Abstract
Background
Opioid stewardship policies could adversely affect pain management for patients with cancer. Yet patients with cancer are also at risk for opioid‐related harms. This study sought to determine trends in opioid prescribing by clinical stratum and pain for patients with cancer from 2016 to 2020.
Methods
A retrospective study was conducted of opioid‐naive adults with newly diagnosed cancer from 2016 to 2020 (N = 10,232) in a large Connecticut health system. Logistic regression was used to calculate changes in the predicted probability of opioid prescribing from 2016 to 2020. Two subpopulations were examined: patients treated surgically (n = 4405) and patients with metastatic cancer (n = 2158). Flowsheet pain scores for patients with metastatic cancer were used to stratify by no pain (all scores, 0) versus any pain. The main outcomes were new (≥1 prescription in the 0–6 months after diagnosis) and additional (0–6 and 7–9 months) opioid prescriptions.
Results
A decline was observed in the predicted probability of new (71.1% to 64.6%; p < .001) and additional prescribing (27.2% to 24.2%; p = .07 [not significant]) declined. Among surgical patients, the predicted probability of new opioid prescribing fell (96.0% to 88.6%; p < .001), whereas additional prescribing was stable (13%). For patients with metastatic cancer with pain, new opioid prescribing was stable (56%). For those reporting no pain, the predicted probability of new opioid prescribing declined from 61.6% to 36.1% (p < .001).
Conclusions
In the context of widespread policy changes, this study showed a modest decline in new and additional opioid prescribing for patients with cancer. In metastatic cancer, prescribing remained stable for patients reporting pain and declined steeply for those reporting no pain.
Keywords: cancer‐related pain, cancer survivorship, Connecticut, opioid epidemic, opioid prescribing, pain management, postoperative pain, retrospective study
Short abstract
This study investigates trends in new and persistent opioid prescribing for patients with cancer from 2016 to 2020 by focusing on patients with metastatic cancer and those undergoing curative‐intent surgery. A modest decline was found in new opioid prescriptions overall, with a notable reduction for patients with metastatic cancer reporting no pain whereas remaining stable for those with metastatic cancer–related pain.
INTRODUCTION
In response to the opioid crisis, public health efforts have focused on reducing inappropriate opioid prescribing via regulation. 1 , 2 , 3 In addition to state and federal efforts, health care delivery networks have implemented policies to reduce inappropriate prescribing. 4 Evidence suggests that state‐level policies reduce overall opioid prescribing 5 ; whether this represents intended declines in inappropriate use is unknown. There is a further knowledge gap as to how these overlapping efforts affect patients. 5 Since 2016, Connecticut has implemented several broad‐reaching public acts, such as 7‐day prescription limits for nonmalignant pain, a public service advertising campaign, and an electronic prescription mandate. At the same time, the Yale New Haven Health System (YNHHS)–Smilow Cancer Hospital and its regional affiliates, which care for approximately 40% of patients with cancer in Connecticut, attempted to reduce inappropriate opioid prescribing via internal regulations, such as electronic health record (EHR) default prescribing of fewer days’ supply. 6 , 7 It is imperative to understand the impact of these changes on patients with cancer.
Prior work in the Medicare (https://www.medicare.gov) population showed that new and additional opioid prescribing declined over time at a slower rate in patients with cancer than in patients without cancer, and varied by treatment intent. 8 , 9 , 10 , 11 , 12 , 13 Yet these data may not be generalizable to younger patients with cancer, who differ in opioid use, 14 , 15 cancer‐related pain, 16 cancer prognosis and survivorship needs, and insurance coverage and health care accessibility.
Opioid prescribing must be examined within a clinical context. Patients with cancer remain at risk for opioid‐related harms, including overdose. 17 One study showed a doubling in emergency department visits for opioid overdose in patients with cancer between 2006 and 2015, with the highest risk among those with chronic pain, substance use, and mood disorders. 18 Likewise, persistent opioid use is considered an adverse outcome in cancer survivorship, although its measurement is highly variable across studies. Although continuous daily use portends a higher risk of overdose, 19 , 20 any additional use is more sensitive in identifying long‐term opioid exposure. 8 , 21 Our focus on new opioid prescribing near diagnosis and subsequent progression to additional use fills a gap in the literature.
Correspondingly, there is ongoing concern that spillover from opioid stewardship policies unintentionally reduces access to opioids for patients with metastatic cancer, 22 , 23 , 24 which may lead to the undertreatment of cancer pain. 25 A study of patients with cancer near the end of life found declines in opioid prescriptions and increasing pain‐related emergency visits from 2007 to 2017. 13 In metastatic cancer, understanding the relationship between patient‐reported pain and opioid prescribing is fundamental to examining the impacts of shifting trends. In this study, the inclusion of patient‐reported pain scores and both older and younger adults is novel, which adds clinical context to prescribing trends and increasing generalizability to non‐Medicare patients.
There is a body of literature examining persistent opioid use after surgery for opioid‐naive patients exposed to new opioids perioperatively 26 , 27 ; this is an important outcome in the context of cancer as well. We examine persistent (additional) use trends, given that risk trajectories for harm from long‐term opioids are likely similar in patients with and without a cancer history. 28 Furthermore, we evaluate prescribing trends and patient‐reported pain in patients with metastatic cancer, for whom the necessary treatment of cancer‐related pain is the basic intent of most opioid prescribing.
MATERIALS AND METHODS
Study design and data source
We conducted a retrospective observational study of patients with cancer identified via the YNHHS Tumor Registry. Treatment and prescription data for patients were obtained from the EHR. The Yale University Institutional Review Board determined this study to be exempt.
Study population/cohort
We included patients aged ≥18 years with a first primary solid tumor malignancy, including breast, prostate, lung, colorectal, renal, bladder, liver, brain/nervous system, esophagus, pancreas, and oral cavity/pharynx cancer, diagnosed between 2016 and 2020. We chose 2016 to align with the implementation of Epic (https://www.epic.com) across YNHHS.
We limited our sample to patients with ≥3 encounters at YNHHS within 2 months before through 6 months after the cancer diagnosis to exclude patients for whom YNHHS was not the primary oncology provider. We further restricted the sample to those who were opioid naive, which is defined as no opioid prescription during the 1–4 months before the cancer diagnosis. We implemented a 1‐month washout period to prevent the exclusion of patients who had received opioids as part of their diagnostic assessment, such as for biopsies (Figure S1).
Construction of variables
We identified opioid prescribing by matching EHR prescriptions with the Centers for Disease Control and Prevention's Opioid National Drug Code and Oral MME (morphine milligram equivalent) Conversion File (https://www.cdc.gov/opioids/data‐resources/index.html). We defined new opioid prescribing as receipt of an opioid prescription within 6 months after diagnosis, which corresponds to the time frame used to identify oncological surgery and to the timely completion of most active cancer treatment in the curative setting. For metastatic cancer, we chose to match the 6‐month time frame for consistency in the analyses and to examine access to opioids near the time of cancer diagnosis. We included all outpatient prescriptions (e.g., oral, transdermal, etc.). Additional opioid prescribing was defined as new opioid prescribing plus an additional prescription during months 7 to 9 (Figure S2), which is consistent with literature using a 3‐month window after new opioid use to identify persistent or additional use. 26 , 27 For additional opioids, we excluded patients who died during the first 6 months or did not have any YNHHS encounter during months 7 to 9.
We created two clinical strata of interest: (1) patients with nonmetastatic disease who received oncological surgery (“surgery”), and (2) patients with metastatic disease (“metastatic”). Patients in neither of these strata were classified as “other.” All brain/nervous system tumors were categorized as nonmetastatic. We identified receipt of oncological surgery with International Classification of Diseases, Ninth and Tenth Revisions and Healthcare Common Procedure Coding System codes (Table S1), and restricted the sample to surgeries within 6 months of diagnosis.
Within the metastatic cohort, we evaluated pain scores recorded in EHR flowsheet data during the 6 months after diagnosis and before the first opioid prescription, if given, and selected the highest pain score during that window (Figure S2). We classified patients as having any pain (highest score ≥1 on a 10‐point scale) versus no pain (all pain scores, 0). As a sensitivity analysis, we classified pain scores as 0, 1–3 (mild), and ≥4 (moderate/severe). Patients with no pain scores recorded during the ascertainment period were excluded. As an exploratory analysis, we identified the therapeutic class of the first opioid prescription and death within 6 months to evaluate whether dyspnea or end‐of‐life care was potentially driving prescribing for patients reporting no pain. Therapeutic class is a descriptor based on First Databank (https://fdbhealth.com/) classification linked to EHR prescribing data for each prescription. We examined pain scores over time to explore trends in prescribing thresholds, pain documentation frequency, and reported pain levels.
We included the following patient characteristics and demographic factors: sex, age, race and ethnicity, smoking status, and comorbidity. Race and ethnicity, which were included as social rather than biological constructs, were categorized as Asian, Hispanic, non‐Hispanic Black, non‐Hispanic White, and other/unknown. Smoking status, recorded before or at the time of cancer diagnosis, was categorized as never, ever, or unknown. We assessed comorbidity with the Elixhauser Comorbidity Index. 29 , 30 , 31 , 32 , 33
Statistical analysis
We calculated the frequency and percentage of new and additional opioid prescribing according to clinical strata. We used logistic regression to examine associations between clinical stratum and opioid prescription by adjusting for the covariates described above, as well as an interaction between year of diagnosis and clinical stratum. Subsequently, we calculated the predicted probability of opioid prescribing by year of diagnosis and clinical stratum, including all covariates in the model at their observed values. We used this model to calculate the relative change in new and additional opioid prescribing from 2016 to 2020. For patients with metastatic cancer, we calculated the frequency and percentage of opioid prescribing by pain category, and subsequently the relative change in the predicted probability of opioid prescribing by year and pain category. Given the potential impact of coronavirus disease 2019 (COVID‐19), we calculated the frequency of encounters before and after March 1, 2020.
We performed statistical analysis with SAS software, version 9.4 (SAS Institute, Cary, North Carolina) and Stata, version 17 (StataCorp, College Station, Texas). A p value of less than .05 was considered statistically significant.
RESULTS
The cohort included 10,232 patients. Of these, 4405 patients (43.1%) received surgery for nonmetastatic cancer, and 2158 patients (21.1%) had metastatic cancer (Table 1). The majority were female (61.8%), and 52.6% of patients were ever smokers. The mean age was 62.9 years (standard deviation 12.5). During the 2 months before through 6 months after the cancer diagnosis, patients had a mean of 13 visits and median of 11 visits, both before and after March 1, 2020. Median time to surgery was 41 days; 89.5% of surgeries occurring within 1 year from diagnosis occurred in the first 6 months.
TABLE 1.
Demographic characteristics of the sample.
| Total (N = 10,232) | Received new opioids (n = 6962) | Received additional opioids a (n = 1500) | ||||||
|---|---|---|---|---|---|---|---|---|
| No. | % b | No. | % c | p d | No. | % c | p d | |
| Sex | <.001 | <.001 | ||||||
| Male | 3905 | 38.2 | 2430 | 62.2 | 646 | 32.5 | ||
| Female | 6327 | 61.8 | 4532 | 71.6 | 854 | 22.1 | ||
| Age, years | <.001 | <.001 | ||||||
| 18–39 | 407 | 4.0 | 284 | 69.8 | 85 | 32.1 | ||
| 40–60 | 3757 | 36.7 | 2834 | 75.4 | 679 | 27.6 | ||
| 61–80 | 5351 | 52.3 | 3493 | 65.3 | 689 | 24.0 | ||
| >80 | 717 | 7.0 | 351 | 49.0 | 47 | 18.5 | ||
| Race and ethnicity | <.001 | <.001 | ||||||
| Non‐Hispanic White | 7787 | 76.1 | 5254 | 67.5 | 1075 | 24.4 | ||
| Non‐Hispanic Black | 971 | 9.5 | 686 | 70.7 | 196 | 33.9 | ||
| Hispanic | 702 | 6.9 | 525 | 74.8 | 128 | 28.3 | ||
| Asian | 207 | 2.0 | 164 | 79.2 | 32 | 21.9 | ||
| Other/unknown | 565 | 5.5 | 333 | 58.9 | 69 | 26.6 | ||
| Smoking status | <.001 | <.001 | ||||||
| Never smoker | 3858 | 37.7 | 2786 | 72.2 | 461 | 18.9 | ||
| Ever smoker | 5377 | 52.6 | 3768 | 70.1 | 935 | 30.3 | ||
| Unknown | 997 | 9.7 | 408 | 40.9 | 104 | 32.2 | ||
| Comorbid conditions | <.001 | <.001 | ||||||
| 0 | 5876 | 57.4 | 3869 | 65.8 | 894 | 27.8 | ||
| 1 or 2 | 3124 | 30.5 | 2261 | 72.4 | 429 | 21.9 | ||
| ≥3 | 1232 | 12.0 | 832 | 67.5 | 177 | 26.3 | ||
| Stage | <.001 | <.001 | ||||||
| 0 | 608 | 5.9 | 520 | 85.5 | 50 | 12.0 | ||
| I | 2964 | 29.0 | 2189 | 73.9 | 233 | 11.9 | ||
| II | 1800 | 17.6 | 1137 | 63.2 | 294 | 28.1 | ||
| III | 1697 | 16.6 | 1112 | 65.5 | 310 | 31.2 | ||
| IV | 2546 | 24.9 | 1641 | 64.5 | 560 | 48.0 | ||
| Unknown | 617 | 6.0 | 363 | 58.8 | 53 | 19.3 | ||
| Clinical stratum | <.001 | <.001 | ||||||
| Metastatic | 2158 | 21.1 | 1337 | 62.0 | 462 | 51.5 | ||
| Surgery | 4405 | 43.1 | 4068 | 92.4 | 498 | 13.8 | ||
| Other | 3669 | 35.9 | 1557 | 42.4 | 540 | 40.2 | ||
The denominator for additional opioid use is restricted to patients who had new opioid prescribing, did not die during the first 6 months, and had ≥1 Yale New Haven Health System encounter during months 7–9 (n = 5850).
The percentage column under “Total” shows the column percentage, which reflects the proportion of the total sample that is in each covariate stratum.
The percentage columns under “Received new opioids” and “Received additional opioids” show row percentages, which reflect the proportion of patients in each covariate stratum who received opioids.
p values from a χ2 test between each covariate and receipt of a new or additional opioid.
Overall trends in new and additional opioid prescribing
In the full sample, new opioid prescribing decreased from 72.2% in 2016 to 62.8% in 2020 (Figure 1A), and additional opioid prescribing decreased from 26.9% to 24.8% (Figure 1B). The predicted probability of new opioid prescribing fell from 71.1% (95% confidence interval [CI], 69.4% to 72.9%) in 2016 to 64.6% (95% CI, 62.7% to 66.6%) in 2020, which represents a −9.1% (95% CI, −12.6% to −5.7%) relative change (p < .001) (Tables 2 and 3). Among patients who had received an initial opioid, additional opioid prescribing fell from 27.2% (95% CI, 24.9% to 29.5%) in 2016 to 24.2% (95% CI, 21.6% to 26.8%) in 2020, which represents a −11.0% (95% CI, −23.0% to −1.0%) relative change (p = .07) (Tables 2 and 3).
FIGURE 1.
Trends in new (A) and additional (B) opioid prescribing by clinical stratum with confidence intervals.
TABLE 2.
Predicted probability of new and additional opioid use by clinical stratum.
| Strata | Year of diagnosis, % (95% CI) | ||||
|---|---|---|---|---|---|
| 2016 | 2017 | 2018 | 2019 | 2020 | |
| New opioid use | |||||
| Overall (N = 10,232) | 71.1 (69.4–72.9) | 69.2 (67.4–71.0) | 67.5 (65.8–69.3) | 67.1 (65.3–68.9) | 64.6 (62.7–66.6) |
| Surgery (n = 4405) | 96.0 (94.8–97.2) | 93.4 (91.8–95.0) | 92.3 (90.6–94.0) | 90.0 (88.1–92.0) | 88.6 (86.2–90.9) |
| Other a (n = 3669) | 44.6 (41.1–48) | 43.6 (40.0–47.2) | 43.6 (40.1–47.0) | 41.7 (38.2–45.2) | 38.5 (34.9–42.0) |
| Metastatic (n = 2158) | 65.6 (60.7–70.4) | 63.2 (58.9–67.6) | 57.7 (53.1–62.2) | 63.3 (59.1–67.5) | 60.3 (55.6–65.1) |
| No pain (n = 652) | 61.6 (52.8–70.4) | 51.4 (43.6–59.3) | 37.5 (29.3–45.8) | 43.4 (35.4–51.3) | 36.1 (27.4–44.8) |
| Any pain (n = 1100) | 56.0 (48.9–63.2) | 57.8 (51.5–64.2) | 56.5 (50.3–62.7) | 61.8 (55.9–67.7) | 55.7 (48.6–62.7) |
| Mild pain (n = 245) | 43.3 (29.8–56.8) | 50.1 (37.4–62.8) | 51.7 (37.8–65.6) | 55.2 (41.3–69.0) | 43.6 (27.2–60.0) |
| Moderate/severe pain (n = 855) | 60.9 (52.6–69.1) | 60.3 (53.0–67.7) | 57.8 (50.9–64.7) | 63.4 (57.0–69.9) | 58.7 (51.0–66.5) |
| Additional opioid use b | |||||
| Overall (N = 5850) | 27.2 (24.9–29.5) | 26.8 (24.4–29.2) | 26.1 (23.9–28.4) | 23.7 (21.4–26.0) | 24.2 (21.6–26.8) |
| Surgery (n = 3609) | 14.5 (12.1–16.8) | 15.5 (12.9–18.0) | 13.0 (10.7–15.3) | 12.7 (10.2–15.2) | 12.9 (10.1–15.8) |
| Metastatic (n = 898) | 51.7 (44.7–58.7) | 50.0 (42.8–57.1) | 59.4 (52.1–66.8) | 46.9 (40.2–53.5) | 50.8 (42.9–58.7) |
| Other (n = 1343) | 45.1 (39.4–50.7) | 41.8 (35.9–47.7) | 39.0 (33.6–44.5) | 37.5 (31.8–43.3) | 36.8 (30.6–43.0) |
Patients with nonmetastatic cancer who did not receive surgery.
Analyses were restricted to patients with new opioid use who lived at least 6 months from diagnosis and had a Yale New Haven Health System encounter during months 7–9 postdiagnosis.
TABLE 3.
Relative change in predicted probability of new and additional opioid use by clinical stratum, 2016–2020.
| Strata | Relative change from 2016 to 2020, % (95% CI) | p for relative change | p for comparison of difference in relative change |
|---|---|---|---|
| New opioid use | |||
| Overall (N = 10,232) | −9.1 (−12.6 to −5.7) | <.001 | NA |
| Surgery (n = 4405) | −7.7 (−10.5 to −5.0) | <.001 | Reference |
| Other a (n = 3669) | −13.7 (−24.1 to −3.3) | .01 | .28 |
| Metastatic (n = 2158) | −8.0 (−17.9 to 2.0) | .12 | .97 |
| No pain (n = 652) | −41.4 (−57.8 to −25.1) | <.001 | Reference |
| Any pain (n = 1100) | −0.7 (−18.6 to 17.3) | .94 | .001 |
| Mild pain (n = 245) | 0.8 (−48.9 to 50.5) | .98 | .11 |
| Moderate/severe pain (n = 855) | −3.5 (−21.8 to 14.8) | .71 | .002 |
| Additional opioid use b | |||
| Overall (N = 5850) | −11.0 (−23.0 to 1.0) | .07 | NA |
| Surgery (n = 3609) | −10.7 (−35.3 to 13.8) | .39 | Reference |
| Metastatic (n = 898) | −1.7 (−22.0 to 18.5) | .89 | .58 |
| Other (n = 1343) | −18.3 (−35.4 to −1.2) | .04 | .62 |
Abbreviation: NA, not applicable.
Patients with nonmetastatic cancer who did not receive surgery.
Analyses were restricted to patients with new opioid use who lived at least 6 months from diagnosis and had a Yale New Haven Health System encounter during months 7–9 postdiagnosis.
New and additional opioid prescribing by clinical stratum
Receipt of a new opioid prescription was substantially more common in the surgery than metastatic stratum (χ2 test, p < .001) (Figure 1A). For the surgery stratum, 95.9% and 88.4% of patients in 2016 and 2020 received a new opioid prescription, respectively (Figure 1A). The predicted probability of new opioid prescribing was 96.0% (95% CI, 94.8% to 97.2%) in 2016 and 88.6% (95% CI, 86.2% to 90.9%) in 2020 (Table 2). This −7.7% (95% CI, −10.5% to −5.0%) relative change was statistically significant (p < .001). For the metastatic stratum, new opioid prescribing was 66.2% and 60.3% in 2016 and 2020, respectively (Figure 1A). The predicted probability of new opioid prescribing was 65.6% (95% CI, 60.7% to 70.4%) in 2016 and 60.3% (95% CI, 55.6% to 65.1%) in 2020. This −8.0% (95% CI, −17.9% to 2.0%) relative change was not statistically significant (p = .12). For the other stratum, new opioid prescribing was 44.5% and 39.0% in 2016 and 2020, respectively (Figure 1A). The predicted probability was 44.6% (95% CI, 41.1% to 48.0%) in 2016 and 38.5 (95% CI, 34.9% to 42.0%) in 2020, which represents a −13.7% (95% CI, −24.1% to −3.3%) relative change, which was statistically significant (p = .01) (Tables 2 and 3). The relative changes were not significantly different between strata (p = .97 for the association between the metastatic and surgery strata; p = .28 for other and surgery strata) (Table 3).
Additional opioid prescribing (months 7–9) was substantially less common in the surgery versus metastatic stratum (χ2 test, p < .001; data not shown). In the surgery stratum, 14.7% and 12.4% of patients received an additional opioid in 2016 and 2020, respectively (Figure 1B). Similarly, the predicted probability of an additional opioid was 14.5% (95% CI, 12.1% to 16.8%) in 2016 and 12.9% (95% CI, 10.1% to 15.8%) in 2020 (p = .39 for change over time; Table 2). In the metastatic stratum, receipt of an additional opioid remained stable over time at 51.5% in 2016 and 50.9% in 2020 (Figure 1B). The predicted probability of additional opioid prescribing was 51.7% (95% CI, 44.7% to 58.7%) in 2016 and 50.8% (95% CI, 42.9% to 58.7%) in 2020. This −1.7% (95% CI, −22.0% to 18.5%) relative change was not statistically significant (p = .89) (Table 3). In the other stratum, 44.9% and 36.8% of patients received an additional opioid in 2016 and 2020, respectively (Figure 1B). The predicted probability of additional opioid prescribing fell from 45.1% (95% CI, 39.4% to 50.7%) to 36.8% (95% CI, 30.6% to 43.0%), which represents a −18.3% (95% CI, −35.4% to −1.2%) statistically significant relative change (p = .04) (Tables 2 and 3). The relative changes were not statistically significantly different between strata (p = .58 for metastatic and surgery; p = .62 for other and surgery) (Table 3).
Initial opioid prescribing in the metastatic setting by pain scores
Of the 2158 patients with metastatic cancer, 1752 (81.2%) had a pain score (Table 4). Pain was documented a median of 3.0 times (interquartile range [IQR], 1.0–7.0), which did not differ by year (p = .31). Of the 1752 patients, 37.2% (n = 652) had a pain score of 0. Of the 406 patients who were excluded because of missing pain scores, 402 received an opioid but did not have a pain score recorded before receipt of the first opioid. Thus, by definition, new opioid use was nearly universal in this group.
TABLE 4.
Median pain score by year of diagnosis and opioid receipt among metastatic patients with a pain score reported.
| Year of diagnosis |
All patients (N = 1752) |
Received an opioid (n = 935) |
Did not receive an opioid (n = 817) |
|||
|---|---|---|---|---|---|---|
| Median | IQR | Median | IQR | Median | IQR | |
| All years | 3 | 0–6 | 4 | 0–7 | 2 | 0–5 |
| 2016 | 3 | 0–6 | 3 | 0–7 | 3 | 0–5 |
| 2017 | 3 | 0–6 | 3 | 0–7 | 2 | 0–5 |
| 2018 | 3 | 0–7 | 5 | 0–7 | 2 | 0–6 |
| 2019 | 4 | 0–7 | 5 | 0–7 | 2 | 0–5 |
| 2020 | 4 | 0–6 | 5 | 0–7 | 2 | 0–5 |
| p a | .6 | .03 | .73 | |||
Abbreviation: IQR, interquartile range.
Wilcoxon rank sum test comparing pain scores across year of diagnosis.
In 2016, opioid prescribing was similar for patients with metastatic cancer reporting no pain (62.3%) and those reporting any pain (57.2%). However, in 2020, 35.7% of patients reporting no pain and 55.3% reporting any pain received a new opioid prescription (Figure 2). The predicted probability of a new opioid remained stable from 2016 (56.0%; 95% CI, 48.9% to 63.2%) to 2020 (55.7%; 95% CI, 48.6% to 62.7%) (p = .94 for change over time) (Tables 2 and 3). However, for patients reporting no pain, the predicted probability of a new opioid declined from 61.6% (95% CI, 52.8% to 70.4%) in 2016 to 36.1% (95% CI, 27.4% to 44.8%) in 2020, which represents a −41.4% (95% CI, −57.8% to −25.1%) relative change (p < .001). Relative changes over time were significantly different between patients with metastatic cancer reporting no pain versus any pain (p < .001). In the sensitivity analysis, prescribing remained stable for mild pain (0.8% relative change; 95% CI, −48.9% to 50.5%) (p = .98) and moderate/severe pain (−3.5% relative change; 95% CI, −21.8% to 14.8%) (p = .71). Relative changes over time were significantly different between patients with metastatic cancer reporting no pain versus moderate/severe pain (p < .002) (Table 3).
FIGURE 2.
Trends in new opioid prescribing by pain category in patients with metastatic cancer with confidence intervals.
In the metastatic cohort, the therapeutic class of initial opioid prescriptions was analgesia for 94.6% and cough/cold for 5.4%. For patients with metastatic cancer reporting no pain, analgesia was listed for 86.6% and cough/cold for 13.4%. Of patients with metastatic cancer reporting no pain who received an opioid, 22.7% died within 6 months of diagnosis, which is similar to the overall metastatic stratum (21.0%) and to all patients with metastatic cancer who received an opioid (25.8%). Pain scores were stable over time (p = .60 across years) (Table 4). For those receiving opioids, the median highest pain score increased from 3.0 (IQR, 0.0–7.0) in 2016 to 5.0 (IQR, 0.0–7.0) in 2020, a statistically significant change (p = .02), which reflects the migration of patients reporting no pain out of this group. Among those not receiving opioids, the median pain score was 3.0 (IQR, 0.0–5.0) in 2016 and 1.5 (IQR, 0.0–5.0) in 2020, and was not significantly different across years (p = .72).
DISCUSSION
Our analyses showed a 9.1% relative decline in the predicted probability of new and 11.0% relative decline in the predicted probability of additional opioid prescribing in adult patients with cancer in a large Connecticut health system from 2016 to 2020, which coincided with efforts to reduce unnecessary opioid prescribing for nonmalignant pain. We found that for patients with metastatic cancer, pain scores were stable over time, while new opioid prescribing decreased for patients with metastatic cancer reporting no pain.
When examined by clinical stratum, patients with metastatic cancer and patients undergoing oncological surgery experienced relative declines in new opioid prescribing that were similar in direction and magnitude (approximately 8%), although significant in the surgical but not the metastatic group, which had lower initial prescribing rates and a smaller sample size. On the other hand, the predicted probability of additional opioid use was 14.5% in the surgery group in 2016 and did not significantly decrease over time. Generally, “persistent use” can be defined as (1) additional use requiring ≥1 opioid fills in some period after the index date, or (2) continuous use requiring several prescriptions over time (e.g., 90 days of continuous use). Studies using additional use definitions 34 , 35 generally identify higher rates of persistent opioid use than those using measures of continuous use. 10 , 12 , 36 In this health system, despite near‐universal initial prescribing after oncological surgery, rates of additional opioid prescribing remained low for patients undergoing surgery. For patients with metastatic cancer, additional use was stable (approximately −2% relative change).
These findings differ from national trends for new opioid use from 2012 to 2017 in elderly patients undergoing oncological surgery, for whom relative declines were less than 3%, and in elderly patients with metastatic disease (approximately 9% relative decline). These findings are comparable to national trends from 2012 to 2017 for additional use in elderly patients with metastatic cancer. 8 In that study, 50.0% of elderly patients with metastatic cancer received initial opioids, and 31.7% received additional opioids in months 3–6, with an 11.5% relative decline in early additional use over time. That analysis was not able to evaluate whether declines in opioid prescribing translated into suboptimal pain management for patients with metastatic cancer.
The ability to include granular pain score data for patients with metastatic cancer is an advantage of this study. By using EHR flowsheet pain scores, we found that opioid prescribing remained stable over the study period for patients with any pain but declined by 41.4% for those reporting no pain. This may represent a significant shift in the opioid prescribing culture and a growing attention to opioid stewardship, given concern for diversion of pills into the community 37 and other potential harms. With the institutional focus on opioid stewardship, a 36.1% rate of opioid prescribing in 2020 for those reporting no pain was unexpected. Pain score data do not identify other indications for opioid prescribing, such as dyspnea or end‐of‐life care. We found that 22.7% of patients reporting no pain who received an opioid died within 6 months, which may suggest that the opioid was given for end‐of‐life symptom control. Furthermore, this study relies on EHR pain scores, which do not include pain that was not reported or reported but not recorded. Individually, pain scores are captured at discrete points, and do not assess intermittent or incidental pain. If pain documentation used in this study underestimated the pain experienced by patients, the trend in reduced prescribing to patients with metastatic cancer would be concerning. Pain is a particularly complex symptom, and pain score data may underestimate individual patient‐level experience or impact of pain. Although flowsheet data are generally entered by nursing staff obtaining vital signs, physicians or advanced practice providers prescribe the opioids; numerical pain data may not reflect the clinician’s more detailed pain history. Nevertheless, the median highest pain score of 3.0 in the metastatic cohort is reassuring in that patients reporting moderate/severe cancer‐related pain received needed opioids. On the basis of our findings, we speculate that in response to the increased scrutiny and a desire to eliminate unneeded prescriptions, prescribers attempted to reduce prescribing for those who may not require opioid analgesia, which is reflected in significant declines in routine, near‐universal prescribing postoperatively and in prescribing to patients with metastatic cancer not reporting pain.
There are several limitations to our study. Some patients may have received opioids elsewhere, for which pain scores may indicate pain control on existing therapy, which thus deflates pain score trends. Unobserved prior opioid prescribing could misclassify non–opioid‐naive patients as experiencing new persistent opioid use, or conversely, unobserved later prescriptions could underestimate additional opioid prescribing. We limited our study to patients receiving ≥3 encounters at YNHHS to avoid including patients categorically seen only for a second opinion; although we believe that overall this increases the generalizability of our results, nonetheless there may be some remaining misclassification of new use that we do not expect would result in significant bias. We also did not examine inpatient prescribing. We used a binary definition of opioid use, for which there is a precedent in the literature. 38 , 39 Although this does not reflect high‐dose usage, 19 , 40 it is useful in understanding physician prescribing trends for new and additional opioid use. We used metastatic cancer stage as a proxy for palliative intent. We did not evaluate for early recurrence or progression. Furthermore, we could not distinguish surgical from other cancer pain in the surgery cohort, such as from radiation toxicity or disease burden. Radiation toxicity, which may be delayed and resolve slowly over time, could account for additional opioid use for a subset of curative‐intent patients. 41 Nevertheless, the large majority of those receiving an initial opioid prescription in the surgery stratum did not demonstrate persistent use. We examined physician prescribing; the prescription fill and medication usage rates are unknown. The impact of any one regulatory, ecological, or cultural shift, such as the COVID‐19 pandemic, 42 , 43 , 44 cannot be determined from these data, given the complicated interplay between many important factors. 2 These and other shifts have dramatically affected cancer care and opioid prescribing, and are part of the ongoing changing environment in which patients are receiving care. We examine changes over time and impact on patients with cancer; as such, we cannot determine causation. Finally, our findings are specific to Connecticut, a state both highly ranked in health care quality and actively facing a serious opioid overdose crisis; generalizability to other US regions is qualified. 45 , 46
In conclusion, in the context of widespread policy changes for opioid prescribing and stewardship, we found decreasing trends in new and additional opioid prescribing for patients with cancer in a large health care system. Nevertheless, most patients continued to receive opioids early in treatment, particularly if undergoing oncological surgery. Among patients with metastatic cancer, opioid prescribing remained stable for those with pain but declined steeply for those reporting no pain. The clinical impact of these changes remains undetermined, and our findings raise questions about the completeness of pain reporting in the EHR, the appropriate level of judiciousness for opioid prescribing in cancer care, and the adequacy of access to opioids for patients with metastatic cancer. Further research is needed to characterize optimal opioid pain management in cancer.
AUTHOR CONTRIBUTIONS
Laura Van Metre Baum: Conceptualization; investigation; writing—original draft; visualization; writing—review editing; methodology; validation; project administration. Pamela R. Soulos: Conceptualization; funding acquisition; writing—review and editing; visualization; validation; formal analysis; project administration; supervision; data curation; investigation; software. Madhav KC: Conceptualization; investigation; formal analysis; writing—review and editing; data curation; software; project administration. Molly M. Jeffery: Conceptualization; investigation; funding acquisition; writing—review and editing; methodology. Kathryn J. Ruddy: Writing—review and editing; funding acquisition; investigation; methodology. Catherine C. Lerro: Methodology; writing—review and editing; funding acquisition; project administration. Hana Lee: Supervision; formal analysis; methodology; writing—review and editing. David J. Graham: Writing—review and editing; conceptualization; funding acquisition. Donna R. Rivera: Conceptualization; writing—review and editing; funding acquisition. Mark Liberatore: Funding acquisition; conceptualization; writing—review and editing. Michael S. Leapman: Writing—review and editing; funding acquisition; conceptualization. Vikram Jairam: Conceptualization; funding acquisition; writing—review and editing; methodology. Michaela A. Dinan: Methodology; conceptualization; writing—review and editing; supervision. Cary P. Gross: Supervision; conceptualization; investigation; funding acquisition; writing—original draft; writing—review and editing; methodology; resources. Henry S. Park: Resources; supervision; writing—review and editing; methodology; validation; visualization; conceptualization; investigation; funding acquisition.
CONFLICT OF INTEREST STATEMENT
Molly M. Jeffery has received funding from the Food and Drug Administration for studies of opioid use for acute pain (to the institution), National Institutes of Health National Institute on Drug Abuse for studies of treatment of opioid use disorder (to the institution), National Institutes of Health Addiction Health Services Research Conference for unrelated studies (to the institution), American Cancer Society for unrelated studies (to the institution), and National Institutes of Health National Center for Advancing Translational Sciences for studies of unintended prolonged opioid use (to the institution); her spouse owns (purchased) shares in Goodness Growth Holdings. Vikram Jairam has received funding from an American Cancer Society institutional research grant. Cary P. Gross has received research funding from the National Comprehensive Cancer Network (NCCN) Foundation (funds provided to the NCCN by AstraZeneca), as well as funding from Johnson & Johnson to help devise and implement new approaches to sharing clinical trial data. Henry S. Park has received funding from RefleXion Medical and Merck (to the institution) and honoraria from AstraZeneca (consultant and advisory board), Bristol‐Myers Squibb (speaking), Galera Therapeutics (advisory board), G1 Therapeutics (speaking), RefleXion Medical (speaking and consultant), and Regeneron Pharmaceuticals (advisory board). The other authors declare no conflicts of interest.
Supporting information
Supplementary Material
ACKNOWLEDGMENTS
We acknowledge the contributions of Corinne Woods, Kate Gelperin, and Amy Ho from the US Food and Drug Administration (FDA). This publication was supported by the FDA of the US Department of Health and Human Services (HHS) as part of a financial assistance award (Centers of Excellence in Regulatory Science and Innovation; U01FD005938) totaling $369,769. The contents are those of the authors and do not necessarily represent the official views of, nor an endorsement by, the FDA/HHS or the US Government. Some of the authors are employees of the FDA; however, other officials at the FDA had no role in the design and conduct of the study; the collection, analysis, and interpretation of the data; the preparation of the manuscript; or the decision to submit the manuscript for publication. The manuscript was subject to administrative review before submission but the content was not altered by this review.
Baum LVM, Soulos PR, KC M, et al. Opioid prescribing trends and pain scores among adult patients with cancer in a large health system. Cancer. 2025;e70027. doi: 10.1002/cncr.70027
This study was presented as an abstract at the American Society of Clinical Oncology Annual Meeting; May 31–June 4, 2024; Chicago, Illinois.
DATA AVAILABILITY STATEMENT
The data underlying this article were provided by the Yale New Haven Health System via the Joint Data Analytics Team from the Epic data system, and were deemed exempt by the Yale University Institutional Review Board. Thus, given patient‐level protected health information and the proprietary nature of the data, data will not be shared.
REFERENCES
- 1. Lyden J, Binswanger IA. The United States opioid epidemic. Semin Perinatol. 2019;43(3):123‐131. doi: 10.1053/j.semperi.2019.01.001 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2. Lee B, Zhao W, Yang K‐C, Ahn YY, Perry BL. Systematic evaluation of state policy interventions targeting the US opioid epidemic, 2007–2018. JAMA Netw Open. 2021;4(2):e2036687. doi: 10.1001/jamanetworkopen.2020.36687 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3. Duensing K, Twillman R, Ziegler S, et al. An examination of state and federal opioid analgesic and continuing education policies: 2016–2018. J Pain Res. 2020;13:2431‐2442. doi: 10.2147/jpr.s267448 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4. Beaudoin FL, Banerjee GN, Mello MJ. State‐level and system‐level opioid prescribing policies: the impact on provider practices and overdose deaths, a systematic review. J Opioid Manag. 2016;12:109‐118. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5. Mauri AI, Townsend TN, Haffajee RL. The association of state opioid misuse prevention policies with patient‐ and provider‐related outcomes: a scoping review. Milbank Q. 2020;98(1):57‐105. doi: 10.1111/1468-0009.12436 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6. Bachhuber MA, Nash D, Southern WN, et al. Effect of changing electronic health record opioid analgesic dispense quantity defaults on the quantity prescribed: a cluster randomized clinical trial. JAMA Netw Open. 2021;4:e217481. doi: 10.1001/jamanetworkopen.2021.7481 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7. Montoy JCC, Coralic Z, Herring AA, Clattenburg EJ, Raven MC. Association of default electronic medical record settings with health care professional patterns of opioid prescribing in emergency departments: a randomized quality improvement study. JAMA Intern Med. 2020;180(4):487‐493. doi: 10.1001/jamainternmed.2019.6544 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8. Van Metre Baum L, KC M, Soulos PR, et al. Trends in new and persistent opioid use in older adults with and without cancer. J Natl Cancer Inst. 2023;116(2):316‐323. doi: 10.1093/jnci/djad206 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9. Jairam V, Yang DX, Pasha S, et al. Temporal trends in opioid prescribing patterns among oncologists in the Medicare population. J Natl Cancer Inst. 2021;113(3):274‐281. doi: 10.1093/jnci/djaa110 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10. Salz T, Lavery JA, Lipitz‐Snyderman AN, et al. Trends in opioid use among older survivors of colorectal, lung, and breast cancers. J Clin Oncol. 2019;37(12):1001‐1011. doi: 10.1200/jco.18.00938 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11. Agarwal A, Roberts A, Dusetzina SB, Royce TJ. Changes in opioid prescribing patterns among generalists and oncologists for Medicare Part D beneficiaries from 2013 to 2017. JAMA Oncol. 2020;6(8):1271‐1274. doi: 10.1001/jamaoncol.2020.2211 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12. Roberts A, Fergestrom N, Neuner J, Winn AN. New‐onset persistent opioid use following breast cancer treatment in older adult women. Cancer. 2019;126(4):814‐822. doi: 10.1002/cncr.32593 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13. Enzinger AC, Ghosh K, Keating NL, Cutler DM, Landrum MB, Wright AA. US trends in opioid access among patients with poor prognosis cancer near the end‐of‐life. J Clin Oncol. 2021;39(26):2948‐2958. doi: 10.1200/jco.21.00476 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14. Jairam V, Yang DX, Verma V, Yu JB, Park HS. National patterns in prescription opioid use and misuse among cancer survivors in the United States. JAMA Netw Open. 2020;3(8):e2013605. doi: 10.1001/jamanetworkopen.2020.13605 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15. Chino F, Kamal A, Chino J. Incidence of opioid‐associated deaths in cancer survivors in the United States, 2006–2016: a population study of the opioid epidemic. JAMA Oncol. 2020;6(7):1100‐1102. doi: 10.1001/jamaoncol.2020.0799 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16. Green CR, Hart‐Johnson T. Cancer pain: an age‐based analysis. Pain Med. 2010;11(10):1525‐1536. doi: 10.1111/j.1526-4637.2010.00957.x [DOI] [PubMed] [Google Scholar]
- 17. Baum L. Response to opioid epidemic or pain crisis? JCO Oncol Pract. 2020;16(3):146. doi: 10.1200/jop.19.00706 [DOI] [PubMed] [Google Scholar]
- 18. Jairam V, Yang DX, Yu JB, Park HS. Emergency department visits for opioid overdoses among patients with cancer. J Natl Cancer Inst. 2019;112(9):938‐943. doi: 10.1093/jnci/djz233 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19. Henry SG. Evaluating the risks of opioid use for chronic pain: moving beyond overdose. J Gen Intern Med. 2016;31(5):453‐454. doi: 10.1007/s11606-016-3659-5 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20. Liang Y, Turner BJ. Assessing risk for drug overdose in a national cohort: role for both daily and total opioid dose? J Pain. 2015;16(4):318‐325. doi: 10.1016/j.jpain.2014.11.007 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21. Thiels CA, Habermann EB, Hooten WM, Jeffery MM. Chronic use of tramadol after acute pain episode: cohort study. BMJ. 2019;365:l1849. doi: 10.1136/bmj.l1849 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22. Bulls HW, Bell LF, Orris SR, et al. Exemptions to state laws regulating opioid prescribing for patients with cancer‐related pain: a summary. Cancer. 2021;127(17):3137‐3144. doi: 10.1002/cncr.33639 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23. Page R, Blanchard E. Opioids and cancer pain: patients’ needs and access challenges. J Oncol Pract. 2019;15(5):229‐231. doi: 10.1200/jop.19.00081 [DOI] [PubMed] [Google Scholar]
- 24. Townsend TN, Salz T, Haffajee RL, Caram ME, Chino F, Bohnert AS. Has declining opioid dispensing to cancer patients been tailored to risk of opioid harms? J Pain Symptom Manage. 2022;63(2):179‐188. doi: 10.1016/j.jpainsymman.2021.09.014 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25. LeBaron VT, Camacho F, Balkrishnan R, Yao NA, Gilson AM. Opioid epidemic or pain crisis? Using the Virginia All Payer Claims Database to describe opioid medication prescribing patterns and potential harms for patients with cancer. J Oncol Pract. 2019;15(12):e997‐e1009. doi: 10.1200/jop.19.00149 [DOI] [PubMed] [Google Scholar]
- 26. Jivraj NK, Raghavji F, Bethell J, et al. Persistent postoperative opioid use: a systematic literature search of definitions and population‐based cohort study. Anesthesiology. 2020;132(6):1528‐1539. doi: 10.1097/aln.0000000000003265 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27. Delaney LD, Waljee JF. New persistent opioid use: definitions and opportunities. Reg Anesth Pain Med. 2021;46(2):97‐98. doi: 10.1136/rapm-2020-102121 [DOI] [PubMed] [Google Scholar]
- 28. Merlin JS, Black AC, Midboe AM, et al. Long‐term opioid therapy trajectories and overdose in patients with and without cancer. BMJ Oncol. 2023;2(1):e000023. doi: 10.1136/bmjonc-2022-000023 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29. Elixhauser A, Steiner C, Harris DR, Coffey RM. Comorbidity measures for use with administrative data. Med Care. 1998;36(1):8‐27. doi: 10.1097/00005650-199801000-00004 [DOI] [PubMed] [Google Scholar]
- 30. Lieffers JR, Baracos VE, Winget M, Fassbender K. A comparison of Charlson and Elixhauser comorbidity measures to predict colorectal cancer survival using administrative health data. Cancer. 2011;117(9):1957‐1965. doi: 10.1002/cncr.25653 [DOI] [PubMed] [Google Scholar]
- 31. Mehta HB, Sura SD, Adhikari D, et al. Adapting the Elixhauser Comorbidity Index for cancer patients. Cancer. 2018;124(9):2018‐2025. doi: 10.1002/cncr.31269 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32. Sharma N, Schwendimann R, Endrich O, Ausserhofer D, Simon M. Comparing Charlson and Elixhauser comorbidity indices with different weightings to predict in‐hospital mortality: an analysis of national inpatient data. BMC Health Serv Res. 2021;21(1):13. doi: 10.1186/s12913-020-05999-5 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33. Yurkovich M, Avina‐Zubieta JA, Thomas J, Gorenchtein M, Lacaille D. A systematic review identifies valid comorbidity indices derived from administrative health data. J Clin Epidemiol. 2015;68(1):3‐14. doi: 10.1016/j.jclinepi.2014.09.010 [DOI] [PubMed] [Google Scholar]
- 34. Lee JS, Hu HM, Edelman AL, et al. New persistent opioid use among patients with cancer after curative‐intent surgery. J Clin Oncol. 2017;35(36):4042‐4049. doi: 10.1200/jco.2017.74.1363 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35. Falcone M, Luo C, Chen Y, et al. Risk of persistent opioid use following major surgery in matched samples of patients with and without. Cancer Epidemiol Biomarkers Prev. 2020;29(11):2126‐2133. doi: 10.1158/1055-9965.epi-20-0628 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36. Vitzthum LK, Riviere P, Sheridan P, et al. Predicting persistent opioid use, abuse, and toxicity among cancer survivors. J Natl Cancer Inst. 2019;112(7):720‐727. doi: 10.1093/jnci/djz200 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 37. Baum LVM, Bruzelius E, Kiang MV, Humphreys K, Basu S, Baum A. Analysis of unused prescription opioids and benzodiazepines remaining after death among Medicare decedents. Drug Alcohol Depend. 2021;219:108502. doi: 10.1016/j.drugalcdep.2020.108502 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38. Dasgupta N, Wang Y, Bae J, et al. Inches, centimeters, and yards: overlooked definition choices inhibit interpretation of morphine equivalence. Clin J Pain. 2021;37(8):565‐574. doi: 10.1097/ajp.0000000000000948 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 39. Morphine Milligram Equivalents: Current Applications and Knowledge Gaps, Research Opportunities, and Future Directions. Food and Drug Administration Public Workshop; 2021. [Google Scholar]
- 40. Webster LR. Risk factors for opioid‐use disorder and overdose. Anesth Analg. 2017;125(5):1741‐1748. doi: 10.1213/ane.0000000000002496 [DOI] [PubMed] [Google Scholar]
- 41. Majeed H, Gupta V. Adverse Effects of Radiation Therapy. StatPearls Publishing; 2025. [PubMed] [Google Scholar]
- 42. Riera R, Bagattini ÂM, Pacheco RL, Pachito DV, Roitberg F, Ilbawi A. Delays and disruptions in cancer health care due to COVID‐19 pandemic: systematic review. JCO Glob Oncol. 2021;7:311‐323. doi: 10.1200/go.20.00639 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 43. Lee B, Yang K‐C, Kaminski P, et al. Substitution of nonpharmacologic therapy with opioid prescribing for pain during the COVID‐19 pandemic. JAMA Netw Open. 2021;4(12):e2138453. doi: 10.1001/jamanetworkopen.2021.38453 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 44. Rikin S, Perez HR, Zhang C, et al. Changes in outpatient opioid prescribing during the COVID‐19 pandemic: an interrupted time series analysis. J Prim Care Community Health. 2022;13:21501319221076926. doi: 10.1177/21501319221076926 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 45. The Connecticut Opioid REsponse (CORE) Initiative : Advising the State of Connecticut on Evidence‐Based Spending of Opioid Settlement Funds. Yale Program in Addiction Medicine; 2025. [Google Scholar]
- 46. Andrews E. US News: CT Is Third Best State for Healthcare. Connecticut Health Policy Project; 2025. [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Supplementary Material
Data Availability Statement
The data underlying this article were provided by the Yale New Haven Health System via the Joint Data Analytics Team from the Epic data system, and were deemed exempt by the Yale University Institutional Review Board. Thus, given patient‐level protected health information and the proprietary nature of the data, data will not be shared.