Currently Viewing:
The American Journal of Managed Care April 2009
Currently Reading
Does Opioid Therapy Affect Quality of Care for Diabetes Mellitus?
Adam J. Rose, MD, MSc; John A. Hermos, MD; Susan M. Frayne, MD, MPH; Leonard M. Pogach, MD, MBA; Dan R. Berlowitz, MD, MPH; and Donald R. Miller, ScD
Assessing Outcomes in Child Psychiatry
Joseph Gold, MD; Ralph J. Buonopane, PhD; Robert A. Caggiano, EdD; Mark Picciotto, PhD; Christine Vogeli, PhD; Natalie T. Kanner, BA; Rebecca L. Babcock, BS; Zhonghe Li, MA; Michael Jellinek, MD; Stacey J. Drubner, JD, MPH, MSW; Kenneth J. Sklar, EdD; and J. Michael Murphy, EdD
Moving From A to Z: Successful Implementation of a Statin Switch Program by a Large Physician Group
Fangyan Z. Sy, PharmD; Hae Mi Choe, PharmD; Diane M. Kennedy, BS; Connie J. Standiford, MD; Dawn M. Parsons, RPh, MBA; Keith D. Bruhnsen, MSW; James G. Stevenson, PharmD; and Steven J. Bernstein, MD
The Transfer of Uninsured Patients From Academic to Community Primary Care Settings
Sheldon M. Retchin, MD, MSPH; Sheryl L. Garland, MHA; and Emmanuel A. Anum, MBChB, MPH, PhD
Primary Care Strength Linked to Prevention Programs for Cardiovascular Disease
Jan van Lieshout, MD; Michel Wensing, PhD; Stephen M. Campbell, PhD; and Richard Grol, PhD
Pharmaceutical Company Influence on Nonsteroidal Anti-Inflammatory Drug Prescribing Behaviors
Aanand D. Naik, MD; Aaron L. Woofter, MD; Jessica M. Skinner, BA; and Neena S. Abraham, MD, MSc

Does Opioid Therapy Affect Quality of Care for Diabetes Mellitus?

Adam J. Rose, MD, MSc; John A. Hermos, MD; Susan M. Frayne, MD, MPH; Leonard M. Pogach, MD, MBA; Dan R. Berlowitz, MD, MPH; and Donald R. Miller, ScD

Within the Veterans Affairs system, diabetes performance measures were similar in patients who received chronic opioid therapy and in those who did not.

Objective: To examine whether veterans who received chronic opioid therapy had worse diabetes performance measures than patients who did not receive opioids.

Study Design: Retrospective cohort study.

Methods: We identified all patients with diabetes mellitus receiving care in US Department of Veterans Affairs facilities during 2004. Cases received at least 6 prescriptions for chronic opioids during 2004, while controls were randomly selected from among patients with diabetes who received no opioids. We compared process measures (glycosylated hemoglobin and low-density lipoprotein cholesterol levels tested and an eye examination performed) and outcome measures (glycosylated hemoglobin level ≤9.0% and low-density lipoprotein cholesterol level ≤130 mg/dL) between groups.

Results: Cases (n = 47,756) had slightly worse diabetes performance measures than controls (n = 220,912) after adjustment for covariates. For example, 86.4% of cases and 89.0% of controls had a glycosylated hemoglobin test during fiscal year 2004 (adjusted odds ratio, 0.69; P <.001). Among cases, receipt of higher-dose opioids was associated with additional decrement in diabetes performance measures, with a dose-response relationship.

Conclusions: Chronic opioid therapy among patients within the Veterans Affairs system is associated with slightly worse diabetes performance measures compared with patients who do not receive opioids. However, patients receiving higher dosages of opioids had additional decrements in diabetes performance measures; these patients may be appropriate targets for interventions to improve their care for pain and diabetes.

(Am J Manag Care. 2009;15(4):217-224)

Within the Veterans Affairs system, patients who received opioids for chronic pain had slightly worse diabetes performance measures than patients who did not receive opioids.


  • Comparisons included measurement of glycemic and lipemic control, achievement of moderate or better glycemic and lipemic control, and a yearly eye examination.
  • Among the group receiving opioids, the receipt of higher daily opioid doses predicted worse results for all 5 diabetes performance measures. A dose-response relationship was observed, lending additional credibility to this finding.


Understanding the role of clinical complexity as a determinant of quality of care is a major research goal.1 In previous studies,2-9 the effect of clinical complexity on quality of care has varied depending on the diagnoses, the clinician and patient, and the clinical environment. Seeking to harmonize these mixed results into a unifying theory, Piette and Kerr10 proposed that symptomatic conditions may have a greater effect on quality of care than asymptomatic conditions and that conditions with dissimilar management goals (“discordant conditions”) may have a greater effect than those with similar goals (“concordant conditions”).

By this reasoning, chronic pain could have a considerable adverse effect on quality of care for unrelated conditions. Pain is highly symptomatic, and pain management is discordant with the management of other conditions.11 While the use of opioids to treat chronic noncancer pain is increasingly accepted,12 opioid therapy may present additional challenges due to the potential for abuse, dependence, and diversion and due to conflicts over appropriate dosages.13-19 However, opioid therapy could also facilitate care for unrelated conditions. Patients receiving opioids may visit the clinic more often, allowing more opportunities for medical management.10 Adequate treatment of pain may improve the patient’s functional status and quality of life,12 allowing greater focus on self-care activities.

Diabetes mellitus, a common, costly, and highly morbid condition,20,21 is a good condition in which to examine this possibility. Adequate management of diabetes requires collaboration among clinicians and the patient within a system of care,22-27 and explicit guidelines and diabetes performance targets exist with which to examine the adequacy of diabetes care.28-30 Krein et al31 showed that among patients with diabetes, chronic pain is a barrier to the completion of self-care activities such as taking medications, exercising, and pursuing a prudent diet. However, the effect of pain on process and outcome measures of diabetes care is unknown. In addition, no study has specifically examined the effect of opioid therapy on the quality of care for unrelated chronic conditions, but there is reason to believe that opioid therapy may impart more complexity and challenge than pain alone.32

To clarify whether the net effect of opioid therapy is to promote or impede care for diabetes, we analyzed a large database of patients with diabetes in the US Department of Veterans Affairs (VA) and identified those receiving chronic opioid therapy. We compared patients receiving chronic opioids versus patients not receiving opioids regarding selected diabetes performance measures. We hypothesized that the distractions and concerns associated with chronic opioid therapy, as well as perhaps other characteristics of patients with chronic pain, would be reflected in worse diabetes performance measures. We also hypothesized that among those receiving opioids there would be a dose-response relationship between higher opioid dosages and decrements in diabetes performance measures.


Study Sample

We identified subjects from the Diabetes Epidemiology Cohort, which comprises all patients with diabetes seen in the VA.21 The Diabetes Epidemiology Cohort links administrative, laboratory, and pharmacy data from the VA with Medicare claims, providing a rich data set for analysis.21,33 We first looked at all veterans treated for diabetes during fiscal year (FY) 2004 whose diabetes had been diagnosed before the start of FY 2002. Based on earlier work,21 we defined patients as having diabetes if they had at least 2 International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) codes for diabetes or any prescriptions for antiglycemic medications within a 2-year period.

We excluded patients who had an ICD-9-CM diagnosis of malignant neoplasm (other than basal or squamous carcinoma of the skin) within 2 years of study inception. The management of cancer-related pain is qualitatively different; moreover, diabetes performance measures may not apply to patients with active malignant neoplasms. We also excluded all patients receiving methadone hydrochloride or buprenorphine hydrochloride–naloxone hydrochloride for treatment of opioid dependence. Finally, we excluded patients who had fewer than 2 VA primary care visits in FY 2004, as a large portion of their diabetes care may not appear in our database.

This study was approved by the Institutional Review Board of Bedford VA Medical Center.

Independent Variable: Chronic Opioid Therapy

Our independent variable was the prescription of chronic opioids. We considered the following “major” opioids: codeine, fentanyl citrate, hydrocodone, hydromorphone hydrochloride, methadone, morphine sulfate, and oxycodone; all are Schedule II or III controlled substances according to the US Drug Enforcement Administration.34 Any formulation suitable for outpatient administration was considered, including tablets, patches, elixirs, and sprinkles. We also included formulations that combine opioids with other drugs such as acetaminophen. Buprenorphine, butorphanol, nalbuphine hydrochloride, pentazocine, and propoxyphene, which are less potent, were considered “minor” opioids.34

Patients who received at least 6 prescriptions for major opioids during FY 2004, with or without additional minor opioids, constituted the chronic opioid group (cases). This cutoff of 6 prescriptions was chosen to distinguish treatment for chronic pain from treatment for acute pain and is consistent with previous definitions of chronic pain.17,18 Patients who received any major or minor opioids during FY 2004 but did not meet criteria for the case group were excluded from the study. We randomly selected controls from among the remaining patients, who had received no opioids during FY 2004, to achieve a control group approximately 4 times as numerous as the case group.

Dependent Variables: Diabetes Performance Measures

Our 3 process measures, which could be completed at any time during FY 2004, were testing of glycosylated hemoglobin (A1C) level, testing of low-density lipoprotein cholesterol (LDL-C) level, and a dilated eye examination. Our 2 outcome measures were at least 1 A1C level of 9.0% or less and at least 1 LDL-C level of 130 mg/dL or less during FY 2004 (to convert A1C level to proportion of total hemoglobin, multiply by 0.01; to convert cholesterol level to millimoles per liter, multiply by 0.0259). If no test results were available among VA data, patients were considered to have levels above these thresholds. These diabetes performance measures are based on VA clinical practice guidelines for diabetes and reflect a minimal standard of care.28,29 We also examined lower targets for glycemic and lipemic control (ie, A1C level ≤8.0% and LDL-C level ≤100 mg/dL).


Age was divided into the following 4 categories: 54 years or younger, 55 to 64 years, 65 to 74 years, and 75 years or older. Race/ethnicity was categorized into the following 4 groups: white non-Hispanic, black non-Hispanic, all others, and missing. The VA priority status, which characterizes the degree of entitlement to VA care, was defined as follows: poverty, full disability, partial disability, or none of the above.

More or less intensive management of diabetes may be indicated depending on life expectancy and comorbidities.29 We focused on the following complications of diabetes by identifying conditions with at least 1 ICD-9-CM code during FYs 1997 through 2004: cellulitis, gangrene/ulcer, other diabetic infections, congestive heart failure, other heart diseases, cerebrovascular disease, peripheral vascular disease, renal disease, usaffect diabetes care.2-5 Using similar ICD-9-CM code–based definitions, we identified the following mental health conditions: major depression, bipolar disorder, anxiety disorders, posttraumatic stress disorder, and schizophrenia. We also recorded the number of VA primary care visits; more visits might allow more opportunities to complete diabetes performance measures. We also examined pain diagnoses, dividing them into the following 4 broad categories: neuropathic pain, musculoskeletal pain, chronic headache, and psychogenic pain. Using ICD-9-CM codes, we categorized patients according to whether or not they had any diagnoses in each category (vs none).

We hypothesized that patients receiving higher daily doses of opioids might be at risk for additional decrements in diabetes performance measures, as the receipt of higher dosages suggests difficulties in pain management and possibly physiologic tolerance and an increased risk of dependence.13,17,35 We used a standard equivalency table36 to convert all opioid dosages to oral morphine equivalents. We calculated a mean daily dose of opioid therapy in FY 2004 for each patient in the study and categorized patients into quartiles based on their daily opioid doses.

Finally, we assigned each patient to 1 VA medical center so that we could control for site of care. Our assignment was based on the site the patient visited most often for diabetes care during FY 2004. If 2 sites were visited equally, we selected the site visited closest to the end of the year.

Statistical Analysis

We began our analysis with bivariate comparisons of demographics, comorbidities, and healthcare utilization between cases and controls. Using X2 tests, we then performed unadjusted comparisons of the proportions fulfilling each of the 5 diabetes performance measures. We performed adjusted analyses using generalized estimating equations to account for the clustering of outcomes by site of care, while adjusting for other covariates (sex, age, race/ethnicity, VA priority status, pain diagnoses, diabetic complications [including neuropathic pain], mental health and diabetic eye disease. Mental health conditions may also conditions, and the number of VA primary care visits during the study). We did not adjust for eye disease when studying the eye examination process measure.

To investigate the possible effect of missing data on our results, we repeated key analyses among subsets of patients who were likely to have complete data. We restricted process measures to patients 65 years or older, who would presumably use Medicare when not using the VA and thus would have complete data for process measures. We restricted outcome measures to patients who had an A1C or LDL-C test within the VA at least once during the study (ie, those for whom laboratory values were available).

Finally, we added the mean daily opioid dose to our models and examined its ability to risk stratify the cases regarding diabetes performance measures. Our analyses were conducted using SAS version 9.1 (SAS Inc, Cary, NC).



Copyright AJMC 2006-2020 Clinical Care Targeted Communications Group, LLC. All Rights Reserved.
Welcome the the new and improved, the premier managed market network. Tell us about yourself so that we can serve you better.
Sign Up