Association of Initial Opioid Prescription Duration and an Opioid Refill by Pain Diagnosis: Evidence from Outpatient Settings in Ten US Health Systems

Abstract

Objective:

The Centers for Disease Control and Prevention’s 2022 Clinical Practice Guideline for Prescribing Opioids for Pain cautioned that inflexible opioid prescription duration limits may harm patients. Information about the relationship between initial opioid prescription duration and a subsequent refill could inform prescribing policies and practices to optimize patient outcomes. We assessed the association between initial opioid duration and an opioid refill prescription.

Methods:

We conducted a retrospective cohort study of adults ≥19 years of age in 10 US health systems between 2013 and 2018 from outpatient care with a diagnosis for back pain without radiculopathy, back pain with radiculopathy, neck pain, joint pain, tendonitis/bursitis, mild musculoskeletal pain, severe musculoskeletal pain, urinary calculus, or headache. Generalized additive models were used to estimate the association between opioid days’ supply and a refill prescription.

Results:

Overall, 220,797 patients were prescribed opioid analgesics upon an outpatient visit for pain. Nearly a quarter (23.5%) of the cohort received an opioid refill prescription during follow-up. The likelihood of a refill generally increased with initial duration for most pain diagnoses. About 1 to 3 fewer patients would receive a refill within 3 months for every 100 patients initially prescribed 3 vs. 7 days of opioids for most pain diagnoses. The lowest likelihood of refill was for a 1-day supply for all pain diagnoses, except for severe musculoskeletal pain (9 days’ supply) and headache (3–4 days’ supply).

Conclusions:

Long-term prescription opioid use increased modestly with initial opioid prescription duration for most but not all pain diagnoses examined.

INTRODUCTION

Amid the ongoing opioid overdose crisis, concerns remain regarding the transition from acute to chronic opioid use and the significant health and mortality risks that prolonged opioid use poses.^1–6 In response, states, health systems, and insurers have imposed strict limits on the duration of opioid prescriptions for acute pain.^7,8 The Centers for Disease Control and Prevention (CDC)’s clinical opioid prescribing guideline from 2022 described concerns about misapplications of the 2016 guideline, including inflexible duration limits, that could have contributed to patient harm, such as undertreated pain and worsening pain outcomes.^9–11 Understanding initial analgesic opioid prescription duration and long-term use for patients presenting with common acute pain conditions could inform prescribing policies and practices to improve patient management and outcomes.

In this study, we assessed the association of initial opioid prescription days’ supply and the likelihood of a subsequent opioid refill among adult patients presenting with acute pain in outpatient settings. We leveraged data from a large cohort of commercially and publicly insured patients who were newly initiated on opioid analgesics between 2013 and 2018 from 10 US health systems. Following prior work, we evaluated this relationship separately for pain conditions commonly treated in outpatient settings that include tendonitis/bursitis, urinary calculus, headache, and back, neck, joint, and musculoskeletal pain.^12,13 We discuss implications of results for establishing clinical practices that minimize risks of developing prolonged opioid use and unintended harms.

METHODS

Setting and data source

All data for the cohort study were drawn from a multi-site prescription opioid registry funded by the National Institute on Drug Abuse, National Drug Abuse Treatment Clinical Trials Network (CTN-0084). The registry comprises harmonized electronic health record (EHR) and claims data from 10 health systems that represent a mix of commercially and publicly insured populations and include: Baylor Scott and White (Texas); Essentia Health System (Minnesota, North Dakota, Wisconsin); Geisinger (Pennsylvania); Henry Ford Health (Michigan); Kaiser Permanente Colorado; Kaiser Permanente Mid-Atlantic States (Maryland, Virginia, Washington DC); Kaiser Permanente Northern California; Kaiser Permanente Northwest (Oregon and Southwest Washington); Kaiser Permanente Southern California; and Meyers Health Care Institute/Fallon Health System (Massachusetts). The registry uses a common data model to collect data on patients ≥18 years of age who either were prescribed opioids in outpatient settings or received an opioid use disorder (OUD) diagnosis between January 1, 2012 and December 31, 2018. Details regarding development of the registry and data collected have been previously described.¹⁴ Briefly, these data contain socio-demographic, membership, and clinical information, including diagnoses from medical encounters and prescriptions from pharmacy dispensing or medication orders. Seven health systems in the registry captured dispensing, 2 systems had medication orders, and 1 system had dispensing for members and orders for non-members. The study was approved by the Kaiser Permanente Northern California Institutional Review Board (IRB), with a waiver of informed consent and with IRB ceding from all other study sites. The study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline.

Study Population

The study population consisted of patients ≥19 years of age in the registry who had a first opioid analgesic prescription associated with a pain-related outpatient visit between 2013 and 2018. Nine pain diagnoses commonly observed in outpatient settings, identified from prior studies, were examined: back pain without radiculopathy, back pain with radiculopathy, neck pain, joint pain, tendonitis/bursitis, mild musculoskeletal pain, severe musculoskeletal pain, urinary calculus, and headache.^12,13 For each patient, the first observed outpatient visit with one of these diagnoses was retained and represented the index pain visit. When there were multiple pain diagnoses or prescriptions on the same index date, one diagnosis was randomly selected for inclusion in the registry. We required a baseline period of ≥1 year prior to the index pain visit to collect accurate data on patient characteristics. As a result, we excluded individuals in the registry with an index pain visit before January 1, 2013 and individuals 18 years old because data prior to age 18 were not available. Patients were considered newly initiated on opioid therapy if they had not received any opioid analgesic prescription in the year prior to the index visit. The first opioid analgesic prescription within 7 days following the index visit was considered the index (initial) opioid prescription. The follow-up period began after the index prescription, extended up to 180 days after the index days’ supply end date, and ended on December 31, 2018.

Individuals in the registry were excluded from the study if they: 1) had missing information on sex or days’ supply or dose of the index opioid prescription; 2) had an index opioid prescription that exceeded 30 days’ supply, which may indicate prescribing for chronic rather than acute pain; 3) died during follow-up, as death would preclude subsequent receipt of an opioid prescription; 4) had an acute inpatient hospitalization or a same-day surgery during the 1-year baseline period, as prescribing for postoperative pain may warrant different clinical considerations than for non-postoperative pain¹⁵; 5) did not have continuous health plan membership (measured with enrollment data in 7 health systems, a utilization-based algorithm in 2 systems, or both in 1 system)^14,16 during the 1-year baseline and follow-up periods to allow for accurate capture of opioid prescriptions; or 6) had any outpatient opioid prescriptions during the baseline period.

Measures

Study Outcome

The primary outcome was an opioid refill prescription, defined as any opioid analgesic prescription filled or ordered after the index prescription and in the 90-day period after the index days’ supply end date. In sensitivity analyses, the outcome was examined in 30-, 60-, and 180-day periods after the index prescription supply end date.

Exposures and Covariates

The exposure of interest was initial opioid prescription days’ supply with values from 1 to 30 (representing the minimum and maximum values). From the registry, we also extracted the following covariates for each patient: sex; age at the index outpatient visit (19-<40, 40-<65, 65-<80, ≥80 years); self-reported race and ethnicity (Asian, Black, Hispanic, White, other or unknown); study site; the Deyo version of the Charlson comorbidity index score (CCI; 0, 1–2, 3–4, ≥5); and a benzodiazepine prescription, OUD, alcohol use disorder (AUD), nicotine use disorder, other substance use disorder (SUD), and mental health disorder in the baseline period. For the index prescription, we calculated daily morphine milligram equivalents (MME), which was divided into three categories (<50, 50 to <90, and ≥90 MME).

Statistical Analysis

We first calculated descriptive statistics of patient and clinical characteristics and the initial opioid duration. In adjusted analyses, we used marginal standardization to estimate the proportion of patients expected to receive an opioid refill had the study population received each of the 30 possible initial prescription days’ supply (1 to 30).^17–19 The estimated proportion represents the population-based, marginal risk associated with each value of opioid days’ supply, accounting for individual-level patient covariates.^17–19 Marginal standardization was conducted in four steps and implemented separately for each pain diagnosis. First, we used an approach adapted from Scully and colleagues to fit a generalized additive model (GAM) to measure the association between initial opioid days’ supply and a refill.¹⁵ Models used spline smoothing to fit the continuous measure of days’ supply and were adjusted for patient and clinical covariates noted above. Second, we used coefficients from the fitted spline-smoothed GAMs to estimate probabilities of a refill for each patient by setting the initial days’ supply at each value from 1 to 30. Third, we averaged the estimated probabilities of a refill for each days’ supply value across the cohort and multiplied the resulting probabilities by 100 to get percentages. Fourth, we used bootstrapping with 1,000 replications to calculate 95% confidence intervals.²⁰ In sensitivity analyses, because the opportunity to detect a refill may depend on the length of follow-up, we repeated the procedure above with follow-up set at 30-, 60-, and 180-days after the index prescription supply end date.

RESULTS

Characteristics of the Study Cohort

During the study period, 2,454,417 patients had a pain-related outpatient visit, of whom 530,407 (21.6%) had an opioid analgesic prescription within 7 days of the index pain visit (Supplemental Figure 1). After exclusion criteria were applied, the final analytic cohort consisted of 220,797 patients. The mean (standard deviation [SD]) age of the cohort was 51.0 (16.5) years (Table 1). A total of 110,443 (50.0%) patients were women, 16,547 (7.5%) were Asian, 22,201 (10.1%) were Black, 113,553 (51.4%) were White, 54,885 (24.9%) were Hispanic of any race, and 13,611 (6.2%) had other or unknown race and ethnicity. More than three-quarters (77.2%) of the cohort had a CCI score of 0. In the baseline period, <1% of the cohort had an OUD diagnosis (<0.1%), AUD diagnosis (0.8%), or other SUD diagnosis (0.4%), and 4.6% had a nicotine use disorder. Additionally, a total of 25,348 (11.5%) of patients had at least one mental health related diagnosis, and 14,768 (6.7%) had a benzodiazepine prescription.

Table 1.

Characteristics of patients with an index outpatient visit for pain and an associated index opioid analgesic prescription from 10 US health systems, 2013–2018^a

Characteristic, no. (%)	Study Cohort (N=220,797)
Sex
Female	110,443 (50.0)
Male	110,354 (50.0)
Age, mean (SD)	51.0 (16.5)
Age Group
19 - <40	58,031 (26.3)
40 - <65	114,052 (51.7)
65 - <80	39,027 (17.7)
≥80	9,687 (4.4)
Race and Ethnicity
Asian	16,547 (7.5)
Black	22,201 (10.1)
Hispanic	54,885 (24.9)
White	113,553 (51.4)
Other or unknown	13,611 (6.2)
Charlson comorbidity index (CCI) score
0	170,544 (77.2)
1–2	39,077 (17.7)
3–4	7,972 (3.6)
≥5	3,204 (1.5)
Baseline diagnosis	148 (<0.1)
Opioid use disorder
Alcohol use disorder	1,769 (0.8)
Nicotine use disorder	10,220 (4.6)
Other substance use disorder	789 (0.4)
Mental health disorder	25,348 (11.5)
Benzodiazepine prescription in baseline period	14,768 (6.7)
Index pain diagnosis
Back pain without radiculopathy	74,057 (33.5)
Back pain with radiculopathy	19,882 (9.0)
Neck pain	20,213 (9.2)
Joint pain	53,111 (24.1)
Tendonitis/bursitis	13,689 (6.2)
Mild musculoskeletal pain	15,441 (7.0)
Severe musculoskeletal pain	1,038 (0.5)
Urinary calculus	8,402 (3.8)
Headache	14,964 (6.8)
Index opioid prescription days’ supply
1–3 days	45,417 (20.6)
4–6 days	72,952 (33.0)
7 days	49,375 (22.4)
≥8 days	53,053 (24.0)
Index opioid prescription dose, daily morphine milligram equivalents (MME)
<50 MME	199,437 (90.3)
50 to <90 MME	18,578 (8.4)
≥90 MME	2,782 (1.3)

^aPersons included in the study had one of the pain visits of interest and an outpatient opioid prescription within 7 days of the visit date.

Duration of the Index Opioid Prescription

More than three-quarters (76%) of the cohort were prescribed opioids for 7 days or fewer: 1–3 days (20.6%), 4–6 days (33.0%), 7 days (22.4%), and ≥8 days (24.0%) (Table 1). The overall mean (SD) days’ supply of the initial opioid prescription was 7.4 (5.8), and the median (IQR) was 5 (2) (Table 2). There was variability in initial days’ supply across and within pain diagnoses. The mean duration ranged from 5.2 days for urinary calculus to 8.2 days for joint pain, and the median (IQR) ranged from 5 (2) for back pain without radiculopathy and neck pain to 7 (3) for joint pain. Days supplied for joint pain showed the most variability (SD=6.6), whereas urinary calculus showed the least (SD=3.4). Most patients received an index opioid prescription with daily MME of <50 (90.3%) (Table 1).

Table 2.

Index opioid prescription days’ supply and opioid refill prescription after the index prescription and in 90-day follow up period by pain diagnosis. Data from 10 US health systems, 2013–2018.

Pain Diagnosis	Index opioid fill days’ supply, mean (SD)	Index opioid fill day’s supply median (IQR)	Any opioid refill after the index fill and in 90-day follow-up period, no. (%)
All pain types	7.4 (5.8)	5 (2)	51,967 (23.5)
Back pain w/o radiculopathy	7.3 (5.7)	5 (2)	16,461 (22.2)
Back pain w/ radiculopathy	7.6 (5.7)	7 (3)	6,329 (31.8)
Neck pain	7.2 (5.7)	5 (2)	4,472 (22.1)
Joint pain	8.2 (6.6)	7 (5)	13,782 (25.9)
Tendonitis/bursitis	7.6 (5.9)	6 (3)	2,836 (20.7)
Mild musculoskeletal pain	6.6 (4.8)	5 (3)	3,337 (21.6)
Severe musculoskeletal pain	7.0 (5.3)	5 (3)	284 (27.4)
Urinary calculus	5.2 (3.4)	5 (4)	2,015 (24.0)
Headache	6.2 (4.9)	5 (4)	2,451 (16.4)

Abbreviations: SD = standard deviation, IQR = inter-quartile range

Association of Initial Opioid Days’ Supply and an Opioid Refill Prescription

Nearly a quarter (23.5%) of the cohort received an opioid prescription refill during the 90-day follow-up (Table 2). Receipt of a refill varied by pain diagnosis. Patients whose index visit was for back pain with radiculopathy had the highest proportion with a refill (31.8%), and those with headache had the lowest proportion (16.4%).

The adjusted proportion of the cohort with an opioid refill estimated from the marginal standardization procedure is shown for each possible value of the index days’ supply (1 to 30 days) in Figure 1. The proportion with a refill generally increased with initial duration, but there was notable heterogeneity in patterns across pain diagnoses. For example, the expected proportion of patients with a refill changed relatively little as days’ supply increased for mild musculoskeletal pain. For severe musculoskeletal pain, the relationship exhibited a non-linear trend.

Figure 1.

Adjusted probabilities of an opioid refill prescription associated with initial opioid prescription duration, by pain diagnosis. Data from 10 US health systems, 2013–2018.

Note: Plots are based on a generalized additive model estimated separately for each pain diagnosis. Dashed lines depict 95% CI bands.

We focus on an example scenario in which all patients initially received 3 and 7 days (the mean) of opioids to illustrate change in the estimated likelihood of a refill. The respective expected proportions with refills for 3 and 7 days were 19.8% and 22.5% for back pain without radiculopathy, 29.5% and 32.9% for back pain with radiculopathy, 19.7% and 22.8% for neck pain, 23.1% and 26.1% for joint pain, 18.2% and 21.3% for tendonitis/bursitis, 20.5% and 22.2% for mild musculoskeletal pain, 27.9% and 26.3% for severe musculoskeletal pain, 23.6% and 24.3% for urinary calculus, and 15.7% and 16.5% for headache (Table 3).

Table 3.

Adjusted likelihood of an opioid refill prescription associated with select initial opioid days’ supply, by pain diagnosis. Data from 10 US health systems, 2013–2018.

	Adjusted probabilities (95% CI)a
	3 days	5 days	7 days	10 days	14 days	30 days
Back pain without radiculopathy	19.8 (19.4,20.4)	21.5 (21.2,21.9)	22.5 (22.1,22.9)	23.2 (22.6,23.8)	24.3 (23.4,25.1)	33.2 (31.5,35.0)
Back pain with radiculopathy	29.5 (28.3,30.8)	31.5 (30.7,32.3)	32.9 (31.9,33.8)	32.8 (31.5,34.1)	32.2 (30.3,34.1)	37.1 (33.5,40.4)
Neck pain	19.7 (18.8,20.7)	21.5 (20.8,22.2)	22.8 (21.9,23.6)	24.0 (22.7,25.3)	24.5 (22.7,26.1)	29.0 (25.8,32.2)
Joint pain	23.1 (22.4,23.7)	24.6 (24.1,25.1)	26.1 (25.6,26.7)	27.4 (26.7,28.2)	28.2 (27.2,29.3)	36.0 (34.2,37.9)
Tendonitis/bursitis	18.2 (17.1,19.4)	20.0 (19.2,20.8)	21.3 (20.3,22.2)	21.8 (20.5,23.3)	21.9 (20.1,23.8)	28.1 (24.3,31.9)
Mild musculoskeletal pain	20.5 (19.4,21.7)	21.6 (20.9,22.4)	22.2 (21.2,23.2)	23.0 (21.5,24.5)	23.7 (21.5,25.9)	22.1 (18.0,26.8)
Severe musculoskeletal pain	27.9 (23.0,33.1)	27.0 (23.8,30.4)	26.3 (22.8,30.5)	26.0 (20.8,31.4)	30.5 (22.2,38.6)	32.2 (15.9,51.5)
Urinary calculus	23.6 (22.2,24.9)	23.8 (22.6,24.9)	24.3 (22.7,25.8)	25.4 (22.8,28.2)	25.8 (21.3,30.4)	30.0 (18.9,42.4)
Headache	15.7 (14.8,16.5)	16.0 (15.3,16.7)	16.5 (15.7,17.3)	16.7 (15.4,18.0)	17.2 (15.2,19.4)	25.6 (20.9,30.6)

^aAdjusted probabilities for all values of the initial opioid prescription duration are reported in Supplemental Table 1.

For all pain conditions except severe musculoskeletal pain, the expected proportion of patients with a refill was reduced if all patients were given an initial days’ supply of 3 rather than 7 days, with the greatest reduction for back pain with radiculopathy (−3.4%) and lowest reduction for urinary calculus (−0.7%). For severe musculoskeletal pain, 1.6% more patients would be expected to receive a refill if all patients were given a 3 days’ supply rather than a 7 days’ supply. In other words, for every 100 patients with pain conditions other than severe musculoskeletal pain initially prescribed 3 instead of 7 days of opioids, we can expect that about 1 to 3 fewer patients would receive a subsequent refill within 3 months depending on the pain diagnosis, all else equal. In contrast, about 2 more patients with severe musculoskeletal pain would receive a refill for every 100 patients initially supplied 3 instead of 7 days. The lowest predicted proportion with a refill was for a 1-day supply for all pain diagnoses, except for severe musculoskeletal pain (9 days’ supply) and headache (3–4 days’ supply) (Supplemental Table 1).

In sensitivity analyses, we set the follow-up period to 30-, 60-, and 180-days after the index prescription day’s supply end date. In the cohort, 16.6%, 20.7%, and 29.6% received any subsequent opioid fill in the 30-, 60-, and 180-day follow-up, respectively (Supplemental Table 2). Overall, estimated proportions of the cohort with a refill for values of the index days’ supply increased with length of follow-up (Supplemental Table 3, Supplemental Table 4, Supplemental Table 5). Across all follow-up lengths, while estimates differed in magnitude, patterns of a refill by initial duration were generally consistent (Supplemental Figure 2, Supplemental Figure 3, Supplemental Figure 4).

DISCUSSION

In this large cohort of adults presenting at an outpatient visit for one of nine pain diagnoses, on average 23.5% of patients received a refill prescription in the 90 days after the initial opioid supply end date. The likelihood of a refill generally increased with number of initial days’ supply for most but not all pain diagnoses. For most pain diagnoses, we can expect that about 1 to 3 fewer patients would receive an opioid refill within 3 months for every 100 patients initially prescribed 3 instead of 7 days of opioids. These findings suggest that small absolute changes in initial opioid duration translate to modest changes in the likelihood of continued opioid treatment for individual patients presenting with acute pain in outpatient settings.

Findings contribute to existing research on the relationship between initial opioid analgesic prescription duration and continued opioid use by drawing on evidence from a large cohort of commercially and publicly insured patients from outpatient settings across multiple health systems. Findings are consistent with prior work using data prior to release of the 2016 CDC opioid prescribing guideline that indicates median durations of 4–7 days for initial opioid analgesic prescriptions among patients presenting with acute pain in primary care settings.¹² Studies have also found that while the likelihood of obtaining a refill depends on the pain condition, generally less than a third of patients do so in the month after their initial opioid analgesic prescription and even a lower proportion in the following year.^12,21,22 Additionally, approximately a fifth of patients presenting with pain at an outpatient visit in this study received a first opioid analgesic prescription, which is consistent with current clinical recommendations that non-opioid therapies, such as nonsteroidal anti-inflammatory drugs, are preferred for acute pain and may be equally, if not more, effective for management of some pain conditions.^10,23,24

Several mechanisms could explain the observed association between initial duration of prescribed opioid analgesics and a refill. First, inadequate pain treatment resulting from an insufficient initial duration could lead to need for additional prescriptions. Second, longer pain duration and heightened severity over time after the initial pain assessment could increase the likelihood of sustained opioid consumption. Third, days’ supply itself could serve as a marker of pain severity, and a longer duration of opioid prescription analgesics may indicate greater pain management needs. Fourth, patients can unintentionally transition to longer-term opioid therapy when opioids are continued without reevaluation.¹⁰ Finally, iatrogenic opioid dependence could result from an initial treatment with opioid analgesics.²⁵

This study suggests that patients should be prescribed the fewest days of opioids necessary to effectively manage pain for most conditions if the goal is to minimize the likelihood of continued opioid use. The exceptions are headache and severe musculoskeletal pain, where the likelihood of a refill was lowest for an initial supply of 3–4 and 9 days, respectively. In principle, universal opioid duration limits could standardize clinical practices and help curb excess opioid medications that may be diverted and contribute to overdose risk, but rigid limits may lead to unintended consequences,⁸ such as undertreatment of pain. Inadequate pain treatment could have further consequences for patients, such as worsening pain outcomes, greater cost burdens due to copayments for additional visits and prescription fills, and a delay in return to work. Results of this study suggest a small absolute reduction or increase in the initial opioid duration (e.g., 7 vs. 3 days) is associated with a modest change in the likelihood of a refill prescription at the individual-level. At the population-level, however, a difference of a few percentage points in the level of continued opioid use may translate to a substantial number of patients with a refill prescription. In balancing individual patient needs and population health, prescription duration limits could serve as a starting point, but departures from such limits should be considered when warranted to maximize the clinical benefit of opioid therapy, minimize potential harms of undertreatment, and adopt individualized, patient-centered decision-making. Variation in patient needs by pain condition should also be considered.

This study has limitations. Capture of external pharmacy claims may be incomplete for patients who obtain opioid prescriptions by cash or other insurance coverage, which could lead to under-estimating the likelihood of a refill. We are unable to determine why patients received (e.g., had unresolved pain or a new pain concern) or did not receive (e.g., did not need or were not empowered to request) a subsequent opioid prescription. While we adjusted for study site in analyses, prescribing policies imposed by individual insurance providers could have affected initial opioid supply and refills. We lacked information on potential confounding factors, such as pain severity over time. We estimated models separately for each pain diagnosis, and results may differ by diagnosis due to patient composition. There is potential for exposure misclassification due to multiple pain diagnoses and prescriptions on the index date. To ensure accurate assessment of refills, we excluded patients who died or did not have continuous enrollment during follow-up. This could have affected results if these patients differed from the final study cohort. Future research is needed to investigate opioid prescription duration and other outcomes, such as overdose and health plan disenrollment,²⁶ which could inform efforts to determine the optimal duration that maximizes clinical benefit and minimizes risks and unintended patient harm.

CONCLUSION

Patients presenting with pain in outpatient settings were generally prescribed a short duration of opioid analgesics, and most did not receive a subsequent opioid prescription. The likelihood of continued opioid therapy increased modestly with duration of the initial opioid supply for most but not all pain diagnoses.