While outcomes could be interpreted differently, they likely reflect multifaceted perceptions with plausible roles in decision making. When considering risks and potential benefits of tobacco products, adolescents hold views that include multiple aspects of social and physical risks. For e-cigarettes, specifically, adolescents cite multiple influences, both related to the product itself and their social context. Adolescents’ smokeless tobacco use motivations likewise comprise multiple factors, including flavors, perceived nicotine strength, and loyalty to preferred brands. The present study shows that several of these factors each independently contribute to multi-faceted perceptions. Among these dimensions, curiosity and ease-of use are strong predictors of tobacco use behaviors.Youth who perceived flavored smokeless tobacco and flavored e-cigarettes as easier to use than unflavored options were more likely to be susceptible to smokeless tobacco use and to initiate future e-cigarette use, respectively. The present discrete choice findings demonstrate that multiple independent product related factors are associated with constructs shown to predict future tobacco use.A study limitation is that discrete choice experiments ask participants to make hypothetical choices that may not resemble the actual setting in which purchase or use decisions are made. Provided only limited information, study participants may have based some selections on word associations outside the context of e-cigarettes or moist snuff, for example, curing drying connecting the words ‘tobacco’ or ‘cigalike’ with dangers expressed in anti-smoking messages or considering ‘fine cut’ to indicate high quality rather than the coarseness of moist snuff tobacco.
However, adolescents, especially those inexperienced with tobacco use, are likely to possess limited product information in real-word settings and may make the same cognitive associations when evaluating a tobacco product from words on a package, advertisement, or warning. In this study, the presentation of images prior to initiating the text only items may have helped mimic flavor imagery on packages or advertisements; however, it cannot be ruled out that the specific images or colors chosen had some influence on respondents’ choices. A strength of discrete choice survey methods is that the contributions of multiple characteristics are considered in combination, corresponding better to real-world product choices. However, this strength relies on the unverifiable assumption that participants make rational trade-offs between product characteristics when evaluating options. To our knowledge, this study represents the first to apply discrete choice techniques to a school-based sample of youth for both smokeless tobacco and e-cigarettes, including pod-type devices. A school-based design may yield a more representative sample in terms of social-economic profile and tobacco use experiences than an online panel. However, as a limitation, data collected in rural regions of California may not generalize nationally, including to other rural geographical locations. As a non-random sample, the generalizability of this study population is also limited. Advantageously, a rural sample is likely to have greater familiarity with moist snuff products. However, the small total number of moist snuff users did not yield ideal power to examine interactions by use status. Notably, public messaging from health authorities in California emphasizing the potential harms of e-cigarettes and nicotine could have resulted in more concern about nicotine in this sample than would be observed elsewhere.Escalating rates of opioid use disorder and overdose deaths in the United States have become a public health crisis.
Opioids were involved in more than 42,000 deaths in 2016, which is five times higher than in 1999 . To address this public health crisis, considerable efforts are being devoted to better identify risk factors and implement strategies to reduce opioid misuse and mortality. Prescription drug monitoring programs are prominent among the major efforts to reduce opioid misuse, diversion, and related morbidity and mortality. PDMPs are state-run electronic databases . Although there is variation across states, most PDMPs collect and manage prescription records for controlled substances dispensed by pharmacies. These data provide important information for public health agencies, health systems, and providers on prescribing patterns that can inform clinical practice and public health interventions to prevent substance use disorder. PDMP data can also allow the identification of high-risk patterns of prescription drug use and trends in prescribing patterns. Previous studies have shown that mortality attributable to prescription drugs decreases with declines in the prescribing rates of drugs . The few existing studies have called for additional investigations of longitudinal prescribing patterns associated with health outcomes, including mortality . In our prior studies based on a sample of 2,576 OUD patients from a large general healthcare system in Los Angeles, California, we found high rates of mortality among OUD patients compared to either the general population or a matched sample of 5,152 patients without OUD from the same healthcare system . Linking individuals’ medical, death, and PDMP records, the present study expands our previous efforts to investigate the longitudinal opioid prescribing patterns among patients with OUD and matched patients without OUD in relation to their mortality.Data were collected for all patients from the PDMP, mortality records , and medical records .
California’s PDMP is called the Controlled Substance Utilization Review and Evaluation System , maintained by the California Department of Justice . CURES is a database of prescription records of Schedule II, III, and IV controlled substance prescriptions dispensed in California. Each prescription record in the PDMP data includes product name, strength, units of measure , drug formulation , date filled, days’ supply and number of refills. For this study, we included all opioid prescriptions except buprenorphine . For comparison purposes, the opioid medication was converted to morphine equivalents by summing up daily morphine milligram equivalent per prescription and the number of supply days over a year’s time period. Four PDMP outcome measures were considered: number of fills, days of supply, units, and morphine equivalents. Mortality data,dry cannabis available through the end of 2014, were obtained from the National Death Index of the Center for Disease Control and Prevention . Among the 7,728 patients, 1,002 patients passed away. NDI death records include date and cause of death for deaths that occurred as of December 31, 2014. Covariates were based on medical records. Sociodemographic variables included: sex, race, age, and health insurance. Clinical variables were defined from diagnoses for physical health or diseases , substance use disorders , and psychiatric or mental health condition . The lists of ICD-9-CM codes utilized in this study have been previously described and may be obtained upon request from the corresponding author. We conducted t-tests and chi-square tests to examine differences in demographics and diagnoses between patients with OUD versus controls. For longitudinal analysis of opioid prescribing , we used a growth modeling approach to examine group differences by OUD and mortality status at 2014 , controlling for demographics and physical and mental health conditions. For all participants, 4 years of PDMP data were analyzed. Because not all patients had PDMP records, we conducted sensitivity analyses by applying the same growth models based only on the 5,621 patients who had at least one PDMP record. Because the modeling results were similar, we report findings based on the total sample, although the model estimates were generally lower based on the total sample compared to those based on the 5,621 patients who had at least one PDMP. All statistical tests were based on a significance level of α ≤ .05. Analyses were conducted using SAS 9.4. The study results revealed that patients with OUD in our sample received more opioid prescriptions and were prescribed opioids at higher doses than controls, and that higher levels of opioid prescription were associated with greater mortality risk. Moreover, the most striking and clinically relevant finding from this study is that escalating prescribing patterns were associated with heightened mortality risk for both OUD patients and controls, significantly more so among the OUD patients. For example, deceased OUD patients were receiving opioid prescriptions at a yearly increment of 7.84 ME grams greater than those received by non-OUD alive patients, ending with an average of 87.1 MME per day in the last year of observation, as captured in the PDMP system; other potential sources of prescription and non-prescription opioids could not be ascertained.
Additionally, patients with older age, public health insurance, and cancer or chronic pain were more likely to have received more opioid prescriptions over the four years of the observation period. The study had several limitations. The first limitation is that the California PDMP system does not include opioid prescriptions for all patients; exclusions include, for example, the Veterans Affairs , military, inpatient hospitals settings, methadone clinics, out-of-state pharmacies, and Internet sources. Approximately 72.8% of the study sample had a record in the California PDMP database. In this article, we have reported our results based on the total sample given that sensitivity analyses based on those who had at least one record in the PDMP system showed similar patterns of significant effects. However, the reported modeling estimates should be considered conservative given that they are generally lower than those based on the sample with at least one PDMP. Another limitation is that the study is based on patients seen in a single health system serving predominantly white patients living in the Los Angeles area of the United States, which may limit the study’s generalizability. Our findings are also dependent on the extent, accuracy, and validity of the data available in the EHR and PDMP datasets. For example, routine screening for SUDs is not standardized or mandatory in the health system, leaving room for under counts of SUDs and inconsistent documentation of these conditions, particularly in cases of less severe disorders. While we have used a matched control study design, causality underlying the observed findings in this study cannot be ascertained. For example, it may be that end–of-life palliative care leads to escalating opioid prescribing for comfort and not that escalating opioid prescribing leads to mortality. Relative to non-OUD patients, OUD patients had higher rates of chronic pain but equivalent rates of cancer . In our previous study on causes of death among this sample, cancer deaths accounted for similar percentages of death among patients with or without OUD, while overdose deaths were much higher among OUD patients than non-OUD patients . Therefore, end-of-life care cannot entirely account for the observed association between mortality and escalating opioid prescribing, particularly among OUD patients. Factors associated with rising rates of opioid prescriptions and increasing dosages in the OUD group, including co-prescribing of other central nervous system depressants , other substance use, and medical comorbidities were not addressed in this analysis but warrant further study. Future studies may also examine the role of provider specialty and setting in opioid prescribing patterns associated with mortality. The present study has several strengths. Foremost, this is the first study that has linked patient medical records, medication prescription records, and mortality to investigate longitudinal opioid prescribing patterns among OUD and non-OUD patients in relation to mortality. Another important strength is that in the examination of the relationship between opioid prescribing patterns and mortality, the study controlled for patient severity of physical and mental conditions by using a matched control sample and incorporating them as covariates in the model. Given the current crisis of OUD and overdose deaths in the United States and urgent needs to better understand the prescribing patterns and patient characteristics in relation to mortality, our findings have important implications for improved policy and clinical practice addressing this public health issue. The study findings revealed higher rates and doses of opioid prescription among individuals with OUD, relative to non-OUD controls. Most significant is that mortality risks are associated with escalating patterns of opioid prescription regardless of whether the patients had OUD or not. Providers treating patients with OUD need to be alerted to escalating opioid prescribing patterns in light of the finding that the deceased OUD patients demonstrated the sharpest escalation of number of opioid prescriptions and total dosage. To facilitate clinical application, an evolution in the summary and visualization of patient PDMP records should be provided. There is increasing recognition of the importance of using PDMPs as critical tools for supporting clinical decision-making and for decreasing the risk of multiple prescribers and overdoses. This study also confirms the great value of using PDMP data to advance scientific knowledge as demonstrated in other studies . For example, by routinely incorporating PDMP data into diagnostic and medical care, these data can be useful for predicting and/or corroborating clinical diagnoses, gauging treatment responses, and monitoring treatment outcomes .