or benzodiazepines. Opioids are known to cause cardiac, endocrine, gastrointestinal, and central nervous system disturbances, which may precipitate hospitalizations and/or interact with other disease processes.66 In addition, falls, fractures, and motor vehicle accidents are associated with opioid and/or benzodiazepine receipt.13,67,68 Patients receiving long-term opioids are more likely to have alcohol and drug use disorders,69 and patients receiving opioids and benzodiazepines are likely to receive multiple additional medications.1,70 We observed increased risk associated with alcohol use disorders and polypharmacy. Polypharmacy is associated with nonadherence, adverse drug reactions, drug–drug interactions, diminished activities of daily living, increased health service utilization, and cognitive impairments and falls, even after controlling for disease burden.27,71–73
Our results also showed that black race and Hispanic ethnicity were negatively associated with all-cause mortality in multivariable Cox regression. Other studies have found that blacks and Hispanics are less likely to receive opioids and benzodiazepines compared with whites,5,38,74 and that black race is inversely associated with opioid overdose.55 In contrast, among HIV-infected patients, nonwhite race has been shown to be associated with increased risk of mortality.75 Future research should investigate whether protective factors independent of and related to opioid and benzodiazepine receipt mitigate harm among black and Hispanic patients.
Our study has limitations. First, although propensity score matching offers the potential to control for confounding by indication, it can account for only known and observed patient characteristics.50 It is possible that unknown or unmeasured characteristics (eg, pain intensity, degree of disability) affected our results. In addition, the use of administrative codes may have limited our ability to identify conditions important to long-term opioid and/or benzodiazepine receipt.76 Our analysis accounted only for long-term opioid and/or benzodiazepine receipt during 2009. As is standard in large-scale epidemiologic studies, our exposure variables were based on prescription fills, not medication consumed. Patients may not have still been receiving the long-term prescriptions at the time of death. Previous analyses of opioid overdose have demonstrated that 40% of overdose victims did not have active opioid prescriptions at the time of death, highlighting the challenges in ascribing a causal relationship to prescription and overdose.55
Of note, surveillance for death occurred in FY 2010, and patients who died in FY 2009 were excluded, creating the potential for biasing our results toward the null hypothesis.77 In addition, we studied only VHA pharmacy usage; patients may have accessed other licit or illicit medication sources. The use of opioids and/or benzodiazepines obtained illicitly can increase the risk for overdose.60 Finally, our results may not be generalizable to women and men who do not receive medical care in the VHA system.
In recent years, the VHA has sought to restrain coprescription of opioids and benzodiazepines as part of its Opioid Safety Initiative. In light of increasing prescriptions and limited efficacy of long-term opioids and benzodiazepines, our results support the imperative within the VHA and nationally to mitigate the risk associated with receipt of long-term opioids and/or benzodiazepines, and use caution in coprescribing, especially among HIV-infected individuals, who are at increased overall risk for death. Our study adds to a nascent understanding of the overlapping harms associated with psychoactive substance coprescribing and should help inform interventions and research seeking to balance the risks and benefits in patients who are prescribed long-term opioids and benzodiazepines.
1. Mojtabai R, Olfson M. National trends in psychotropic medication polypharmacy
in office-based psychiatry. Arch Gen Psychiatry. 2010;67:26–36.
2. Olsen Y, Daumit GL, Ford DE. Opioid prescriptions by U.S. primary care physicians from 1992 to 2001. J Pain. 2006;7:225–235.
3. Reid MC, Engles-Horton LL, Weber MB, et al.. Use of opioid medications for chronic noncancer pain syndromes in primary care. J Gen Intern Med. 2002;17:173–179.
4. Braden JB, Fan MY, Edlund MJ, et al.. Trends in use of opioids by noncancer pain type 2000-2005 among Arkansas Medicaid and HealthCare enrollees: results from the TROUP study. J Pain. 2008;9:1026–1035.
5. Pletcher MJ, Kertesz SG, Kohn MA, et al.. Trends in opioid prescribing by race/ethnicity for patients seeking care in US emergency departments. JAMA. 2008;299:70–78.
6. Becker WC, Gordon K, Edelman EJ, Kerns R, Crystal S, Dziura J, Fiellin L, Gordon A, Goulet J, Justice A, Fiellin D. Trends in opioid analgesic receipt among patients with and without HIV. Society of General Internal Medicine National Meeting, April 2013, Denver CO, April 2013.
7. Hawkins EJ, Malte CA, Imel ZE, et al.. Prevalence and trends of benzodiazepine use among Veterans Affairs patients with posttraumatic stress disorder, 2003-2010. Drug Alcohol Depend. 2012;124:154–161.
8. Noble M, Treadwell JR, Tregear SJ, et al.. Long-term opioid management for chronic noncancer pain. Cochrane Database Syst Rev. 2010;1:CD006605.
9. Okie S. A flood of opioids, a rising tide of deaths. N Engl J Med. 2010;363:1981–1985.
10. Uhlenhuth EH, Balter MB, Ban TA, et al.. Trends in recommendations for the pharmacotherapy of anxiety disorders by an international expert panel, 1992-1997. Eur Neuropsychopharmacol. 1999;9(suppl 6):S393–S398.
11. Reinhold JA, Mandos LA, Rickels K, et al.. Pharmacological treatment of generalized anxiety disorder. Expert Opin Pharmacother. 2011;12:2457–2467.
12. Offidani E, Guidi J, Tomba E, et al.. Efficacy and tolerability of benzodiazepines versus antidepressants in anxiety disorders: a systematic review and meta-analysis. Psychother and Psychosom. 2013;82:355–362.
13. Charlson F, Degenhardt L, McLaren J, et al.. A systematic review of research examining benzodiazepine-related mortality. Pharmacoepidemiol Drug Saf. 2009;18:93–103.
14. Buscemi N, Vandermeer B, Friesen C, et al.. The efficacy and safety of drug treatments for chronic insomnia in adults: a meta-analysis of RCTs. J Gen Intern Med. 2007;22:1335–1350.
15. Amarasuriya UK, Myles PR, Sanders RD. Long-term benzodiazepine use and mortality: are we doing the right studies? Curr Drug Saf. 2012;7:367–371.
16. Khademi H, Malekzadeh R, Pourshams A, et al.. Opium use and mortality in Golestan Cohort Study: prospective cohort study of 50,000 adults in Iran. BMJ. 2012;344:e2502.
17. Solomon DH, Rassen JA, Glynn RJ, et al.. The comparative safety of analgesics in older adults with arthritis. Arch Intern Med. 2010;170:1968–1976.
18. Gomes T, Juurlink DN, Dhalla IA, et al.. Trends in opioid use and dosing among socio-economically disadvantaged patients. Open Med. 2011;5:e13–22.
19. Huang AR, Mallet L, Rochefort CM, et al.. Medication-related falls in the elderly: causative factors and preventive strategies. Drugs Aging. 2012;29:359–376.
20. Solomon DH, Rassen JA, Glynn RJ, et al.. The comparative safety of opioids for nonmalignant pain in older adults. Arch Intern Med. 2010;170:1979–1986.
21. Dublin S, Walker RL, Jackson ML, et al.. Use of opioids or benzodiazepines and risk of pneumonia in older adults: a population-based case-control study. J Am Geriatr Soc. 2011;59:1899–1907.
22. Li L, Setoguchi S, Cabral H, et al.. Opioid use for noncancer pain and risk of myocardial infarction amongst adults. J Intern Med. 2013;273:511–526.
23. Wasan AD, Butler SF, Budman SH, et al.. Psychiatric history and psychologic adjustment as risk factors for aberrant drug-related behavior among patients with chronic pain. Clin J Pain. 2007;23:307–315.
24. Sullivan MD, Edlund MJ, Zhang L, et al.. Association between mental health disorders, problem drug use, and regular prescription opioid use. Arch Intern Med. 2006;166:2087–2093.
25. Jones JD, Mogali S, Comer SD. Polydrug abuse: a review of opioid and benzodiazepine combination use. Drug Alcohol Depend. 2012;125:8–18.
26. Vogel M, Knopfli B, Schmid O, et al.. Treatment or “high”: benzodiazepine use in patients on injectable heroin or oral opioids. Addict Behav. 2013;38:2477–2484.
27. Edelman EJ, Gordon KS, Glover J, et al.. The next therapeutic challenge in HIV: polypharmacy
. Drugs Aging. 2013;30:613–628.
28. Bing EG, Burnam MA, Longshore D, et al.. Psychiatric disorders and drug use among human immunodeficiency virus-infected adults in the United States. Arch Gen Psychiatry. 2001;58:721–728.
29. Hansen L, Penko J, Guzman D, et al.. Aberrant behaviors with prescription opioids and problem drug use history in a community-based cohort of HIV-infected individuals. J Pain Symptom Manage 2011;42:893–902.
30. Edelman JE, Gordon K, Akgun K, Gibert C, Lo Re III V, McNicholl I, Rimland D, Skanderson M, Tate J, Womack J, Wyatt C, Justice AC. HIV+ individuals on ART are at risk of polypharmacy
: more medication increases mortality. ID Week 2013; October 2-6, 2013; San Francisco, CA; Abstract 76. https://idsa.confex.com/idsa/2013/webprogram/Paper42415.html
. Accessed April 1, 2015.
31. Salas M, Hofman A, Stricker BH. Confounding by indication: an example of variation in the use of epidemiologic terminology. Am J Epidemiol. 1999;149:981–983.
32. Tetrault JM, Tate JP, McGinnis KA, et al.. Hepatic safety and antiretroviral effectiveness in HIV-infected patients receiving naltrexone. Alcohol Clin Exp Res. 2012;36:318–324.
33. Butt AA, Chang CC, Kuller L, et al.. Risk of heart failure with human immunodeficiency virus in the absence of prior diagnosis of coronary heart disease. Arch Intern Med. 2011;171:737–743.
34. Freiberg MS, Chang CC, Skanderson M, et al.. The risk of incident coronary heart disease among veterans with and without HIV and hepatitis C. Circ Cardiovasc Qual Outcomes. 2011;4:425–432.
35. Fultz SL, Skanderson M, Mole LA, et al.. Development and verification of a “virtual” cohort using the National VA Health Information System. Med Care. 2006;44:S25–S30.
36. Smith MW, Joseph GJ. Pharmacy data in the VA health care system. Med Care Res Rrev. 2003;60:92S–123S.
37. Centers for Medicare and Medicaid Services. ICD-9 Provider and diagnostic codes. 2011. Available at: https://http://www.cms.gov/Medicare/Coding/ICD9ProviderDiagnosticCodes/index.html
. Accessed July 20, 2013.
38. Edelman EJ, Gordon K, Becker WC, et al.. Receipt of opioid analgesics by HIV-infected and uninfected patients. J Gen Intern Med. 2013;28:82–90.
39. Korff MV, Saunders K, Thomas Ray G, et al.. De facto long-term opioid therapy for noncancer pain. Clin J Pain. 2008;24:521–527.
40. Cowper DC, Kubal JD, Maynard C, et al.. A primer and comparative review of major US mortality databases. Ann Epidemiol. 2002;12:462–468.
41. Fisher SG, Weber L, Goldberg J, et al.. Mortality ascertainment in the veteran population: alternatives to the National Death Index. Am J Epidemiol. 1995;141:242–250.
42. Bush K, Kivlahan DR, McDonell MB, et al.. The AUDIT alcohol consumption questions (AUDIT-C): an effective brief screening test for problem drinking. Ambulatory Care Quality Improvement Project (ACQUIP). Alcohol Use Disorders Identification Test. Arch Intern Med. 1998;158:1789–1795.
43. Seale JP, Boltri JM, Shellenberger S, et al.. Primary care validation of a single screening question for drinkers. J Stud Alcohol. 2006;67:778–784.
44. Justice AC, Modur SP, Tate JP, et al.. Predictive accuracy of the Veterans Aging Cohort Study index for mortality with HIV infection: a North American cross cohort analysis. J Acquir Immune Defic Syndr. 2013;62:149–163.
45. Akgun KM, Tate JP, Pisani M, et al.. Medical ICU admission diagnoses and outcomes in human immunodeficiency virus-infected and virus-uninfected veterans in the combination antiretroviral era. Crit Care Med. 2013;41:1458–1467.
46. Erlandson KM, Allshouse AA, Jankowski C. Prospective comparison of three functional assessments with the Veteran's Aging Cohort Study index in virologically suppressed HIV-infected adults. 2nd International Workshop on HIV and Aging, Baltimore, MD, October 2011.
47. Womack JA, Goulet JL, Gibert C, et al.. Physiologic frailty and fragility fracture in HIV-infected male veterans. Clin Infect Dis 2013;56:1498–1504.
48. Justice AC, Freiberg MS, Tracy R, et al.. Does an index composed of clinical data reflect effects of inflammation, coagulation, and monocyte activation on mortality among those aging with HIV? Clin infect dis. 2012;54:984–994.
49. Glynn RJ, Schneeweiss S, Sturmer T. Indications for propensity scores and review of their use in pharmacoepidemiology. Basic Clin Pharmacol Toxicol. 2006;98:253–259.
50. Freemantle N, Marston L, Walters K, et al.. Making inferences on treatment effects from real world data: propensity scores, confounding by indication, and other perils for the unwary in observational research. BMJ. 2013;347:f6409.
51. Joffe MM, Rosenbaum PR. Invited commentary: propensity scores. Am J Epidemiol. 1999;150:327–333.
52. Brookhart MA, Schneeweiss S, Rothman KJ, et al.. Variable selection for propensity score models. Am J Epidemiol. 2006;163:1149–1156.
53. Austin PC. The use of propensity score methods with survival or time-to-event outcomes: reporting measures of effect similar to those used in randomized experiments. Stat Med. 2014;33:1242–1258.
54. Dunn KM, Saunders KW, Rutter CM, et al.. Opioid prescriptions for chronic pain and overdose: a cohort study. Ann Intern Med. 2010;152:85–92.
55. Bohnert AS, Valenstein M, Bair MJ, et al.. Association between opioid prescribing patterns and opioid overdose-related deaths. JAMA. 2011;305:1315–1321.
56. Chou R, Ballantyne JC, Fanciullo GJ, et al.. Research gaps on use of opioids for chronic noncancer pain: findings from a review of the evidence for an American Pain Society and American Academy of Pain Medicine clinical practice guideline. J Pain. 2009;10:147–159.
57. Fishbain DA, Cole B, Lewis J, et al.. What percentage of chronic nonmalignant pain patients exposed to chronic opioid analgesic therapy develop abuse/addiction and/or aberrant drug-related behaviors? A structured evidence-based review. Pain Med. 2008;9:444–459.
58. Martell BA, O'Connor PG, Kerns RD, et al.. Systematic review: opioid treatment for chronic back pain: prevalence, efficacy, and association with addiction. Ann Intern Med. 2007;146:116–127.
59. Minozzi S, Amato L, Davoli M. Development of dependence following treatment with opioid analgesics for pain relief: a systematic review. Addiction. 2013;108:688–698.
60. Hall AJ, Logan JE, Toblin RL, et al.. Patterns of abuse among unintentional pharmaceutical overdose fatalities. JAMA. 2008;300:2613–2620.
61. Gomes T, Mamdani MM, Dhalla IA, et al.. Opioid dose and drug-related mortality in patients with nonmalignant pain. Arch Intern Med. 2011;171:686–691.
62. Jann M, Kennedy WK, Lopez G. Benzodiazepines: a major component in unintentional prescription drug overdoses with opioid analgesics. J Pharm Pract. 2014;27:5–16.
63. Gruber VA, McCance-Katz EF. Methadone, buprenorphine, and street drug interactions with antiretroviral medications. Curr HIV/AIDS Rep. 2010;7:152–160.
64. McCance-Katz EF, Sullivan LE, Nallani S. Drug interactions of clinical importance among the opioids, methadone and buprenorphine, and other frequently prescribed medications: a review. Am J Addict. 2010;19:4–16.
65. Green TC, McGowan SK, Yokell MA, et al.. HIV infection and risk of overdose: a systematic review and meta-analysis. AIDS. 2012;26:403–417.
66. Benyamin R, Trescot AM, Datta S, et al.. Opioid complications and side effects. Pain Physician. 2008;11:S105–S120.
67. Saunders KW, Dunn KM, Merrill JO, et al.. Relationship of opioid use and dosage levels to fractures in older chronic pain patients. J Gen Intern Med. 2010;25:310–315.
68. Woolcott JC, Richardson KJ, Wiens MO, et al.. Meta-analysis of the impact of 9 medication classes on falls in elderly persons. Arch Intern Med. 2009;169:1952–1960.
69. Saunders KW, Von Korff M, Campbell CI, et al.. Concurrent use of alcohol and sedatives among persons prescribed chronic opioid therapy: prevalence and risk factors. J Pain. 2012;13:266–275.
70. Parsells Kelly J, Cook SF, Kaufman DW, et al.. Prevalence and characteristics of opioid use in the US adult population. Pain. 2008;138:507–513.
71. Shah BM, Hajjar ER. Polypharmacy
, adverse drug reactions, and geriatric syndromes. Clin Geriatr Med. 2012;28:173–186.
72. Azar MM, Springer SA, Meyer JP, et al.. A systematic review of the impact of alcohol use disorders on HIV treatment outcomes, adherence to antiretroviral therapy and health care utilization. Drug Alcohol Depend. 2010;112:178–193.
73. Cargiulo T. Understanding the health impact of alcohol dependence. Am J Health-syst pharm. 2007;64:S5–S11.
74. Yang HW, Simoni-Wastila L, Zuckerman IH, et al.. Benzodiazepine use and expenditures for Medicare beneficiaries and the implications of Medicare Part D exclusions. Psychiatr Serv. 2008;59:384–391.
75. McGinnis KA, Fine MJ, Sharma RK, et al.. Understanding racial disparities in HIV using data from the veterans aging cohort 3-site study and VA administrative data. Am J Public Health. 2003;93:1728–1733.
76. Sinnott PL, Siroka AM, Shane AC, et al.. Identifying neck and back pain in administrative data: defining the right cohort. Spine (Phila Pa 1976) 2012;37:860–874.
77. Suissa S. Immortal time bias in observational studies of drug effects. Pharmacoepidemiol Drug Saf. 2007;16:241–249.