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

Virginia T. LeBaron, PhD; Fabian Camacho, MS; Rajesh Balkrishnan, PhD; Nengliang (Aaron) Yao, PhD; Aaron M. Gilson, PhD


J Oncol Pract. 2019;15(12):e997-e1009. 

In This Article


Our results provide important context for managing pain for patients with cancer, especially those who reside in areas significantly affected by the opioid epidemic. Perhaps most importantly, our findings suggest possible significant undertreatment of cancer-related pain. Of the total number of patients with cancer, over all the years, less than 25% were in the three-or-more-POM category. More than 60% of patients never received a C-II prescription, which is consistent with the findings of studies estimating undertreatment of pain in 60% to 90% of patients with cancer.[3,4] The large percentages of patients with cancer who were never prescribed a C-II are concerning for a number of reasons, especially when we consider the results per year (Data Supplement). First, the no–C-II patients remain more than 80% of the total sample, each year, even after accounting for the upscheduling (from C-III to C-II) of commonly prescribed hydrocodone products in 2014. Second, anecdotal data and emerging empirical evidence demonstrate that patients with legitimate pain needs, including patients with cancer, experience significant difficulty accessing POMs.[26–28] Because POM regulations have continued to tighten since the end date of our analysis (2015), it is likely that the number of patients with cancer in the no–C-II category is now higher.

Specific prescribing practices may also suggest suboptimal management of cancer pain. For example, the most frequently prescribed C-II POMs were opioid-acetaminophen combinations, with ceiling effects that limit dosing because of acetaminophen toxicity. It is also noteworthy that eight prescription claims were for meperidine; although a small number, meperidine has been removed from most pharmacy formularies because of its potential for serious neurotoxic adverse effects, and it is not recommended for the treatment of cancer-related pain.[29,30] In addition, more than a quarter (28%) of three-or-more-POM category patients also received three or more benzodiazepine prescriptions within the same year. Although the clinical context for these benzodiazepine prescriptions is unknown, coprescribing benzodiazepines and opioids is associated with increased harms,[31] and current prescribing guidelines[32,33] do not recommend this practice. Finally, few prescriptions (4.6%) were written for deterrent C-II formulations; whether this was because of a lack of prescriber knowledge, formulary unavailability, or payment or reimbursement constraints is unknown.

These factors reinforce the continued need to improve safe POM prescribing practices. Although guidelines and educational interventions for safe opioid prescribing are being actively developed and implemented, most focus on patients with chronic noncancer pain[33] or are designed to reduce opioid prescribing, for example, in the immediate postoperative period[34] or in the emergency department.[35] Our results further support the need for prescriber resources that focus specifically on the unique needs of patients with cancer,[36,37] particularly patients in active treatment or at the end of life, a population for whom limiting or reducing POMs may be inappropriate, or where a patient's symptom management needs may not always align with guideline recommendations (eg, the patient with cancer who has a clinical indication for both benzodiazepines and POMs).

Another compelling finding is the low number of patients with cancer prescribed C-II POMs who were then hospitalized for an OUD (less than 1%), consistent with recent research.[38] Of particular interest is our finding that POM frequency did not affect the number of OUD hospitalizations. In other words, OUD hospitalization prevalence did not change whether a patient was prescribed a C-II three or more times per year, one to two times per year, or never. Although it is imperative that POM prescribing for patients with cancer be accompanied by risk assessments and close monitoring,[39–41] our finding underscores the fact that OUD concerns should not deter access to POMs, a mainstay therapy for patients with cancer pain. This is especially important because general POM prescribing guidelines, which recommend strategies such as limited duration of therapy or dosage thresholds, may strongly influence prescribers' willingness to appropriately manage pain, even in a high-need cancer population.

Although pain management should always be tailored to individual need, logistic regression results can help inform more efficient and effective delivery of health care services. For example, patients with cancer with multiple malignancies who are 30 to 64 years of age and reside in particular geographic regions (such as Wise County) may especially benefit from education regarding POM benefits and risks and safe handling. In addition, male patients with Medicaid were less likely to be three-or-more category patients, suggesting that lower-income patients may be at risk of poor pain control.

Geospatial explorations also identified targeted regions for future interventions and additional investigation. Paradoxically, fatal POM overdose rates are highest in areas where the fewest patients with cancer are receiving C-II prescriptions. This finding raises some interesting questions and highlights inherent data limitations. One important consideration is that we do not know exactly who is overdosing on POMs. For example, Figure 2 represents the general population, and we do not know if these individuals are patients or, more specifically, if they are patients with cancer. In addition, where are those who overdose getting the POMs attributed to their overdose (eg, from a legitimate prescriber or from a seller on the street)? We need to understand the contextual characteristics of those who overdose from POMs to answer these important questions, but our preliminary findings signal that patients with cancer are not the primary contributors to POM fatalities. This finding may provide reassurance to prescribers when their clinical judgment determines that POMs are in fact the appropriate and necessary medication to optimally manage a patient's cancer pain. In addition, our maps provide important clinical context to current prescribing rate maps, such as the CDC Prescribing Rate Maps;[42] the CDC maps show high levels of POMs prescribed in all counties of HPDs 1 and 2, but they are not specific to certain patient populations, such as patients with cancer.

Given the high premium on provider time, especially in underserved rural areas, having a clearer idea of at-risk populations can help inform resource allocations. For example, on the basis of our geospatial findings, it may be particularly effective to implement policies regarding safe POM handling and prescribing in counties where the highest probabilities exist of a patient being prescribed a C-II POM, and to focus overdose prevention and intervention efforts where we see the highest rates of POM fatalities.

The CV-APCD does not include data on stage of cancer, date of diagnosis, or clinical data to determine appropriateness of prescribing POMs. We attempted to address this limitation by carefully selecting our sample to capture those patients most likely to receive C-IIs for cancer-related pain. Other important CV-APCD limitations include the following: (1) to meet de-identification requirements, statistical modeling relied on five-digit zip codes (v census tract), which limited geospatial analysis precision; (2) quantity dispensed was not a reliable variable, which precluded the ability to accurately calculate a milligram morphine equivalent[43] daily dose metric for each patient; (3) we captured OUD hospital admissions but not patients referred for community-based treatments or those who experienced other opioid-related harms; we also lacked the data to assess patient history of substance misuse, chronic pain, or mental illness that may influence OUD risk; (4) patients who pay cash for prescriptions were excluded; (5) race and ethnicity were not reliable variables and could not be included; (6) we did not have a comparison non-cancer control group because of cost constraints related to data access.

This study raises a critical issue about how to optimally access and integrate multiple data sets to most accurately investigate POM prescribing and potential harms. As a primary data source, the CV-APCD provided important information, but it had limitations. In fact, the CV-APCD was obtained as a substitute after access to our intended data source, the Virginia Prescription Drug Monitoring Program (PDMP), remained unavailable. PDMPs are state-run databases that record controlled substance prescriptions written and dispensed, regardless of payer source.[44] Analyzing PDMP data to characterize potential POM-related harms through evidence of pharmacy hopping or doctor shopping, and ideally to link to OUD admissions, is crucial for developing data-informed policies.[45]

Another challenge involves obtaining relevant local-level data. We planned to analyze a POM consumption metric at the local level. However, as of this analysis, we were unable to obtain average POM stock data from local pharmacies, and local DEA opioid consumption data have historically been only publicly available at the three-digit ZIP-code level.[46] POM consumption data at a more geographically granular level are available but remain proprietary (Quintiles IMS Transactional Data Warehouse). However, recent developments may make these data increasingly more accessible.[47] Access to comprehensive quality local POM consumption data would provide additional insights into disparities in pain relief and how POM availability may, or may not, contribute to opioid-related harms.

Investments at the local and state level to most accurately capture POM fatality data would help inform policies and prescribing guidelines. Despite methodologic limitations, opioid overdose data are often the key catalyst for initiatives with far-reaching implications; therefore, ensuring these data are as accurate as possible is critical. The Virginia Office of Chief Medical Examiner forensic reports[7] are unique in their attention to toxicology details. Policies that support county- and state-level comprehensive reporting of POM deaths (eg, clearly separating overdoses attributed to legitimate POM prescriptions from illicit street opioids, such as synthetic fentanyl, and including contextual details of the deceased's medical condition) would be extremely helpful to better understand how many opioid deaths are related specifically to patients with cancer.

Future research should (1) seek a larger sample, from multiple state APCDs or within the CV-APCD, to replicate our key findings; (2) link APCD data to state cancer registry data or Centers for Medicare & Medicaid Services/SEER Medicare data to provide important diagnostic, demographic, and clinical context; (3) use a control group (patients without cancer) to analyze differences between the populations; and (4) conduct qualitative fieldwork to understand pain management guideline implementation and how POM regulations affect patients with cancer, particularly those who live in regions heavily affected by the opioid epidemic.

A clearer view of geographic patterns and predictors of both POM prescribing and potential harms can inform targeted interventions and policy initiatives that achieve a balanced approach to POMs, ensuring access for patients in need while reducing risk to both patients and communities. Our research makes an important contribution by exploring how the current opioid epidemic relates to rural patients with cancer.