Patterns of Opioid Prescription, Use, and Costs Among Patients With Advanced Cancer and Inpatient Palliative Care Between 2008 and 2014

Abstract

PURPOSE:

An understanding of opioid prescription and cost patterns is important to optimize pain management for patients with advanced cancer. This study aimed to determine opioid prescription and cost patterns and to identify opioid prescription predictors in patients with advanced cancer who received inpatient palliative care (IPC).

MATERIALS AND METHODS:

We reviewed data from 807 consecutive patients with cancer who received IPC in each October from 2008 through 2014. Patient characteristics; opioid types; morphine equivalent daily dose (MEDD) in milligrams per day of scheduled opioids before, during, and after hospitalization; and in-admission opioid cost per patient were assessed. We determined symptom changes between baseline and follow-up palliative care visits and the in-admission opioid prescription predictors.

RESULTS:

A total of 714 (88%) of the 807 patients were evaluable. The median MEDD per patient decreased from 150 mg/d in 2008 to 83 mg/d in 2014 (P < .001). The median opioid cost per patient decreased and then increased from $22.97 to $40.35 over the 7 years (P = .03). The median MEDDs increased from IPC to discharge by 67% (P < .001). The median Edmonton Symptom Assessment Scale pain improvement at follow-up was 1 (P < .001). Younger patients with advanced cancer (odds ratio [OR[, 0.95; P < . 001) were prescribed higher preadmission MEDDs (OR, 1.01; P < .001) more often in the earlier study years (2014 v 2009: OR, 0.18 [P = .004] v 0.30 [P = .02]) and tended to use high MEDDs (> 75 mg/d) during hospitalization.

CONCLUSION:

The MEDD per person decreased from 2008 to 2014. The opioid cost per patient decreased from 2008 to 2011 and then increased from 2012 to 2014. Age, prescription year, and preadmission opioid doses were significantly associated with opioid doses prescribed to patients with advanced cancer who received IPC.

INTRODUCTION

Pain is one of the most distressing symptoms experienced by patients with cancer.¹ Approximately 60% to 90% of patients with cancer endure moderate to severe pain.¹ Adequate and prompt pain management is essential for them to achieve better quality of life and treatment outcomes.² The increased awareness of the need to have prompt treatment of pain was facilitated by the WHO declaration that “pain relief [is] a basic human right.”^3,4 The WHO also recommends opioid therapy as an effective response to moderate or severe cancer pain for most patients.⁵ In the United States, the reduction of the stigma around opioid medications in the early 1990s partially fueled the escalation in opioid prescriptions from 1990 through 2010.^6,7 It is estimated that 75% of patients with cancer in the United States are prescribed opioids for cancer pain management.^8,9

The palliative care (PC) community is a strong advocate for equal access to pain treatment.⁹ Yet, most patients with cancer who have intractable pain remain undertreated because of opioid underuse.^{9, 6,10-12} The Centers for Disease Control and Prevention reported 63,632 drug overdose deaths in 2016; 42,249 (66.4%) involved an opioid.¹³ Recent studies have found that the risk of opioid misuse was also an increasing concern in approximately 20% of patients with cancer who used opioids.^14,15 Therefore, assessment of the opioid prescription pattern will facilitate a better understanding of optimal opioid prescription for cancer pain management and thereby reduce the risk of opioid underuse and misuse among patients with cancer.^6,8

Opioid costs play another role in shaping the opioid prescription pattern in the cancer care continuum. From 1995 to 2010, opioid sales in the United States increased by 400% and accounted for 80% of the world’s opioid supply.^16,17 Opioid medications also accounted for 20.1% of total US pain medication costs from 2000 to 2007.¹⁸ The high costs of opioids impose a significant economic burden on patients with cancer. Tanco et al¹⁹ reported that patients with cancer could face payment denials when insurance companies considered the opioid use inappropriate.

However, few studies have compared the opioid prescription pattern before, during, and after hospitalization; evaluated opioid costs; and determined the opioid prescription predictors in patients with advanced cancer who received inpatient PC (IPC). Given the effect of opioids on health outcomes and quality of life in patients with cancer, a comprehensive evaluation of the opioid prescription and cost patterns and of opioid prescription predictors could inform the future clinical practice of opioid prescription and thereby optimize pain management. This study sought to determine the opioid prescription and cost patterns and to identify opioid prescription predictors in patients with advanced cancer who received IPC in a comprehensive cancer center.

MATERIALS AND METHODS

This retrospective study was approved by the institutional review board of The University of Texas MD Anderson Cancer Center. We reviewed the medical records of consecutive patients with cancer who received IPC in each October from 2008 through 2014 at MD Anderson. Patients were eligible if they had a diagnosis of advanced cancer (defined as metastatic, locally advanced, or locally recurrent incurable cancer), were referred to IPC, were older than age 18 years, and were prescribed scheduled opioids during hospitalization. The rationale for choosing October as the evaluation month was that it had less variability in opioid use and prescription because the PC fellows at MD Anderson had been trained, no major holidays occurred, and the patients’ census was consistent.²⁰ Scheduled opioids included long- and short-acting oral or intravenous (IV) formulations prescribed on a scheduled basis for at least a 24-hour period by the MD Anderson PC team.

We obtained patient sociodemographic and clinical characteristics, symptoms, opioid types, and opioid doses dispensed before, during, and after hospitalization by reviewing electronic medical records. The MD Anderson pharmacy database provided the average wholesale prices (AWPs) of opioids.

We collected patient age, sex, race/ethnicity, education, marital status, primary caregivers, insurance types, alcoholism risk (CAGE [cut down–annoyed–guilty–eye opener] questionnaire), illicit drug use, smoking history, family cancer history, cancer types, cancer treatments, and IPC referral reasons. The length of stay (LOS) in hospital, duration from hospital admission to the initial IPC, and duration from the initial IPC to the first follow-up PC visit after discharge were collected for each patient.

The Edmonton Symptom Assessment Scale (ESAS) was used to assess pain, fatigue, nausea, depression, anxiety, drowsiness, dyspnea, appetite, sleep, and feeling of well-being; it is a part of routine care at MD Anderson PC team. We reviewed ESAS pain scores at the first IPC consultations during admission and at the first follow-up PC visits after discharge. The ESAS, a highly validated and reliable numerical scale (range, 0 to 10; 0 = no symptom, and 10 = worst possible severity), quantifies 10 symptoms that patients with cancer commonly experience 24 hours before assessment.^21,22 We also reviewed performance status (Eastern Cooperative Oncology Group performance scale) and delirium (Memorial Delirium Assessment Scale) data at those time points.

We calculated the sum of the morphine equivalent daily dose (MEDD) in milligrams per day of all opioids prescribed to a given patient as opioid daily dose per patient using the following standardized MEDD calculation equation: (strength per unit × [number of units/days of supply] × MEDD conversion factor = MEDD per day).²³ Opioid costs were calculated using the methods introduced by Curry et al,²⁰ which included opioid cost per patient, cost per MEDD per patient, and cost per MEDD per opioid (Data Supplement, online only); all cost values were normalized to 2008 US dollars.²⁰ For 221 patients whose opioid AWPs were not available, we imputed their opioid-specific AWPs using the AWPs of the same opioid medications in the same given year.

We used median and interquartile range (IQR) to summarize continuous variables (eg, age) and frequencies and proportions to present categoric variables (eg, sex) of patient sociodemographic and clinical characteristics. The Mann-Whitney U test was used to examine differences of continuous variables between two PC visits and weight changes between baseline and 3 months afterward. The χ² test/Fisher’s exact test was applied to assess the variations of categoric variables between two PC visits. The Spearman correlation test was used to analyze the correlations of MEDD per patient among before, during, and after hospitalization. The trends of MEDD per patient, opioid cost per patient, and cost per MEDD per patient from 2008 to 2014 were determined using the Jonckheer-Terpstra test. Multivariable logistic models were used to explore opioid prescription predictors during hospitalization, including age, sex, race/ethnicity, the highest education level, cancer types, preadmission MEDDs, and opioid prescription year. The MEDD per patient during hospitalization was divided into high opioid daily dose (MEDD per patient > 75 mg/d) and low opioid daily dose (≤ 75 mg/d) according to the median value. The covariates were selected on the basis of their clinical and statistical significances (P < .10) to the outcome in the univariable analysis. We computed statistical analyses using SAS 9.3 (SAS Institute, Cary, NC) and reported two-tailed tests with a significance level of less than .05.

RESULTS

We analyzed 714 patients (88%) with opioid use of 807 eligible patients. The median age of the evaluable patients was 57 years (IQR, 47-65 years; Table 1). The patients were primarily women (55%), white (61%), married (69%), and referred to IPC for symptom management (91%). Furthermore, 12% of study patients had a history of illicit drug use, 47% were smokers, 5% had a potential risk for alcoholism (CAGE positive), and 64% had at least one family member with a cancer diagnosis. No significant differences in patient demographics and clinical factors were detected from 2008 through 2014.

TABLE 1.

Patient Demographic and Clinical Characteristics and Symptoms at Baseline (N = 174)

Patients reported moderate to high ESAS symptom scores at the initial IPC during hospitalization, including median scores of 6.0 (IQR, 4.0-8.0) for pain, 6.0 (IQR, 4.0-8.0) for fatigue, 5.0 (IQR, 3.0-8.0) for appetite, 5.0 (IQR, 4.0-7.0) for well-being, and 5.0 (IQR, 1.0-7.0) for sleep. The median hospital length of stay was 5 days (IQR, 3-9 days), the median interval from hospitalization to the initial IPC was 1 day (IQR, 1-4 days), and the median duration from the initial IPC to first follow-up PC visit was 13 days (IQR, 7-21 days). The ESAS scores for pain (6.0 v 5.0; P < .001), depression (2.0 v 0; P < .001), anxiety (3.0 v 0; P < .001), and financial distress (1.0 v 0; P = .022) improved for patients compared with their baseline scores (data not shown).

Patients used oxycodone, methadone, morphine, fentanyl, hydromorphone, oxymorphone, hydrocodone, codeine, and tramadol from 2008 to 2014 (Appendix Fig A1, online only). The median MEDD per patient decreased from 150 mg/d (IQR, 63.8-316.3 mg/d) in 2008 to 83 mg/d (IQR, 36.4-150.0 mg/d) in 2014 (P = .001; Table 2). The median MEDD per patient before hospitalization decreased from 60 mg/d (IQR, 0-150.0 mg/d) in 2008 to 50 mg/d (IQR, 0-120.0 mg/d) in 2014 (P = .09), and the median MEDD per patient at discharge decreased from 75 mg/d (IQR, 30-180 mg/d) in 2008 to 60 mg/d (IQR, 30-120 mg/d) in 2014 (P = .56). Compared with preadmission opioid prescription, the MEDDs for patients with advanced cancer increased from IPC to discharge by 66.7% (P < .001). The MEDD prescribed at hospital discharge was 20% lower than during hospitalization (P < .001; Appendix Fig A1, online only).

TABLE 2.

Opioid Use and Cost Changes (2008-2014)

The trends of opioid cost per patient (P = .03) and cost per MEDD per patient (P = .02) decreased from 2008 to 2011 and increased from 2012 to 2014 (Table 2). From 2008 through 2011, the median opioid cost per patient ($22.97 [IQR, $8.19-$145.13] to $9.85 [IQR, $2.19-$87.02]), and median cost per MEDD per patient ($0.17 [IQR, $0.09-$0.94] to $0.13 [$0.05-$1.44]) decreased gradually. However, between 2012 and 2014, these two cost indicators increased and reached their highest points in 2014. In 2014, the median opioid cost per patient ($40.35 [IQR, $3.68-$258.98]) and median cost per MEDD per patient ($0.84 [IQR, $0.06-$5.16]) exceeded their values in 2008. Furthermore, trends of decreasing median costs per MEDD per opioid were observed for fentanyl ($8.12 in 2008 to $7.29 in 2014; P = .01); methadone ($0.08 in 2008 to $0.06 in 2014; P < .001); hydromorphone ($0.17 in 2008 to $0.04 in 2014; P < .001); and the combination of codeine, hydrocodone, oxymorphone, and tramadol ($2.30 in 2008 to $1.98 in 2014; P = .80; data not shown). However, the median cost per MEDD (data not shown) increased during the 7 years for morphine ($0.01 in 2008 to $1.34 in 2014; P = .004) and oxycodone ($0.27 in 2008 to $0.31 in 2014; P = .24, data not shown).

The multivariable logistic regression model suggested that age negatively predicted high MEDD per patient (OR, 0.95; 95% CI, 0.92 to 0.97; P < .001; Table 3). MEDD per patient before admission (OR, 1.01; 95% CI, 1.01 to 1.02; P < .001) positively predicted high MEDD per patient. Compared with 2008, the odds of using high MEDD per patient between 2009 and 2014 decreased (2009 v 2014: OR, 0.30 [95% CI, 0.11 to 0.85; P = .02] v 0.18 [95% CI, 0.06 to 0.59; P = .004]). In the univariable analysis, pain (OR, 1.12; P < .001), anxiety (OR, 1.03; P = .06), and depression (OR, 1.03; P = .02) positively predicted high in-admission MEDDs. However, such positive associations did not persist in the multivariable logistic regression model.

TABLE 3.

Univariable and Multivariable Logistic Regression Models of Predictors of Opioid Use in Hospitalization (N = 714)

DISCUSSION

Optimal pain control is one of the essential elements to improve quality of life in patients with advanced cancer and should be a critical outcome of quality improvement initiatives in the cancer care continuum.²⁴ This study reported three important findings. First, opioid daily dose for patients with advanced cancer increased after IPC and decreased from 2008 to 2014 before, during, and after hospitalization. Second, opioid cost per patient and cost per MEDD per patient decreased from 2008 to 2011 and then increased from 2012 to 2014. Third, younger age, earlier prescription year, and higher preadmission MEDD per patient were significantly positive predictors of opioid prescription during hospitalization.

The findings of decreased opioid doses at preadmission and discharge but increased opioid doses after IPC in patients with advanced cancer were consistent with previous large cohort studies in Europe and the United States.^9,10,25 Mercadante et al⁹ reported that patients with cancer who had lower opioid MEDDs (45 mg/d) before hospitalization experienced a six-fold increase in opioid use after IPC. The increase in opioid prescriptions after IPC might be linked to the aim of palliative care to improve pain management for patients with advanced cancer who had exacerbated pain by appropriate prescription of sufficient and strong opioids.^24,26 Successful pain control using effective opioid doses is considered to be an indicator of end-of-life quality of care in PC.^24,26 Compared with the opioid use among patients with advanced cancer in these studies, including studies by Klepstad et al²⁷ (mean opioid dose, 150 mg/d) and Mercadante et al²⁸ (mean opioid dose, 134 mg/d), the patients in this study used lower opioid doses during hospitalization and at discharge, which might be explained by varying prescription practice in different countries and the regulatory scrutiny related to the opioid epidemic in the United States. Moreover, the patients in this study were referred to IPC promptly after hospitalization. The safety initiatives in the MD Anderson PC program, which include screening for substance abuse history using validated scales, routine assessment of pain and other symptoms, personalization of pain goals, and use of adjuvant medications, might result in lower opioid doses dispensed upon admission.²⁴ The percentage of total opioid dose change after the IPC consultations decreased from 2008 to 2014, which could be attributed to the improved primary PC provided by the treating oncologists at MD Anderson in this period (Table 2). Furthermore, the association between IPC and less aggressive health care use could lower opioid doses prescribed to control aggressive care-induced pain at discharge.²⁹ However, we need to assess these possibilities more.

This study found that the MEDD per person decreased from 2008 to 2014, which was mainly caused by the decreased use of hydromorphone (31.5% in 2008% to 22.9% in 2014), morphine (28.5% in 2008% to 16.3% in 2014), and other opioids (codeine, hydrocodone, oxymorphone, and tramadol; 4.1% in 2008% to 2.3% in 2014; data not shown). The decreasing annual trend of opioid use in patients with advanced cancer who received IPC was consistent with prior studies by Curry et al²⁰ and Haider et al²⁸ on PC opioid use by patients from 1996 through 2004 and from 2010 through 2015. Aside from the multimodal strategies to encourage more effective and safe opioid prescribing, the knowledge of patients with advanced cancer and MD Anderson health professionals about potential fatal adverse events related to long-term opioid use likely contributed to the decline of opioid prescribing.¹¹ This decreasing trend could also be a response to stricter regulations on opioid prescription at the federal and state levels. The efforts to save hospital costs via prescription of more cost-effective opioids with fewer adverse events could also contribute to this decline in prescribing.^18,20

The changes in types of opioid prescribing could explain the decreasing then increasing pattern of opioid costs from 2007 through 2014 (Appendix Fig A2). The transdermal fentanyl patch, IV fentanyl, and patient-controlled analgesia therapy of hydromorphone and morphine were the top three contributors to opioid costs because of their high AWPs. Methadone is regarded as an effective opioid with low cost and few adverse events in PC.²⁰ The median cost per MEDD of methadone was lower than these three opioids during the 7 years (data not shown). From 2007 through 2011, the decreasing cost proportions of the transdermal fentanyl patch and IV fentanyl (82.7% to 74.7%), hydromorphone (82.7% to 74.7%), and morphine (82.7% to 74.7%) and the increasing cost proportion of methadone (0.4% to 0.5%) could have lowered the opioid costs. However, such a distribution pattern was reversed between 2012 and 2014 when the cost proportions of the transdermal fentanyl patch, IV fentanyl hydromorphone, and morphine increased, but the cost proportion of methadone decreased. This shifting of opioid prescribing types could have led to the increase of opioid costs in these 3 years.

This study determined that younger age, earlier prescription years, and higher preadmission opioid daily doses positively predicted high opioid daily doses during hospitalization. The changing prescription practice related to stricter regulatory policies of opioid prescription plus more safety initiatives at MD Anderson could explain the negative association between prescription year and high in-admission MEDDs, because patient characteristics remained stable over the 7 years. Like prior studies, this study showed that older patients tended to receive lower MEDDs.^28,30,31 PC physicians prescribed lower MEDDs to older patients because of their decreased volume of distribution so the risk of opioid-induced adverse events was reduced. The positive association between pre- and in-admission MEDDs might imply that the preadmission opioids were inadequate to control pain symptoms, because 91% of patients were referred to IPC to manage moderate to severe pain.^6,8-10,26-28 Consequently, the MD Anderson PC team possibly prescribed higher opioid doses in IPC to improve their pain management effectively and quickly. The identified opioid prescription predictors revealed that age, prescription year, and preadmission opioid doses were associated with physician opioid prescribing patterns for patients with advanced cancer, so these predictors could be targeted to optimize opioid prescription for pain management in the IPC setting.

Most patients with advanced cancer, by the time they are referred to IPC, have been prescribed scheduled opioids as needed to control severe pain. However, PRN doses for breakthrough remain standard earlier in the treatment trajectory of cancer pain management. This study only measured the scheduled opioids prescribed to enrolled patients with advanced cancer, which could lead to a potential underestimation of the total prescribed opioid doses. Furthermore, any costs related to opioid administration were excluded. Thus, the cost results should be applied to estimate opioid costs from a payer perspective. In addition, we focused on patients with advanced cancer who had poor symptom control with scheduled opioids; hence, our results are not generalizable to patients with chronic noncancer pain. This study also did not explore how adjuvant medications were prescribed, what their related costs were, and what their influences were on opioid prescription and overall pain medication costs to patients.

These results demonstrate the changing trends of opioid prescribing and costs in patients with advanced cancer. Policymakers as well as health care practitioners must weigh these changes carefully to ensure that actions to address opioid misuse do not lead to unintended consequences that restrict access to pain treatment for patients with cancer. Furthermore, given the increasing costs of opioids, policy efforts to address increasing drug prices should include opioids despite the public pressures to reduce opioid access.

In conclusion, we found that the daily opioid dose per patient decreased from 2008 to 2014 before, during, and after hospitalization. In-admission opioid costs per patient decreased gradually from 2008 to 2011 and increased from 2012 to 2014. Younger age, earlier prescription year, and higher preadmission opioid doses were positive predictors of higher daily opioid doses prescribed to patients with advanced cancer who received IPC.

APPENDIX

Fig A1.

Opioid utilization patterns by opioid type in hospitalization, 2008-2014. (*) Other opioids: codeine, hydrocodone, oxymorphone, and tramadol. Abbreviation: MEDD, morphine equivalent daily dose.

Fig A2.

Opioid cost proportions (%) by opioid type in hospital admissions from 2008 to 2014 (Jonckheere-Terpstra test). (*) Other opioids: codeine, hydrocodone, oxymorphone, and tramadol.

AUTHOR CONTRIBUTIONS

Conception and design: Sriram Yennurajalingam, Zhanni Lu, Suresh K. Reddy, EdenMae C. Rodriguez, Marie J. Waletich-Flemming, Eduardo Bruera

Collection and assembly of data: Sriram Yennurajalingam, Zhanni Lu, EdenMae C. Rodriguez, Kristy Nguyen, Marie J. Waletich-Flemming, Kyu-Hyoung Lim, Aksha Memon, Nhu-Nhu Nguyen, Kristy W. Rofheart, Guoqin Wang, Srikanth Reddy Barla, Janet L. Williams, Eduardo Bruera

Provision of study material or patients: Sriram Yennurajalingam, Guoqin Wang

Data analysis and interpretation: Sriram Yennurajalingam, Zhanni Lu, Suresh K. Reddy, EdenMae C. Rodriguez, Kyu-Hyoung Lim, Srikanth Reddy Barla, Jimin Wu, Eduardo Bruera

Administrative support: Sriram Yennurajalingam

Financial support: Sriram Yennurajalingam

Manuscript writing: All authors

Final approval of manuscript: All authors

Accountable for all aspects of the work: All authors

AUTHORS' DISCLOSURES OF POTENTIAL CONFLICTS OF INTEREST

Patterns of Opioid Prescription, Use, and Costs Among Patients With Advanced Cancer and Inpatient Palliative Care Between 2008 and 2014

The following represents disclosure information provided by authors of this manuscript. All relationships are considered compensated. Relationships are self-held unless noted. I = Immediate Family Member, Inst = My Institution. Relationships may not relate to the subject matter of this manuscript. For more information about ASCO's conflict of interest policy, please refer to www.asco.org/rwc or ascopubs.org/jop/site/ifc/journal-policies.html.

Sriram Yennurajalingam

Research Funding: Bayer (Inst), Genentech (Inst), Roche (Inst), Helsinn Therapeutics (Inst)

Kristy Nguyen

Employment: Walgreens (I)

Eduardo Bruera

Research Funding: Helsinn Healthcare

No other potential conflicts of interest were reported.