Outpatient Prescription Opioid Use in Pediatric Medicaid Enrollees With Special Health Care Needs

We performed a retrospective cohort study of CYSHCN using the MarketScan Medicaid Database (IBM Watson Health, Armonk, NY). This database contains claims data from fee-for-service and managed care plans from 11 deidentified states representing all geographic regions of the United States and has been extensively used for studies of CYSHCN.^18,19 We included children ages 0–18 years continuously enrolled in Medicaid (≥11 months) in 2016. Continuous enrollment was chosen for study inclusion because the database does not contain pharmacy claims that could have occurred during enrollment gaps. CYSHCN were defined by the presence of ≥1 chronic conditions as defined by the Agency for Healthcare Research and Quality (AHRQ) Chronic Condition Indicator (CCI) system.²⁰ The AHRQ CCI system classifies ∼68 000 International Classification of Diseases, 10th Revision, Clinical Modification (ICD-10-CM) codes as chronic or not chronic and assigns each chronic condition code to 1 of 18 mutually exclusive categories that are largely organized by organ system (eg, digestive or respiratory).^21,–23 In the CCI algorithm, a chronic condition is defined as a condition that lasts 12 months or longer and has 1 or both of the following effects: (1) it places limitations on self-care, independent living, and social interactions; and/or (2) it results in the need for ongoing intervention with medical products, services, and special equipment. Adapted for use in children, the AHRQ CCI system is all-inclusive of childhood chronic conditions across the spectrum of prevalence, severity, and complexity.^19,23,–25 

For the patient-level analysis, an OE was defined as any filled prescription for an opioid based on the 2-digit generic product indicator (GPI) code for opioids (2-digit GPI = 65). We reported exposures to specific opioids at the level of the generic product name. For the prescription-level analysis, OEs were further classified as single- or multiple-fill episodes. Single-fill episodes consisted of an opioid fill without a refill or a new fill in the subsequent 7 days. Multiple-fill episodes consisted of an opioid fill with a refill or a new fill in the subsequent 7 days and could comprise several sequential fills. We identified children with multiple fills because they likely indicate ongoing pain and/or prolonged opioid therapy.

We described demographic characteristics including age, sex, and race and/or ethnicity. We assessed the type and number of chronic conditions for each child with ICD-10-CM codes using the AHRQ CCI system. To better understand the potential severity of the chronic conditions, we further identified the subset of CYSHCN with complex chronic conditions (CCCs), known to be associated with increased medical complexity (eg, multiple co-morbidities, technology dependence, polypharmacy), higher morbidity and mortality, and increased resource use.²⁶ 

For each opioid fill, we used the AHRQ Multi-level Clinical Classification System (CCS) to comprehensively assess all health problems (eg, injury) that were filed during health care encounters across the care continuum (ie, inpatient or outpatient) during the 7 days preceding OE. The AHRQ CCS system groups related ICD-10-CM codes into mutually exclusive categories. We determined frequencies of all CCS level 2 diagnoses to determine the most prevalent health problems in the 7 days preceding OE. Of these, we associated OE first to diagnoses with clinical indications for OE (eg, traumatic injury, malignancy). If there was not a forthcoming clinical indication for OE, we then considered other related diagnoses (eg, infection, nontraumatic musculoskeletal condition) and lastly the categories of symptoms or descriptions of the health care encounter itself (eg, ill-defined symptoms, signs, and conditions).

Health care use across the continuum was assessed by using preestablished service categories in the data set. These categories included primary care, outpatient specialty, home health care, ED, inpatient (including inpatient surgery), outpatient surgery, dental surgery, and mental health. Overall, and by single versus multiple-fill episodes, we reported health care use during the 7 days before an opioid fill, and during the 7 and 30 days after an opioid fill.²⁷ In the case of multiple fills, only outcomes surrounding the first fill were considered. Only prescriptions filled between January 8 and December 1 of the calendar year of study were considered for analysis to allow for complete use data.

We used descriptive statistics to describe key characteristics of the study population, including the exposure of CYSHCN to opioids, the exposure by specific chronic condition, and the most common specific chronic medications taken by CYSHCN. Comparisons between categorical variables of interest were conducted by using the χ² test. Statistical significance was set at P < .001. All analyses were performed by using SAS 9.4 (SAS Institute, Cary, NC). This study of deidentified data was exempt from review by the policies of the Colorado Multiple Institutional Review Board.

Among 4 032 955 continuously enrolled children, 64.4% were CYSHCN. Overall, OE occurred in 3.5% of children without a special health care need, 7.4% of CYSHCN, and 14.0% in CYSHCN with a CCC. CYSHCN represented 80.0% of all children who had OE. The 2 597 987 CYSHCN comprised the study population for all subsequent analyses (Table 1).

Children with versus without OE were older (eg, ages 10–18 years: 69.4% vs 48.7%), more likely to be girls (52.0% vs 47.7%), non-Hispanic white (55.0% vs 46.9%), and had more chronic conditions (eg, ≥3 conditions: 49.1% vs 30.6%) as well as CCCs (18.8% vs 9.0%), P < .001 for all. Children with OEs were also more likely to have filled prescriptions from multiple other medication classes during the study year (eg, ≥5 additional medication classes: 54.7% vs 31.2%, P < .001). Specifically, 6.5% of children with OEs had concomitant prescriptions for an anticonvulsant at some point during the study period, 5.9% for an anxiolytic, and 5.4% for a benzodiazepine.

As the number of chronic conditions increased from 1 to 5 or more, the rate of OE increased from 4.9% to 15.2%, P < .001 (Fig 1). The highest rates of OE were observed in children with immune (26.9%), cancer (24.0%), hematology (22.8%), and neurologic (17.8%) conditions. Specific, common examples of these conditions were hypogammaglobinemia (immune), benign neoplasm of skin (cancer), sickle cell anemia (hematology), and migraine headaches and cerebral palsy (neurologic). Also, based on the CCC classifications system, OE was high (22.0%) in children assisted with medical technology.

Of the 271 869 opioid prescriptions, 189 040 (76.7%) were dispensed as part of a single-fill episode, and the remainder were dispensed as part of 57 542 multiple-fill episodes (Table 1). Single-fill episodes had a median duration of 4 days (interquartile range [IQR]: 3–6 days), and multiple-fill episodes consisted of a median of 2 fills with a total median duration of 10 days (IQR: 7–19 days). Table 2 lists the most commonly filled opioids and the median total days supplied. Acetaminophen-hydrocodone comprised 47.5% of all opioid fills, followed by acetaminophen-codeine (21.5%), oxycodone (9.5%), acetaminophen-oxycodone (9.0%), and tramadol (6.2%).

Table 3 lists the top categories of recorded diagnosis codes preceding OE. Those diagnoses most frequently included infections (25.9%), injury (22.3%), nontraumatic musculoskeletal issues (9.6%), dental issues (6.7%), mental health issues (6.6%), gastrointestinal issues (3.7%), nontraumatic neurologic issues (1.6%), and malignancy (1.5%). Among specific types of infection, respiratory infections comprised the majority (64.1%). There was no diagnosis recorded for 12.9% of CYSHCN with an OE.

Approximately 86.3% of all OEs were preceded by contact with a health care provider (Fig 2). ED visits preceded 28.8% of all exposures, followed by outpatient surgery (28.8%), outpatient specialty (20.8%), primary care (18.3%), dental care (7.3%), and inpatient (6.5%) visits. Of the remaining 13.7% without care in the preceding 7 days, 12.0% had a “not elsewhere specified/injections/durable medical equipment” claim and the rest had no claims at all. Regarding follow-up care, 35.1% of single-fill OEs were associated with a follow-up visit within 7 days and 62.2% within 30 days, with the majority of follow-up visits occurring in the primary care or outpatient subspecialist office settings. Examining multiple-fill episodes only, this rose to 51.6% within 7 days and 74.7% within 30 days.

The findings from the current study suggest that CYSHCN in Medicaid may be exposed to opioids at a twofold higher rate than the general pediatric population.¹ Older, non-Hispanic white children were most likely to have OE compared with children of other race and/or ethnicities. The highest rates of OE were observed in children with immune, cancer, and hematologic conditions. Moreover, multiple chronic conditions, technology assistance, and polypharmacy were particularly associated with higher rates of OE. Consistent with previous literature, the majority of opioids were prescribed as a single-episode fill and for short durations, although ∼12% of prescriptions were multiple-fill episodes.²⁸ The types of opioids used were consistent with previous studies, with codeine, hydrocodone, and oxycodone representing the majority of prescriptions.²⁹ The most frequent preceding diagnoses were consistent with potential need for opioid therapy (eg, injury). Although >80% of all children had contact with a prescribing provider in the 7 days before OE, far fewer had subsequent follow-up at 7 and 30 days. Several of these findings warrant additional discussion.

First, further investigation is needed to assess the appropriateness of OE in CYSHCN. Many CYSHCN were exposed to opioids by health care providers, including ED clinicians and dentists, who, because of their typical role and setting in the health care system, may either not have been ideally positioned to achieve strong longitudinal knowledge of the children’s past medical history (ie, CCCs, polypharmacy, etc) or current overall health and well-being, or situated to provide follow-up.^30,31 These patient attributes may be particularly important to contemplate when choosing which medication(s) to prescribe for pain in CYSHCN. Although certain chronic conditions (eg, sickle cell anemia) may necessitate opioid therapy,³² a number of prescriptions were preceded by diagnoses not synonymous with pain (eg, respiratory infections). Recent US Food and Drug Administration warnings have been issued to minimize OE for acute respiratory infections and to reduce use of specific opioids in children (eg, codeine).^27,33,34 High rates of OE in the current study also occurred in children with complex neuromuscular conditions, such as cerebral palsy. These conditions may make it difficult to quantify and assess pain, especially by health care providers lacking clinical familiarity with the child and family. Real-world clinical data are needed to assess the medical decision-making that led to OE in CYSHCN, including which nonopioid medications for pain in CYSHCN were trialed in advance of the exposure.³⁵ 

Second, the high rates of polypharmacy observed in CYSHCN with OE in the current study merit additional exploration. Emerging evidence suggests that polypharmacy is prevalent among CYSHCN, especially those using Medicaid.³⁶ Although opioids themselves can result in dangerous ADEs, including digestive dysmotility, urinary retention, and respiratory depression, they also have the potential for drug-drug interactions with a variety of other medications.^16,37 For example, because of the risk for severe respiratory depression, coma, and death when opioids are used concomitantly with benzodiazepines, the US Food and Drug Administration has issued a black box warning against simultaneous administration of these medications. Despite this warning, some CYSHCN (eg, those with epilepsy and musculoskeletal spasticity) using benzodiazepines in the current study were also exposed to opioids. Additional study is necessary to assess whether those CYSHCN received prescriptions from multiple outpatient providers (eg, opioid from a dental or surgical provider and antiepileptic from a neurologist), potentially without concerted collaboration, decision-making, and monitoring. This practice could lead to concurrent therapy with contraindicated drug combinations. Such combinations, like opioids and benzodiazepines, may be clinically necessary but also require extremely close monitoring to avoid adverse events when used simultaneously. Although our study was not designed to look at concurrent medication use, future pediatric polypharmacy studies should investigate simultaneous exposure to opioids and other drugs, particularly to look for signals of downstream consequences.

Finally, <50% of children with multiple-fill episodes (indicating ongoing pain) had follow-up within 7 days. Although not assessed in the current study, it is possible that the pain experienced by these children resolved or that based on family or physician preference, follow-up occurred by phone or another communication mode that was not associated with a health care claim. However, because long-term opioid use often begins with treatment of acute pain, it is imperative that children who receive multiple fills of opioids are reevaluated in some manner for ongoing need versus discontinuation.¹³ In a recent study investigating persistent opioid use after surgery, ∼5% of opioid-naïve subjects filled additional opioid prescriptions >90 days after surgery, raising concern that acute exposure to opioids for postoperative pain management may be associated with a risk of long-term use.²⁹ When follow-up is indicated, improved strategies for ensuring follow-up are warranted.³⁸ Medical homes for CYSHCN may be best positioned to identify patients with prescriptions for new high-alert medications like opioids and then conduct follow-up evaluations either in person, by phone, or via the electronic health record. Centralized electronic prescription drug monitoring programs contain the necessary data to ascertain such patients.³⁹ Follow-up evaluations should minimize impact on the families of children with high medical complexity who may also have medical caregivers in the home who can relay information to the prescribing provider.

Our findings must be interpreted in the context of several limitations. First, our results were based on filled opioid prescriptions. Children may have used only a portion of the filled prescription, which would lead to an overestimation of risk due to opioids. We attempted to address this issue by separately analyzing children with multiple prescription episodes, which would indicate that they were using their dispensed medication. Second, a percentage of prescribing episodes were not preceded by visits with a prescribing provider. This indicates that children may have received opioids from prescriptions made >7 days before the fill date or from other modes (eg, called-in prescription without a health care encounter). In either scenario, this raises concern that children might be receiving opioids without a timely, in-person assessment of their pain symptoms. Third, because these were administrative claims data, we were not able to determine an individual child’s actual need for opioid pain control. We attempted to report prescription episodes associated with diagnostic codes for conditions indicating likely physical pain (eg, traumatic injury). For the majority of prescription episodes, however, we were not able to identify the reasons for opioid therapy, including their use in palliative and end-of-life care. Fourth, the continuous enrollment attribute of our study population led to underrepresentation of infants, which could have affected the demographic characteristics of the cohort as well as the rate and risk factors for OE in CYSHCN. Subsequent analysis of OE in neonates and infants is warranted.

These limitations notwithstanding, we hope that the current study advances knowledge and awareness of outpatient OE in the vulnerable population of CYSHCN. Medical home clinicians for CYSHCN in particular may leverage these findings to counsel CYSHCN and their families on the likelihood of OE, especially for children with attributes (eg, multiple chronic conditions and polypharmacy) associated with higher OE. To ensure the safety of OE in CYSHCN, the development of evidence-based opioid prescribing guidelines is paramount, including best practices for follow-up evaluation. As opioid prescribing guidelines are implemented, it will be important for subsequent studies to assess trends in OE over time for CYSHCN.

OE in CYSHCN is common, especially among those with multiple chronic conditions and polypharmacy. The findings may be useful to catalyze subsequent investigations on (1) the appropriateness of opioid use in CYSHCN, especially when prescribed from dental and ED settings; (2) opioid interactions with other chronic medications, especially in CYSHCN with polypharmacy; and (3) reasons for insufficient follow-up in CYSHCN prescribed opioids.

ADE

adverse drug event

AHRQ

Agency for Healthcare Research and Quality

CCC

complex chronic condition

CCI

Chronic Condition Indicator

CCS

Clinical Classification System

CYSHCN

children and youth with special health care needs

ED

emergency department

GPI

generic product indicator

ICD-10-CM

International Classification of Diseases, 10th Revision, Clinical Modification

IQR

interquartile range

OE

opioid exposure