The sample is derived from 7,728 patients treated in an academic health center in Los Angeles, consisting of 2,576 patients with OUD diagnosis in the EHR aged 18 to 64 years at their first OUD diagnosis and 5,152 control patients, matched by sex, date of birth , first encounter , and the Elixhauser Comorbidity Index. Data from 2006 to 2014 were collected from an electronic health record system utilizing Epic software. Among the 7,728 individuals, 5,202 individuals during 2010–2014 were matched in PDMP and were included in the study. This study was approved by the Institutional Review Boards at UCLA and the State of California.The primary outcome was all-cause mortality status at the end of follow-up, which was December 31, 2014, for the alive patients or date of death for the deceased patients. Mortality records, available through December 31, 2014, were obtained from the Centers for Disease Control and Prevention National Death Index . The California state Prescription Drug Monitoring Program , also known as CURES is a database of prescription records of Schedule II, III, and IV controlled substance prescriptions dispensed in California. The description of PDMP data and conversion of opioid medications to morphine milligram equivalent have been described previously. We defined most recent opioid use as at least one opioid prescription in the last 30 days of observation . The daily opioid dose was calculated by averaging MME of all opioid prescriptions in the last month of observation over 30 days. BDZ medications were converted to diazepam milligram equivalents for comparison purposes.
We defined most recent BDZ use as at least one BDZ prescription in the last 30 days of observation. The daily BDZ dose was calculated by averaging DME of all BDZ prescriptions in the last month of observation over 30 days. Covariates were obtained from the medical records. Sociodemographic variables included age, sex, race, and insurance status. Clinical variables were defined from the ICD-9 codes. Included comorbidities were physical conditions , cancer, heart disease, respiratory disease, liver disease, sleep disorder, chronic pain, human immunodeficiency virus , weed growing systems and Hepatitis C ; psychiatric conditions ; and substance use disorders .We conducted t-tests for continuous variables and chi-square tests for categorical variables to examine differences in demographics, comorbidities, prescription patterns, and mortality status between OUD and non-OUD patients. For multivariate analyses, a logistic regression model was used to examine the association between the prescription of BDZ together with opioids and mortality. In addition to the primary independent variable , other significant covariates from univariate analyses were included to build additional models. Covariates considered as potential confounders were identified based on a combination of clinical significance and significant p value. Akaike Information Criterion and deviance were calculated. Also, after stratifying by OUD diagnosis, an interaction term of BDZ and opioid use in the last month of observation was included in the models with same covariates as the third model to examine the moderating effect of co-prescription on mortality. All statistical tests were based on a significance level of α ≤ .05. Analyses were conducted using SAS 9.4 . Using the state’s PDMP records linked with EHR and CDC NDI data, this study explored the co-prescription of BDZ and opioids and its association with all-cause mortality among patients in a large healthcare system.
We found that 1) OUD patients had higher rates of other substance use disorders, mental and physical comorbidities relative to non-OUD; 2) OUD patients were prescribed both BDZ and opioids in significantly higher doses compared to non-OUD patients; 3) most recent average daily BDZ or opioid prescription dose was significantly associated with increased all-cause mortality after adjusting for covariates, and 4) there is a significant interaction between BDZ and opioid co-prescription on mortality among non-OUD patients. Though this study analyzed records prior to 2015, national efforts and guidelines have been implemented since that time to help combat problems associated with co-prescription of BDZ and opioids. Findings from this study support recommendations to minimize co-prescription of BDZ and opioids whenever possible, and to use lowest effective doses when prescribing BDZ or opioids. Our study extends prior literature by examining a general patient population from a large health care system with a larger proportion of privately insured individuals and focusing on a population at risk given OUD diagnosis history. Compared to non-OUD patients, we found higher rates of all-cause mortality and higher average doses of BDZ prescribed among OUD patients. Prior research has demonstrated an increased likelihood of BDZ prescription in persons with chronic noncancer pain prescribed higher doses of opioids. We also observed elevated rates of physical and mental health comorbidities in OUD patients relative to non-OUD patients. Accumulating evidence has supported the positive association between opioid use and physical or psychiatric comorbidities, which presents considerable management challenges in clinical practice and contributes to mortality risk . The findings in this study suggest that use of either opioids or BDZ is significantly associated with all-cause mortality, which is consistent with our previous findings and prior studies. Furthermore, our finding of the association between opioid dosage and mortality was in accordance with prior studies demonstrating mortality risk associated with opioid prescribing in a dose-dependent manner.
The CDC guideline recommends risk review and mitigation for opioid dosages greater than 90 MME per day; dosages as low as 50 MME per day can increase the risk for drug-related adverse events for patients with chronic pain. For BDZ, we found 36.3% of deaths among OUD patients and 34.7% among non-OUD patients were BDZ involved. We also found that increased daily BDZ dosage was associated with mortality risk after adjusting for potential confounders, which is consistent with prior literature focused on overdose risk as well as all-cause mortality associated with BDZ. Prior studies have reported BDZ dose relationships with mortality as quantified by cumulative prescribed doses over observation periods; this study extends prior work to include average prescribed dose in the final month of observation. This finding of increased mortality associated with BDZ dose may be related to management of underlying medical conditions closer to time of death, medication interaction effects, or other factors. The association between co-prescription of opioids and BDZ and overdose risk or mortality has been reported among Medicare and Medicaid enrollees and veterans. Findings from this study extend to a primarily privately insured population. In our study, it is worth noting that there was a significant interaction between BDZ and opioid co-prescription on all-cause mortality among patients without OUD diagnosis, but not in those with a diagnosis of OUD, despite elevated rates of physical and psychiatric comorbidities in those with OUD. Although our observational study cannot determine the etiology of this differential finding, a contributing factor might be higher opioid tolerance to the effect of respiratory suppression among OUD patients compared with non-OUD patients. As reported,indoor farming systems higher doses of opioids and BDZ were prescribed in patients with OUD. Clinicians should be cautious about prescribing BDZ to patients using opioids, whether or not they have a diagnosis of OUD, and should be aware of the high rates of comorbidity in this population. More studies are warranted to help elucidate the potential medication interaction of opioid and BDZ. The results of this study revealed several covariates associated with increased risk for all-cause mortality in addition to opioid and BDZ use. We found that self-pay for health care was a strong risk factor associated with mortality compared with public and private insurance; this could be related to health care access or utilization in this population or other factors and warrants further study. Not surprisingly, older age, alcohol use disorder comorbidity, and some physical health comorbidities, such as heart, liver, cancer, and HCV, were significantly associated with all-cause mortality. This study has several limitations. First, findings based on EHR data are limited by provider diagnostic coding and documentation choices, which are influenced by clinical experience and billing requirements. This may affect the accuracy of the data as it was not collected specifically for research purposes; diagnoses may be over- or under-captured. Like other records-based research, diagnoses and prescriptions are limited to this healthcare system and PDMP records; those obtained outside of the system were not captured, and thus dosages may not accurately reflect the total amount consumed.
It is possible that patients used remaining BDZ or opioids from prescriptions prior to the last 30 days of the follow-up or from other sources such as friends or illicit purchases. Therefore, our findings are likely a conservative estimate. Second, the number of overdose deaths is relatively low in this study, and overdose deaths may be misclassified, so we decided to use all-cause mortality instead. Participants were predominantly white, insured individuals living in the Los Angeles area, limiting the generalizability of findings. Lastly, residual confounding is still possible, even after adjusting for many demographic and comorbidity covariates. Advent of the coronavirus 2019 disease pandemic was associated with changes in drinking and drug use among young adults. For alcohol, most studies found increases in the number of days drinking and decreases in the number of drinks consumed per occasion , though other studies have found no significant change or a decrease in the number of days drinking. Studies found no change in the number of days using nicotine and no change or increases in the number of days using cannabis during the pandemic. The emergent literature has three key limitations. First, it is unclear whether the initial effects of the pandemic on drinking and nicotine use in the Spring-Summer of 2020 persisted over time. Most published work has focused on the immediate impact of public health policies to reduce the impact of the pandemic in March 2020 . Second, those studies with more extended follow-up have not been designed to distinguish pandemic effects from maturation effects , leaving it unclear to what extent the observed changes in drinking or nicotine use are specifically due to the COVID-19 pandemic. Developmental increases in drinking and drug use are expected as young adults mature, even absent a pandemic . Thus, characterizing the effects of the pandemic in the medium- and long-term requires a design that can subtract out the developmental change that would be expected outside the pandemic context. Third, initial evidence regarding an important potential moderator of the pandemic’s impact—its impact on financial security—has been mixed. One study found financial strain was linked to greater pandemic-related increases in nicotine use during March and April 2020 while another study found loss of income did not moderate pandemic-related changes in drinking during June 2020. The financial impact of the pandemic on U.S. adults has been heterogenous and time-varying , so both replication and extension of these findings with a longer period followup is warranted. Procedures were approved by Institutional Review Boards at each study site. The NCANDA Study was designed to investigate the impact of heavy alcohol use on neurodevelopment. 831 participants ages 12–21 years old were recruited into NCANDA in 2012–2014 and have been followed prospectively at five study sites across the U.S: Duke University, University of Pittsburgh Medical Center , Oregon Health & Science University , University of California San Diego , and SRI International . Exclusion criteria were intentionally minimized: participants lived within 50 miles of the study site, had no MRI contraindications, had no reported prenatal or perinatal exposures or complications, had no pervasive developmental disorder, had no current or persistent major psychiatric disorder that would interfere with the protocol, and were not taking medications known to affect brain function or blood flow . Each site aimed to recruit a community sample representative of the racial/ethnic distributions of their county. Participants were recruited through announcements at local schools and colleges, public notices, and targeted catchment-area calling. The current study draws data from 348 participants ages 12–15 years old at study entry—older participants were excluded to minimize the potential for cohort effects on drinking and nicotine use . 49% of participants were female. 13% identified as Hispanic; 68% as White, 12% as Black, 7% as Asian, and 8% as Alaskan Native or Pacific Islander. 84% of participants had 1 + parent who completed a Bachelor’s degree. After completing their baseline assessment at study entry, participants were assessed every six months going forward with a combination of in-person assessments and phone interviews .