Prescription opioid use among older adults with arthritis or low back pain




To examine patterns of chronic opioid use in selected groups with arthritis and low back pain and compare them with patterns among persons with ischemic heart disease.


The study database consisted of Medicare beneficiaries who were enrolled in a drug benefit program for low-to-moderate income Pennsylvania residents. We identified selected patients who had a diagnosis of rheumatoid arthritis, osteoarthritis, chronic low back pain, or ischemic heart disease since 1995. Chronic opioid use, defined as at least six 30-day prescriptions in a year, was the endpoint of interest. We examined the proportion of patients meeting this definition during the period 1996–2001 and determined predictors based on multivariable Cox proportional hazards models.


Four percent of subjects with rheumatoid arthritis used opioids chronically in 2001, compared with <1% in each of the other groups. There was no increase in the chronic use of opioids over the 6-year study period. Low-potency opioids were the most commonly prescribed preparations for chronic users from all patient groups. The prior use of medicines for psychiatric illness, including benzodiazepines or barbiturates, was associated with chronic prescription opioid use across all diagnoses. However, subjects with a prior diagnosis of psychiatric illness were less likely to receive chronic opioids.


Chronic opioid use is relatively uncommon, even among older individuals with arthritis or low back pain. The proportion of these individuals receiving such medicines has not increased in the late 1990s. There seems to be a complex relationship between psychiatric medication use, psychiatric diagnoses, and the use of chronic opioids among these individuals.


Twenty-seven percent of American adults report experiencing chronic musculoskeletal pain and experiencing a limitation in activity due to arthritis (1), and community-based surveys suggest that 40% of adults >65 years of age report arthritis in at least 1 joint area (2). Although nonpharmacologic measures are effective for treating certain types of pain, prescription analgesics are a mainstay of clinical management. Two pharmacologic options for chronic pain include opioid and nonopioid drugs. Nonopioid analgesics are typically first line; however, the American Pain Society recognized the important role that opioid analgesics may play for selected patients (3).

Chronic musculoskeletal syndromes related to arthritis and low back etiologies prompt many patients to take opioids. The American College of Rheumatology's recommendations for the medical management of osteoarthritis of the hip and knee discuss tramadol and opioids as treatments to be considered if acetaminophen, nonsteroidal antiinflammatory drugs, and selective cyclooxygenase 2 inhibitors are not sufficient (4). A recent committee of the American Pain Society suggested a central role for opioids in treating severe arthritis pain that was unresponsive to nonsteroidal antiinflammatory drugs (5).

Although randomized controlled clinical trial data support the use of opioids in a variety of painful conditions and several professional societies acknowledge the role of opioids in treating such patients, there is still great controversy over opioid analgesics' role in clinical practice (6). One study using the National Ambulatory Medical Care Survey found that the use of opioid analgesics for chronic pain doubled from 8% in 1980 to 16% in 2000 (7). The third National Health and Nutrition Examination Survey found similar rates of analgesic medication use (8); however, these data are cross-sectional and do not allow one to accurately depict the antecedent predictors of opioid analgesic use. Several other small studies from rheumatology practices have examined opioid use in specific patient populations cared for by selected providers (9, 10). These studies may not represent typical community-based prescription patterns.

We examined the prevalence of chronic prescription opioid analgesic use among low-moderate income Medicare beneficiaries residing in one US state who had been diagnosed with osteoarthritis, rheumatoid arthritis, chronic low back pain, or, as a comparison, ischemic heart disease. A large database of health care claims information was used for these analyses. Although such data do not allow one to determine the appropriateness of an analgesic regimen in a given patient, they do allow inspection of trends and predictors in opioid use.


Study population.

Eligible subjects included all persons who were simultaneous beneficiaries of Medicare and the Pharmaceutical Assistance Contract for the Elderly (PACE) in Pennsylvania. The PACE program provides drug insurance for low-to-moderate–income elderly with annual household incomes between $10,000 and approximately $17,000. Participants had to be enrolled in and active users of Medicare and PACE for at least 3 consecutive 6-month periods during 1996–2001. Beneficiaries demonstrated active use by filling at least 1 prescription as well as having at least 1 health care encounter in each of the 6-month periods. To ensure adequate baseline and followup information, we required that two of the three 6-month periods occurred prior to the subject's index diagnosis (see below) and the third period occurred after the diagnosis.

From this pool of eligible subjects, we selected several groups of participants for the study population. These included subjects who were likely to have chronic pain because of 3 relatively common diagnoses in older adults: osteoarthritis, rheumatoid arthritis, and chronic low back pain. We also included 1 group of subjects without a clear cause for chronic opioid use, patients with ischemic heart disease, for comparison. Although ischemic heart disease is associated with chest pain, most clinicians do not consider it a common reason for chronic opioid analgesics. These groups were defined based on algorithms requiring visits and/or hospitalizations with specific International Classification of Diseases, Ninth Revision, Clinical Modification diagnosis codes (Appendix A). Subjects who qualified for >1 diagnosis group were excluded, except for subjects with rheumatoid arthritis who were allowed to have a diagnosis of osteoarthritis. Also, patients with a diagnosis of a malignant neoplasm were excluded. The study protocol was approved by the appropriate institutional review board.

Study endpoints: prescription opioid use.

We studied the use of prescription opioids including codeine, hydrocodone, hydromorphone, fentanyl, levorphanol, meperidine, methadone, morphine, oxycodone, oxymorphone, pentazocine, propoxyphene, and tramadol. Combination products containing 1 of these opioids as an ingredient were included, except those used as antitussive agents. Prescription drug information was obtained through pharmacy claims to PACE. This database includes the drug name, dose, dispensing date, and quantity for all prescription drugs paid for by PACE. Information can be linked to specific beneficiaries through unique identifiers. During the study period, PACE placed no specific limitations on opioid dispensing.

The primary endpoint of interest was chronic opioid prescription use, defined as at least a 180-day continuous supply of opioids. Because chronic users of any prescription medicine may experience relatively brief episodes without a supply of medicine, we allowed for up to 30 days between the end of one prescription and the beginning of the next. If a participant received multiple overlapping prescriptions for opioids, each overlapping day was only counted once. As secondary endpoints, we also assessed the prevalence of filling at least 1 opioid prescription and 2 overlapping or consecutive opioid prescriptions.

Statistical analyses.

We first assessed relevant characteristics of the study cohort using data from the 12 months prior to the first diagnosis of the relevant condition. For example, if a beneficiary was diagnosed with rheumatoid arthritis in January 1998, data from 1997 were examined for the covariates of interest. These characteristics included demographic, clinical, and health system use variables. Next, we examined the prevalence of the prescription opioid endpoints (chronic use, at least 2 prescriptions, and any 1 prescription) and the specific preparations taken by subjects with chronic use. Finally, we developed multivariable Cox proportional hazards models with chronic opioid use as the dependent variable. Models included all covariates because of a priori hypothesized relationships with the outcome. Separate models were analyzed for each group of subjects with the different diagnoses. P values less than 0.05 were considered statistically significant. All analyses were run using SAS 8.2 (SAS Institute, Cary, NC).


We defined large, mutually exclusive groups of subjects with the diagnoses of interest (Table 1). Notable differences across diagnosis groups included sex distribution; prevalence of nursing home residence; mean number of visits to physicians; mean number of comorbid conditions; prevalence of orthopedic or other surgeries; and the use of nonbenzodiazepine, nonbarbiturate psychiatric medicines. During the 6-year study period, the proportion of subjects that were chronic opioid users ranged from 3.3% of subjects with ischemic heart disease to 24.3% of subjects with rheumatoid arthritis. There were no substantial temporal trends for opioid use over the study period (Figure 1). In all 3 opioid use categories (any use, at least 2 prescriptions, or chronic use), subjects with rheumatoid arthritis were always more likely to be using opioids than the other subject groups. However, subjects with ischemic heart disease had use patterns quite similar to those of subjects with osteoarthritis or chronic low back pain.

Table 1. Characteristics of patients in the year prior to the qualifying diagnosis*
CharacteristicRheumatoid arthritis (n = 2,956)Osteoarthritis (n = 8,276)Chronic low back pain (n = 3,178)Ischemic heart disease (n = 3,689)
  • *

    Values are the number (percentage) unless otherwise indicated. Columns may not add to 100% because of rounding. NSAID = nonsteroidal antiinflammatory drug; GI = gastrointestinal.

Age, years    
 65–741,156 (39.1)2,349 (28.4)1,526 (48.0)1,348 (36.5)
 75–841,307 (44.2)3,690 (44.6)1,241 (39.1)1,641 (44.5)
 ≥85493 (16.7)2,237 (27.0)411 (12.9)700 (19.0)
Female sex2,719 (92.0)7,393 (89.3)2,578 (81.1)2,384 (64.6)
 White2,752 (93.1)7,564 (91.4)3,031 (95.4)3,412 (92.5)
 African American175 (5.9)626 (7.6)106 (3.3)242 (6.6)
 Other29 (1.0)86 (1.0)41 (1.3)35 (1.0)
Nursing home residence109 (3.7)247 (3.0)52 (1.6)101 (2.7)
Hospitalized for any reason356 (12.0)798 (9.6)345 (10.9)660 (17.9)
Number of physician visits, mean ± SD5.1 ± 6.13.3 ± 3.84.5 ± 4.53.9 ± 4.3
Number of medications, mean ± SD4.1 ± 3.63.5 ± 2.94.1 ± 3.35.1 ± 3.6
Comorbidity index, mean ± SD0.8 ± 1.10.6 ± 0.90.6 ± 0.91.1 ± 1.3
Orthopedic surgery or fracture663 (22.4)923 (11.2)299 (9.4)186 (5.0)
Diagnosis of osteoporosis437 (14.8)444 (5.4)211 (6.6)50 (1.4)
Other surgery1,130 (38.2)2,739 (33.1)1,023 (32.2)1,114 (30.2)
NSAID use in prior year1,075 (36.4)1,777 (21.5)439 (13.8)227 (6.2)
Opioid use in prior year979 (33.1)1,268 (15.3)574 (18.1)377 (10.2)
History of ulcer or GI bleed34 (1.2)61 (0.7)20 (0.6)44 (1.2)
Renal disease133 (4.5)298 (3.6)134 (4.2)282 (7.6)
Addiction or substance abuse14 (0.5)22 (0.3)17 (0.5)18 (0.5)
Other psychiatric diagnosis298 (10.1)1,002 (12.1)432 (13.6)372 (10.1)
Benzodiazepine or barbiturate use451 (15.3)1,115 (13.5)578 (18.2)520 (14.1)
Other psychiatric medication use454 (15.4)1,217 (14.7)533 (16.8)435 (11.8)
Figure 1.

Proportion of subjects using opioids based on different definitions: A,any opioid use, B,at least 2 opioid prescriptions, and C,chronic opioid use. These time periods are defined in the Subjects and Methods section. ▪ = rheumatoid arthritis; □ = osteoarthritis; ○ = low back pain; ▵ = ischemic heart disease.

The majority of prescriptions filled by subjects during their first period of chronic opioid use were for low-potency, short-acting opioids (Table 2). This pattern was consistent across subjects in each of the diagnosis cohorts. Several high-potency agents were rarely used by any of the subjects.

Table 2. Opioid preparations for chronic users during first period of chronic opioid use*
PreparationRheumatoid arthritis (n = 719)Osteoarthritis (n = 629)Chronic low back pain (n = 513)Ischemic heart disease (n = 123)
  • *

    Values are the number (percentage of different opioid preparations).

  • Potency of agents was based on equianalgesic dosing tables from reference13.

Low potency    
 Codeine154 (10.9)119 (12.2)57 (11.1)35 (14.5)
 Hydrocodone234 (16.6)156 (16.0)99 (19.3)33 (13.6)
 Meperidine5 (0.4)3 (0.3)4 (0.8)1 (0.4)
 Propoxyphene485 (34.3)418 (42.9)146 (28.5)85 (35.1)
 Tramadol177 (12.5)124 (12.7)57 (11.1)18 (7.4)
High potency    
 Fentanyl84 (5.9)30 (3.1)33 (6.4)14 (5.8)
 Hydromorphone2 (0.1)0 (0.0)2 (0.4)0 (0.0)
 Levorphanol0 (0.0)0 (0.0)0 (0.0)0 (0.0)
 Methadone1 (0.1)0 (0.01 (0.2)1 (0.4)
  Short acting26 (1.8)10 (1.0)9 (1.8)8 (3.3)
  Long acting1 (0.1)0 (0.0)1 (0.2)0 (0.0)
  Short acting165 (11.7)81 (12.7)66 (12.9)36 (14.9)
  Long acting68 (4.8)30 (3.1)33 (6.4)9 (3.7)
 Pentazocine10 (0.7)3 (0.3)6 (1.2)2 (0.8)

The prior use of medicines for psychiatric illness, including benzodiazepines or barbiturates, was associated with chronic prescription opioid use across all diagnosis groups (Table 3). However, subjects with a prior diagnosis of psychiatric illness were less likely to receive chronic opioids. Also, a greater number of physician visits was associated with chronic opioid use. African American race was associated with a lower probability of receiving chronic opioids; this was only statistically significant for the rheumatoid arthritis cohort, but the trend was seen for all groups. Finally, residence in a nursing home was associated with chronic opioid use in several of the subject groups.

Table 3. Multivariable predictors of chronic opioid treatment*
Patient characteristicsRheumatoid arthritisOsteoarthritisChronic low back painIschemic heart disease
  • *

    Values are the adjusted odds ratios (95% confidence intervals). Multivariable models adjusted for all other variables listed. NSAID = nonsteroidal antiinflammatory drug; GI = gastrointestinal.

  • The comorbidity index was calculated using previously described methods for health care utilization data (14).

  • Odds ratio could not be estimated because of small numbers.

Age, years    
 75–840.9 (0.7–1.0)0.9 (0.8–1.1)1.0 (0.7–1.3)0.6 (0.4–0.9)
 ≥850.7 (0.5–0.9)1.0 (0.8–1.3)1.2 (0.8–1.7)0.5 (0.3–0.9)
Male sex0.7 (0.5–0.9)0.7 (0.5–1.0)1.2 (0.8–1.6)0.8 (0.5–1.2)
 African American0.6 (0.4–0.9)0.7 (0.5–1.0)0.5 (0.2–1.3)0.4 (0.2–1.2)
 Other0.4 (0.1–1.2)0.9 (0.4–2.0)1.4 (0.6–3.8)2.7 (0.9–8.9)
Hospitalization in prior year1.1 (0.8–1.4)0.7 (0.4–1.0)0.9 (0.5–1.4)0.6 (0.3–1.0)
Nursing home in prior year1.3 (0.9–2.0)0.7 (0.3–1.7)2.0 (1.0–4.4)2.6 (1.1–6.2)
Number of drugs    
 5–70.8 (0.6–1.0)0.8 (0.6–1.0)0.9 (0.6–1.3)0.6 (0.3–1.1)
 ≥80.8 (0.6–1.0)0.8 (0.6–1.0)1.3 (0.8–2.1)0.6 (0.3–1.4)
Number of physician visits    
 7–121.2 (1.0–1.5)1.3 (1.1–1.6)1.4 (1.0–1.9)1.2 (0.7–1.9)
 ≥131.7 (1.3–2.1)1.6 (1.2–2.2)1.7 (1.2–2.5)2.8 (1.7–4.6)
Comorbidity index
 10.8 (0.6–0.9)0.9 (0.8–1.2)1.0 (0.7–1.2)1.4 (0.9–2.2)
 ≥20.7 (0.6–0.9)1.2 (0.9–1.6)1.1 (0.8–1.6)1.4 (0.9–2.4)
Orthopedic surgery or fracture1.4 (1.2–1.7)0.9 (0.7–1.2)1.2 (0.8–1.8)1.1 (0.5–2.3)
Diagnosis of osteoporosis1.2 (0.9–1.4)0.8 (0.6–1.3)1.1 (0.7–1.7)1.5 (0.5–4.9)
Other surgery1.0 (0.8–1.2)0.9 (0.7–1.1)0.8 (0.6–1.0)1.1 (0.8–1.7)
NSAID use in prior year1.1 (0.9–1.2)1.8 (1.5–2.1)1.6 (1.2–2.1)1.4 (0.8–2.5)
History of ulcer or GI bleed0.9 (0.4–1.8)1.3 (0.5–3.7)2.7 (1.0–7.6)0.7 (0.1–5.0)
Renal disease0.9 (0.6–1.3)0.7 (0.4–1.3)1.2 (0.7–2.2)1.1 (0.6–2.1)
Substance abuse or addiction0.8 (0.2–3.3)2.7 (0.9–8.7)0.5 (0.1–3.8)
Other psychiatric diagnosis0.7 (0.5–0.9)0.5 (0.4–0.7)0.7 (0.5–1.0)0.7 (0.4–1.4)
Benzodiazepine or barbiturate use1.5 (1.2–1.8)1.8 (1.5–2.2)1.9 (1.5–2.6)1.8 (1.2–2.8)
Other psychiatric medication use1.3 (1.0–1.6)1.5 (1.2–1.9)1.9 (1.4–2.5)1.7 (1.0–2.8)


Chronic pain is common and will most likely increase in prevalence as the population ages. Although prescription opioids should not be considered first-line treatment for most types of chronic pain, they play an important adjunctive role and are part of the recommended treatment guidelines for many chronic painful conditions. However, little is known about the use of opioids in typical community-based practice (11). To better understand this, we studied a large group of low-moderate income older adults participating in a statewide drug benefits plan. We compared the prevalence and predictors of chronic opioid use across relevant diagnosis groups, including rheumatoid arthritis, osteoarthritis, chronic low back pain, and ischemic heart disease. As anticipated, we found that subjects with ischemic heart disease used chronic opioids infrequently. The types of opioids used were quite similar across diagnoses, and the subject characteristics associated with chronic opioid use were similar across diagnoses.

These analyses do not allow us to comment on the appropriateness of a given individual's opioid prescriptions. We do not have access to medical records and have no information on a given subject's level of pain. Rather, these data illustrate broad patterns that help us to begin to understand opioid prescribing. These analyses may allow for a critical assessment of physicians' patterns of opioid use. Several specific findings merit discussion.

We found that the prevalence of chronic opioid use among subjects with chronic painful conditions is relatively low. Even when we consider subjects with rheumatoid arthritis where two-thirds receive at least 1 opioid prescription, chronic use is uncommon. Treatment for rheumatoid arthritis should be aimed at reducing inflammation with disease-modifying antirheumatic drugs, but pain control is a major goal and may not always be achieved with antirheumatic treatments. One prior survey of subjects attending a rheumatology practice at a Veterans Affairs hospital found a proportion of patients with rheumatoid arthritis taking chronic opioids similar to this study (9). However, neither prior studies nor the present study answer several important questions about opioid use in persons with rheumatoid arthritis. Do some persons with rheumatoid arthritis remain in pain and yet never receive a trial of prescription opioids? Moreover, do some individuals with rheumatoid arthritis take chronic opioids without an adequate trial of antirheumatic treatment? We suspect that chronic pain from musculoskeletal conditions, such as arthritis, often goes under treated. These are questions for future studies that include more detailed clinical information. If patients with arthritis are found to have under-treated pain, then patient and physician interventions might be pursued.

The majority of prescriptions filled by persons taking chronic opioids were for short-acting agents. Experts in the field recommend that long-acting agents should be preferentially used in patients requiring chronic analgesia (12). It is possible that either long-acting preparations are not as effective as short-acting analgesics in typical practice or that current prescribing practices are suboptimal. Our analysis of the predictors of opioid use suggests that the use of medications for psychiatric conditions (e.g., antidepressants, anxiolytics, and antipsychotics), including barbiturates and benzodiazepines, is associated with an increased probability of chronic opioid use. It may be that some of these psychiatric medicines are being used as adjunctive analgesics. However, it also may be the case that some patients and doctors are predisposed to use medicines for chronic physical and psychic pain. In light of these findings regarding psychiatric medicines, it is interesting to note that subjects with psychiatric diagnoses were less likely to receive chronic opioids. This may reflect doctors' reluctance to use chronic opioids in patients with diagnosed psychiatric illness but not in patients taking medicines used for psychiatric conditions.

Explanations for these correlates of chronic opioid use are complex. One possibility is that persons with certain characteristics, such as African American race or psychiatric diagnoses, are less likely to report pain or ask for treatments. As well, selected groups of individuals, such as African Americans or those with psychiatric diagnoses, may be more reluctant to accept physicians' recommendations of prescription opioids. Furthermore, doctors may selectively prescribe opioids to patients who are more likely to benefit or less likely to be harmed by opioids. Many other explanations can be postulated based on these data, but we are unable to examine such mechanistic issues using health care utilization data.

These results must be viewed in light of several limitations. Study participants were all older adults and beneficiaries of Medicare and a drug prescription program for low-moderate income elderly. These analyses should also be carried out on younger and more affluent patients. The information we have on clinical diagnoses is based on health care claims that probably misclassify some patients. In other words, some participants with health care claims for rheumatoid arthritis may not fulfill diagnostic criteria for this diagnosis. Such misclassification is less of an issue when assessing patterns of prescription medication use versus the incidence of an illness. Finally, and most important, our study database does not contain detailed clinical information on why participants were using prescription opioids. There is a possibility that some of the occasional opioid use was for reasons other than the arthritis or musculoskeletal diagnoses, such as a dental extraction. Also, we have no information on participants' reported pain levels.

In conclusion, we found a relatively low frequency of chronic prescription opioid use among low-moderate income older adults who participate in a drug benefits program. Participants with rheumatoid arthritis were more likely to use chronic opioids than those with osteoarthritis, chronic low back pain, or ischemic heart disease. The use of long-acting prescription opioids among chronic users was surprisingly low. With an aging population, rates of chronic pain will increase. To ensure rational use of analgesics, more detailed studies of patterns of chronic opioid use would be valuable.


Table  . Appendix A: Definitions of Study Groups*
Study groupCriteria
  • *

    ICD-9-CM = International Classification of Diseases, Ninth Revision, Clinical Modification; CPT = Current Procedural Terminology.

OsteoarthritisTwo inpatient or outpatient visits associated with ICD-9-CM diagnosis codes 715.00–715.99
Rheumatoid arthritisTwo inpatient or outpatient visits associated with ICD-9-CM diagnosis code 714.0
Chronic low back painTwo inpatient or outpatient visits associated with ICD-9-CM diagnosis codes 720.00–724.99
Ischemic heart diseaseAn inpatient or outpatient visit associated with ICD-9-CM diagnosis codes 410.00–413.99 (acute myocardial infarction or angina); diagnosis related groups 106, 107, 112, 121, 122, 123, 124, 125, or 132 (acute myocardial infarction or coronary intervention); or CPTs 33510–33519, 33521–33523, 3350, 33533–33536, 33545 (coronary revascularization procedures)