Predictors of Concomitant Use of Prescription Opiods and Benzodiazepines in Rhode Island

Objective: To determine the predictors of concomitant use of benzodiazepines with chronic opioid use in Rhode Island using pharmacy level data. Design: Cross – sectional analysis Setting: Rhode Island Prescription Drug Monitoring Program (PDMP) Database, 2015 Participants: Any person who filled a prescription for an opioid (>30 days’ supply), a benzodiazepine or both at a licensed retail pharmacy in Rhode Island during 2015. Main outcome measure: Concomitant use of opioid analgesics and benzodiazepines, defined as the receipt of any benzodiazepine and opioid pharmacy dispensing having a days’ supply which overlaps by at least 1 day. Results: Of 139,410 patients who were included in our analysis, 15.5% had overlapping (concomitant) opioid and benzodiazepine prescriptions during 2015. Patients who were younger than 45 years of age, and who see more than three prescribers were less likely to be prescribed both opioids and benzodiazepines, in which the days supply overlap. Additionally, each additional prescription a person receives for a controlled substance increased the odds that they were prescribed concomitant therapy.


LIST OF TABLES
Hypothesis: were also involved in 30% of cases. 5 In another study, conducted by Yarbourgh, et al, the researchers analyzed interviews with survivors of opioid overdoses and found that most events involved the use of multiple drugs (83%), and often benzodiazepines (32%). 6 Reisfield, et al, noted that opioids and benzodiazepines are the most commonly co-abused drugs; estimates ranging from 20-50% of patients who are prescribed long term opioid therapy, also received prescriptions for a benzodiazepine. 7 The concomitant use of opioids with benzodiazepines is discouraged in prescribing guidelines, 8 but co-prescribing these drugs by both a single prescriber or simultaneously by multiple prescribers is not uncommon, nor is prescribing for a longer duration than is recommended. 9 This may be due to the fact that benzodiazepines are characterized as relatively safe while having a high therapeutic index, 10 and less likely than opioids to be the sole drug involved in an accidental overdose. 11 Benzodiazepines are recommended as a second-line treatment for some common conditions such as insomnia, and anxiety, and when they are prescribed it is recommended that they are used for a short time period and not in conjunction with certain other classes of medicines, including opioids. Benzodiazepines alter the pharmacodynamic effects of opioids. Studies have shown that the concomitant use amplifies the central nervous system effects of opioids. 12 Additionally, opioids and benzodiazepines have different mechanisms of action which is problematic because both opioids and benzodiazepines are central nervous system depressants; these drugs have the effect of depressing the central nervous system through two different mechanisms when the two classes of drugs are administered together. 11 This can lead to overdose; chronic use of both opioids and benzodiazepines has been associated with "additive effects in sleep disordered breathing." 13 Finally, when benzodiazepines are prescribed for a longer period than is recommended, the patient may experience an increased tolerance to the dose of the drug 10 and dependency. 14  Island who were prescribed concomitant therapy with opioids and benzodiazepines in 2015 and determine if there were patient level or prescription level predicators that indicated a patient was more likely to be prescribed concomitant therapy.

Study design
A cross sectional observational study was conducted using de-identified pharmacy-

Data Source
The Rhode Island PDMP database includes pharmacy-level data for all prescriptions filled for schedule II to IV medications at Rhode Island based pharmacies who hold a controlled substance registration number. As mandated by state law, all pharmacies that dispense schedule II-IV medications to individuals who are "not an inpatient of a hospital, correctional institution or nursing facility" are required to electronically report 19 data elements to the PDMP within 72 hours of filling the prescription. The data set does not include medications dispensed to patients of outpatient methadone maintenance programs. A de-identified subset of PDMP data was used, which included the following variables: age, gender, payer, date of prescription fill, the number of days covered by the prescription, medication name, and a de-identified prescriber drug enforcement administration identification number.

Study population:
The study population was comprised of patients of any age who filled a prescription for an opioid medication, a benzodiazepine, or both in Rhode Island, during 2015.

Patient identification:
Inclusion/exclusion criteria The following patients were included: A.) Patients, of any age, who filled at least one prescription for oxycodone, hydrocodone, morphine, oxymorphone, hydromorphone, tapentadol, methadone, pentazocine, butorphanol, meperidine, levorphanol or codeine tablets, or fentanyl patches, totaling at least a 30 days' supply during 2015. Patients who filled prescriptions for buprenorphine or dose forms of opioids which suggest cancer treatment (such as fentanyl lozenges) were excluded; and B. Patients who filled at least one prescription for any benzodiazepine drug.

Study variables:
Outcome variables: The primary outcome measure was the concomitant use of opioid medications and benzodiazepines. This variable was defined as the receipt of any benzodiazepine and opioid pharmacy dispensing having a days' supply which overlapped by at least 1 day.

Independent variables:
Demographic variables included age (and age group), gender, payment type (including cash and insurance), number of unique opioid and/or benzodiazepine prescribers, if the patient received any prescriptions from the top five, ten and twenty five percent of opioid prescribers in Rhode Island (percentage was determined by the total number of opioid prescriptions written in 2015 by the prescriber in relation to the total number of opioid prescriptions in 2015), type of opioid medication(s) prescribed (short acting versus long acting, and also chemical type (oxycodone, hydrocodone, morphine, oxymorphone, hydromorphone, tapentadol, methadone, pentazocine, butorphanol, meperidine, levorphanol, codeine and fentanyl), total days' supply, total number of dispensings for controlled substances of any type (CII-CIV), whether the drugs of interest were prescribed by the same or different prescribers, if the patient was a high utilizer of multiple pharmacies or prescribers, and the concurrent use of stimulants.

Statistical analysis:
A descriptive analysis was conducted to identify the overall frequency and percentage of each variable included in the study. Each independent variable was compared with concomitant use, sole use of opioids, sole use of benzodiazepines and nonconcomitant use of both opioids and benzodiazepines. We conducted a student t-test for continuous variables, and a Chi-square test for categorical variables to determine the statistical significance of differences between the four outcome groups.
After the initial descriptive analysis, the outcome groups were collapsed into one dichotomous variable: concomitant use (sole use of opioids, sole use of benzodiazepines and non-concomitant use of both opioids and benzodiazepines groups were combined to represent non-concomitant use.) Univariate logistic regression analysis was conducted to evaluate the association between each independent variable with the dependent variable and identify the significant risk factors of concomitant use of opioid and benzodiazepine medication. A predictive multivariate logistic regression model was then built using a backward stepwise approach with all significant independent variables identified from the univariate logistic regression. Independent variables associated with combined use of opioid and benzodiazepine drugs at P<0.2 (for any strata) were retained in the model. Diagnostic tests for collinearity were performed between independent variables, 2-way interactions were assessed, and diagnostics of model fit were examined, as guided by Akaike information criterion and the Hosmer Lemeshow test.
The measure of association between outcome and each independent variable was determined by computing the odds ratio with a 95% confidence interval. Bonferroni Correction was applied for multiple or pairwise comparisons for independent variables with multiple levels.
A sensitivity analyses was conducted. Specifically, the definition of concomitant use of opioid and benzodiazepine medications was expanded to include 0, ≥2, or ≥3 concomitant days of opioid and benzodiazepine prescriptions. For zero concomitant days, the gap between prescriptions was not greater than one day.
Finally, a subgroup analyses was conducted to examine whether the significant risk factors for concomitant use of opioids and benzodiazepines differed by gender.

Hypothesis 2:
Of those individuals who were prescribed at least one instance of concomitant prescriptions, the average number of opioid, benzodiazepine and controlled substance prescriptions and the most commonly prescribed opioid and benzodiazepine medications were identified. The most commonly prescribed opioids were oxycodone and hydrocodone, with roughly 50% of the concomitant use group being prescribed at least one of these medications (54% and 49%, respectively). On average, individuals who received concomitant opioid and benzodiazepine prescriptions, received 10 prescriptions for opioids (SD 7) and 7 prescriptions for benzodiazepines (SD 5). See Table 3.
We analyzed prescribers by de-identified DEA number, and determined the top 25%, 10% and 5% of opioid prescribers, by prescription volume. In the concomitant use group, 21,595 (99.7% of the concomitant use group) individuals filled at least one opioid prescription by a provider who was in the top 25%, 7,359 (34%) filled at least one opioid prescription by a provider who was in the top 10% and 4,073 (18.8%) filled at least one opioid prescription by a provider who was in the top 5%. See Table 4.
We conducted a sub group analysis of those receiving concomitant opioids and  Another study which found similar results, compared laboratory drug screening test result data to provider supplied prescription information, for both prescriptions written by the clinician who order the drug screening, and other drugs that the patient was prescribed, which were obtained from the PDMP. All subjects included in this study were prescribed either a benzodiazepine and/or an opioid. This study concluded that concomitant use of benzodiazepines and opioids were more common in women (P<0.01), and that this result was driven by the increased number of benzodiazepine prescriptions which women received (32.7% vs 23.5%). They also found that the use of benzodiazepines, opioids and the concomitant use of both drugs increased with age. 16 We also found that older age increases risk for concomitant opioid and benzodiazepine use, which is supported by findings that older adults may suffer from more chronic pain, 17 and are prescribed benzodiazepines for a range of maladies including, insomnia and anxiety. 18 We theorized that an individual who was prescribed concomitant opioid and benzodiazepine drugs was more likely to have a greater number of prescribers; this theory was confirmed by our findings. Patients who saw greater than two prescribers were more likely to receive concomitant therapy.
These patients may have more complicated medical conditions requiring them to see more physicians, but the care may not be well coordinated. Another potential explanation of this finding is timing of the implementation of the PDMP with relation to our study year. One goal of implementation is to encourage prescribers to consult the database prior to prescribing medications, however through the end of 2015 (the year of the study), only about 40% of prescribers had registered with the PDMP.
Further analysis with subsequent years of data may find different results; as of September, 2016, 100% of Rhode Island prescribers were registered with the PDMP. 19 Lastly, patients who were prescribed stimulant medications were less likely to be prescribed opioids and benzodiazepines as concomitant therapy. We included stimulant use as a covariate due to its potential for abuse. While stimulant use (and subsequent abuse) has been historically seen in younger populations, 20 stimulant dependence and abuse has been reported in older adult populations as well. 21 Much of the published literature surrounding concomitant use of benzodiazepines and opioids discusses changes in prescribing patterns over time, 22 the detrimental effects of concomitant use of these drugs, including death, 5,11,23,24 and socio-economic factors that predict concomitancy. 13,22,25 While these topics are important and relevant to the current climate of pharmaceutical prescribing, these are not factors that we can consider with the data utilized in our study. In addition to its primary purpose as a real-time tool to aid prescribers and pharmacists in making the best decisions for their patients with timely information about patients' prescription history, PDMP data provide opportunity for state-wide evaluation of patterns of controlled substance use, such as our study, that inform public health efforts to promote safe and evidence-based use of higher-risk medications.

Limitations:
While the PDMP dataset lacks certain basic demographic information, such as race/ethnicity, other variables which are captured in the database have been reported as predictors in published literature and were consistent with our findings, including age, and a single provider who writes concomitant prescriptions. 22 This study has several limitations in its current state, some of which can be addressed in future analysis, and others that are limitations of the dataset. Limitations which will be addressed in subsequent analysis are identifying and determining the significance of other prescribed scheduled medications, identifying prescribers who write the most prescriptions for opioids, benzodiazepines, and the concomitant use of these medications. It is possible that some providers are unaware of their patient's concomitant therapy while others may be prescribing it knowingly. Further sub-group analyses will provide more insight into these factors.
Limitations that we cannot address through a more thorough analysis include a lack of diagnosis information, use of non-scheduled drugs, race, socio-economic status attributes including income, education, transportation, housing status, marital statusall which may be of interest but are beyond the scope of this research and database.
We excluded patients who we determined to be receiving end-of-life care or cancer treatment, without diagnosis code this determination was based on the type of prescription that a patient filled. It is possible that we over or under-estimated the number of patients with these conditions.
We conclude that there are predictive factors that can be observed in pharmacy level data to determine if a patient is at a higher risk for being prescribed concomitant benzodiazepine and opioid therapy. These factors are age greater than or equal to 45 years, using health insurance to pay for controlled substance prescriptions (rather than cash) and receiving controlled substance prescriptions from greater than two prescribers.