Prescription Opioid Use in General and Pediatric Emergency Departments

We conducted a secondary analysis of data from the National Hospital Ambulatory Medical Care Survey (NHAMCS) from 2006 to 2015, a nationally representative survey of ED visits compiled by the CDC. This investigation was certified exempt from institutional review board review.

The NHAMCS derives data from a 4-stage probability sample of visits to the EDs of noninstitutional general and short-stay hospitals to provide estimates of the characteristics of patients seen in US EDs. The details of the survey methods and quality assurance program are public record.^34,35 

The NHAMCS ED record lists all medications administered in the ED and/or prescribed at discharge by their national drug code (NDC). The NDC is an identifier assigned to each medication listed under Section 510 of the US Federal Food, Drug, and Cosmetic Act. NDC code 058 is for analgesics. Subclasses 060 and 191 are the opioid analgesics and combination opioid analgesics, respectively. For the purposes of this investigation, we categorized the subclasses 060 and 191 as opioid analgesics. Medications within the NDC code 058 and without the subcode 060 or 191 were categorized as nonopioid analgesics, as previously described.³⁶  Neither the dose of medication administered nor the number of pills prescribed is recorded in the NHAMCS data set. The outcome of interest was whether (1) an opioid was administered in the ED, (2) an opioid prescription was given at discharge, or (3) a nonopioid analgesic prescription was given at discharge. The predictor variable of interest was ED type (general or pediatric ED). The NHAMCS does not directly differentiate pediatric EDs from general EDs. We operationally defined a pediatric ED as one in which the average age of patients treated was <18 years and <10% of visits were from patients >25 years of age. This definition is similar to that of previous studies of pediatric and general EDs in which NHAMCS data were used.^11,34  Covariates of interest included age category (0–6, 7–12, 13–17, and 18–25 years), mode of arrival (eg, ambulance), reported pain severity, expected payer source, and race and/or ethnicity.

We employed the ultimate cluster design (single-stage sampling) in variance and 95% confidence interval (CI) calculations using masked stratum and primary sampling unit identifiers provided with the NHAMCS public use data set.³⁵  Descriptive statistics were generated and scaled to the population level by using survey weights. Multivariate logistic regressions were employed by using the “svy” function in Stata 13 (Stata Corp, College Station, TX). Three key outcomes were considered: (1) administered opioid in the ED, (2) prescribed opioid at discharge from the ED, and (3) prescribed nonopioid analgesic at discharge from the ED. For the outcome administered opioid in the ED, the sample included all patients treated in an adult or pediatric ED ≤25 years of age. For the outcomes opioid prescription at discharge and nonopioid analgesic at discharge, we restricted the sample to patients ≤25 years of age who were discharged. We conducted a secondary analysis restricting the sample to patients with a discharge diagnosis of fracture or dislocation using Clinical Classification Software (CCS). The CCS scheme collapses >14 000 International Classification of Diseases, Ninth Revision codes into a smaller number of clinically meaningful categories.³⁷  Throughout the study period, the NHAMCS used International Classification of Diseases, Ninth Revision coding rather than International Classification of Diseases, 10th Revision coding. The fracture and/or dislocation sample was defined as those patients ≤25 years of age with CCS codes 226, 228 to 231, or 830 to 839 as 1 of the top 3 discharge diagnoses. This was done to examine if there was differential opioid use between sites of care in a subpopulation of patients with a seemingly higher need for opioids. Mode of arrival and pain severity were included in the multivariate regressions to reduce confounding by disease severity. Age, race and/or ethnicity, and insurance were included because these have been associated with differential analgesic use previously. In the multivariate analysis, the 7 to 12 years age group was used as the reference to highlight differences in use between preadolescents and adolescents. Predicted probabilities were generated by using the “predict” function in Stata after performing the sampling-weighted regression. Several sensitivity analyses were performed, including exploring more restrictive definitions of pediatric ED, exploring more restrictive entry criteria into the fracture and/or dislocation cohort, and restricting the population of interest to only those <18 years old.

During the study period, we observed 305 570 encounter records, which, after applying the survey weights, corresponded to 1 301 553 214 ED visits. Children, adolescents, and young adults (≤25 years) accounted for 111 219 records corresponding to 474 989 623 ED visits. Pediatric EDs accounted for 3.4% of total EDs (108 of 3082) and 9.0% (9967 of 111 219) of sampled ED visits for children, adolescents, and young adults. The mean age of patients was 38.1 years in general EDs and 6.0 years in pediatric EDs. Fractures and/or dislocations accounted for 4.0% of all ED visits. Characteristics of the sample are shown in Table 1.

Among all patients ≤25 years seen in any ED type for any complaint, 10.0% were administered an opioid in the ED. In general EDs, 10.6% of patients were administered an opioid, compared with only 3.5% of those treated in pediatric EDs (P < .001). However, after adjustment for age, mode of arrival, pain severity, race and/or ethnicity, and insurance, the odds of being administered an opioid was similar for children, adolescents, and young adult patients treated in a pediatric versus general ED (odds ratio [OR] 0.88; 95% CI 0.61–1.27; P = .49). The odds of receiving an opioid varied markedly with patient age in both ED types. Compared with those 7 to 12 years old, adolescents aged 13 to 18 years had 1.6 higher odds (OR 1.66; 95% CI 1.47–1.88), and those aged 18 to 25 years had 2.8 higher odds (OR 2.81; 95% CI 2.48–3.18) of being administered an opioid in the ED. See Table 2 for full regression results. For patients with fractures or dislocations, the odds of opioid administration were similar for patients treated in pediatric and general EDs (OR 1.10; 95% CI 0.74–1.64; P = .63). See Table 3 for results from the fracture or dislocation subgroup.

Patients seen in pediatric EDs were less likely to be prescribed opioid medications at discharge than those seen in general EDs. For general EDs, 11.4% of visits resulted in an opioid prescription, compared with only 1.7% of visits for pediatric EDs (P < .001). During the study period, the total number of opioid prescriptions given to children, adolescents, and young adults at discharge from EDs was 47 360 197, of which 46 702 350 originated from general EDs and 657 847 originated from pediatric EDs.

After regression adjustment for age, mode of arrival, pain score, and other variables, children, adolescents, and young adults treated in pediatric EDs were markedly less likely to receive an opioid prescription at discharge compared with those treated in general EDs (OR 0.37; 95% CI 0.27–0.52; P < .001). Age was a significant predictor of opioid prescription. Patients aged 13 to 17 and 18 to 25 years were much more likely to receive an opioid prescription at discharge (OR 1.96 [95% CI 1.71–2.23] and 3.78 [95% CI 3.34–4.27], respectively), whereas patients aged 0 to 6 years were less likely (OR 0.29; 95% CI 0.24–0.35). Among patients ≤25 years old with fractures or dislocations, 39.1% of those treated in general EDs received an opioid prescription at discharge, compared with just 11.9% of those treated in pediatric EDs (P < .001). In the multivariate model (Table 3), patients with fractures or dislocations remained markedly less likely to receive an opioid prescription if treated in a pediatric ED (OR 0.27; 95% CI 0.13–0.54). Predicted probabilities for receiving an opioid prescription at discharge from general and pediatric EDs at a variety of ages are given in Table 4. There were no meaningful changes in the temporal trends in opioid administration or prescribing between general and pediatric EDs during the study period (Fig 1).

For all patients ≤25 years of age, compared with a general ED, being treated in a pediatric ED was associated with higher odds of being prescribed a nonopioid analgesic at discharge from the ED (OR 1.62; 95% CI 1.23–2.14). For patients with fractures or dislocations, the odds of receiving a nonopioid analgesic prescription were similar, whether treated in a pediatric or adult ED (OR 1.71; 95% CI 0.96–3.05).

In this nationally representative study, we found that children, adolescents, and young adults treated in pediatric EDs were less than half as likely to receive an outpatient opioid prescription than similarly aged patients treated in general EDs. For example, the predicted probability of a 15-year-old patient receiving a prescription opioid at discharge is 4.2% if seen in a pediatric ED and 10.7% if seen in a general ED. For an 18-year-old patient, the predicted probabilities are 6.1% and 15.7%, respectively. Importantly, the vast majority of young persons are treated in general EDs (91%). During the study period, there were 46.7 million prescriptions made to children, adolescents, and young adults in general EDs. Had general EDs employed a similar prescribing pattern as pediatric EDs, 28 million fewer opioid prescriptions would have been given to children, adolescents, and young adults. Conversely, we found the initial ED management of pain to be similar between pediatric and general EDs (eg, the predicted probability of a 15-year-old patient being given an opioid in a pediatric ED was 9.1%, compared with 10.5% in a general ED). The difference in opioid prescriptions between general and pediatric EDs begs the questions: Why are they different, and which is right? Are pediatric EDs undertreating pain or general EDs overprescribing opioids? Restrictive prescribing patterns are only justifiable if pain in pediatric patients is adequately treated. Unfortunately, there is no objective standard by which to assess whether analgesic use was appropriate, overly generous, or underused.

The restrictive opioid prescribing practice in pediatric EDs may be more desirable in light of the nation’s opioid crisis and adolescent vulnerability to drugs.^38,39  Biologically, changes in the prefrontal regions and limbic systems are thought to drive youth risk-taking behaviors such as experimentation with alcohol and/or drugs.^38–40  In turn, the adolescent brain appears to be more vulnerable to changes triggered by drug use, which potentially results in higher chances of progression to drug abuse.^41,42  Empirically, children exposed to prescription opioids are more likely to misuse them in the future. Despite a biological and empirical case for minimizing unnecessary prescribing to youth, the majority of prescribing guidelines are muted on the issue. The much-cited CDC guidelines for opioid treatment are specifically for adults and there is no pediatric corollary, and some pediatric pain experts have warned of potential serious negative consequences of applying the adult guidelines to the younger population.^27,28,43,44 

Our research suggests disparities in opioid prescribing across ED types are driven by pediatric EDs substituting nonopioid analgesic prescriptions for opioids and are not due to poor pain recognition or overall unwillingness to use opioids. Historically, EDs have been cited for undertreating pain or for oligoanalgesia in the management of pediatric patients. Rates of opioid administration have been shown to be consistently lower for children compared with adults suffering from conditions ranging from long-bone fractures to acute appendicitis.^9–11,45  However, the oligoanalgesia phenomenon has not generally been considered to be more prominent in pediatric EDs compared with general EDs. In this study, we found that children, adolescents, and young adults had similar odds of being administered an opioid analgesic during their ED visit, whether treated in general EDs or pediatric EDs, a finding that held for the full study cohort and for the restricted cohort of patients diagnosed with fractures and/or dislocations. This strongly suggests that during the acute stabilization phase of injury, when pain is most severe, pediatric EDs are at least as willing as general EDs to provide opioid analgesics to young people. At discharge, patients treated in pediatric EDs have higher odds of receiving a prescription for nonopioid analgesics compared with those in general EDs, again revealing that pediatric ED providers are not ignoring the need to prescribe pain medications. However, the odds of receiving an outpatient opioid prescription from a pediatric ED are roughly one-third the odds of receiving a prescription from a general ED. Several recent controlled trials reveal that, for most patients, nonsteroidal antiinflammatory drugs are as effective at pain control as medium potency opioids without the risk for accidental overdose or subsequent misuse, medically justifying the substitution of nonopioid analgesics for opioids.^46–50 

It is a matter of urgent clinical and academic importance to understand what factors facilitate pediatric emergency medicine physicians’ low opioid prescribing rate or, alternatively, what modifiable variables drive general emergency physicians’ higher rates. Our study cannot provide direct evidence as to why general EDs are more liberal with outpatient opioid prescribing to adolescents and young adults. Adult patients in the ED have a much higher prevalence of severe pain, chronic painful conditions, opioid tolerance, and opioid-induced hyperalgesia and may expect or demand higher rates of opioid prescribing. In this study 17.4% of patients in general EDs reported severe pain, compared with 4.9% of patients in pediatric EDs, and 18% of general ED visits (all ages) resulted in an opioid prescription. We posit that general ED providers therefore become calibrated to a high set point of opioid prescribing dictated by their adult patients, and this prescribing culture trickles down to pediatric patients in this general environment. Specific efforts may be needed to adjust provider prescription expectations when they practice in a mixed adult and pediatric environment.

This study has several notable limitations. First, this is a secondary analysis of abstracted data. However, these data are part of a large nationally representative sample collected by the CDC, an experienced agency with expertise in data collection. Second, opioid medication dose is not collected during the abstraction process. As such, we are only able to determine that general and pediatric EDs administer opioids with similar frequency, but we cannot observe whether they administer similar morphine equivalents. A third limitation is that the NHAMCS does not directly classify pediatric EDs. We conducted several sensitivity analyses with modified definitions of pediatric ED but found no differences in the results. Fourth, each facility was coded as either a general ED or pediatric ED. Our analysis did not allow for identification of a specific pediatric ED embedded within a general ED, which occurs in some larger hospitals. Therefore, pediatric patients treated within the pediatric section of a general ED were counted as having been treated in a general ED. Overall, this would tend to bias our findings to the null. Although we did incorporate pain severity into the regression models, we could not identify the pain score at the time of opioid administration in the ED or at discharge, which might influence prescribing tendencies. We used the fracture and/or dislocation cohort as a proxy for patients who had a high need for pain medication. This definition relied on patients having a fracture or dislocation as 1 of their top 3 diagnoses. By doing so, we ran the risk of misclassifying the true reason for the visit. However, there is no reason to suspect that any misclassification would occur differentially between general versus pediatric EDs. We conducted sensitivity analyses restricting the fracture cohort to patients whose primary reason for visit was a fracture or dislocation; this had no substantial effect on the results.

Although children, adolescents, and young adults had similar odds of being administered opioids while in the ED, they had markedly lower odds of receiving an outpatient opioid prescription at discharge from a pediatric ED compared with a general ED. For patients with fractures, this pattern persisted. The question of which pattern is most appropriate remains somewhat open. However, given research highlighting the risks of subsequent misuse and long-term use after opioid prescribing and recent evidence demonstrating equivalent analgesic effects between oral opioids and nonsteroidal antiinflammatory drugs in patients with extremity injury, providers in general EDs should consider whether a more restrictive opioid prescribing pattern is warranted for young persons.

CCS

Clinical Classification Software

CDC

Centers for Disease Control and Prevention

CI

confidence interval

ED

emergency department

NDC

national drugcode

NHAMCS

National Hospital Ambulatory Medical Care Survey

OR

oddsratio