Chronic Medication Use in Children Insured by Medicaid: A Multistate Retrospective Cohort Study

This was a retrospective cohort study of the MarketScan Medicaid Database (IBM Watson Health, Armonk, NY), which includes demographic and clinical data on paid Medicaid claims for all health care encounters across the care continuum. We included children ages 1 to 18 years continuously enrolled (≥11 months) in Medicaid or the Children’s Health Insurance Program fee-for-service and managed care plans from 10 states within all geographic regions of the United States in 2014. We required continuous enrollment because the database does not contain pharmacy claims that could have occurred during enrollment gaps. Neonates were excluded because of their unique enrollment and health care attributes. This study of deidentified data was exempt from review by the Colorado Multiple Institutional Review Board.

Children were classified into 5 groups of CM use based on their annual number of unique CMs: no medications, 1 medication, 2 to 4 medications, 5 to 9 medications, and ≥10 medications. To provide the most comprehensive description of CM use, we included all paid medication claims, including those for over-the-counter medications. CM was defined by using a previously established and commonly used method of relying on medication claims from existing pharmacoepidemiology studies: a filled ≥30-day prescription with ≥2 refills for the same generic drug at any time during the study period.⁴ We excluded the following items from medication counts: dental agents, devices and nondrug items, diagnostic agents, nutrition ingredients, and vaccines and toxoids.¹⁸ 

Medications were classified by using the American Hospital Formulary Service Classification Compilation as recorded in the 2008 Red Book.¹⁹ Three levels of classification were available, ranging from the highest-level aggregation “therapeutic group” (eg, central nervous system agent) to “therapeutic class” (eg, amphetamine stimulant) to “generic drug name” (eg, methylphenidate hydrochloride).

We reported demographic characteristics, including age, sex, race and/or ethnicity, and Medicaid eligibility. We identified chronic conditions in 2013–2014 on the basis of International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) diagnosis codes using (1) the Agency for Healthcare Research and Quality’s Chronic Condition Indicator system to identify chronic conditions of any complexity and to count the number of organ systems affected by multiple chronic conditions²⁰ and (2) the CCC system version 2 to identify children with CCCs known to be associated with high morbidity and resource use.²¹ 

Health care spending and use was assessed by using the number of CMs for 7 mutually exclusive service groupings: primary care, outpatient specialty, emergency department (ED), inpatient, mental health (inpatient and outpatient), pharmacy, and laboratory and/or radiographic testing.²² We reported the total spending, the median per-member per-year (PMPY) spending, and the percentage using each service associated with the number of CMs.

We used descriptive statistics to describe demographics and clinical characteristics of the study population, their annual CM use, and the most common CMs taken by the subjects. We used χ² tests to compare categorical variables of interest across groups of CM use and Cochran-Armitage tests to assess trends between categorical predictor variables of interest and ordered outcomes of CM groups. Analyses were performed by using SAS 9.4 (SAS Institute, Inc, Cary, NC), and P values <.001 were considered statistically significant because of the large sample size.

Of 4 594 061 children, 18.8% used CMs (Fig 1). Of CM users, 52.2% used 1 CM, 40.3% used 2 to 4 CMs, 7.0% used 5 to 9 CMs, and 0.5% used ≥10 CMs. At the highest-level medication classification, the central nervous system therapeutic group (eg, antidepressants, antipsychotics, stimulants, antiepileptic medications) was the most common group, comprising 28.9% of all CMs. The 10 most common therapeutic classes (of 197 total classes) accounted for 70.6% of all CMs; amphetamine stimulants (eg, methylphenidate), antihistamines (eg, cetirizine), and hypotensive agents (eg, clonidine, presumably used mostly for treatment of attention-deficit/hyperactivity disorder) were the 3 most common CMs (Table 1).

As the number of CMs increased, the diversity of medications increased. Whereas amphetamine stimulants and antihistamines accounted for 48.0% of all CMs for children using only 1 CM, no single medication classes in the ≥10-CM group accounted for ≥8% of CM use. Additionally, as the number of CMs increased, the CMs accounting for the greatest percentage of CM use also changed. Comparing the group with 1 CM to the group with ≥10 CMs, the percentage of children using amphetamine stimulants decreased from 29.0% to 4.0%, whereas the percentage of children using anticonvulsants increased from 2.3% to 7.1%. Other medications with known potential for adverse effects (eg, hypotensive agents, antidepressants, and antipsychotics) also increased in children with higher numbers of CMs.²³ 

Older age was significantly associated with higher numbers of CMs (Table 2). For example, the percentage of children with ≥2 CMs increased from 3.3% in 1- to 4-year-olds to 12.9% in 16- to 18-year-olds (P < .001). CM use also varied by race and/or ethnicity, with a higher rate of subjects with ≥2 CMs being observed in non-Hispanic white (10.6%) compared with non-Hispanic African American (7.5%) and Hispanic (4.6%) races and/or ethnicities (P < .001).

Most (84.0%) children identified on the basis of claims data with a chronic condition of any complexity used no CMs (68.2%) or only 1 CM (15.8%). For some chronic conditions, the majority of patients used CMs and frequently used multiple CMs (eg, immunologic, endocrine, renal, and neurologic), whereas for other conditions (eg, respiratory, musculoskeletal, and dermatologic), the use of multiple CMs was less common (Fig 2). For children with immunologic conditions, 17.6% used 1 CM, 30.9% used 2 to 4 CMs, 15.7% used 5 to 9 CMs, and 3.9% used ≥10 CMs. In comparison, for children with a dermatologic chronic condition, 15.3% used 1 CM, 12.4% used 2 to 4 CMs, 2.5% used 5 to 9 CMs, and 0.2% used ≥10 CMs. For subjects with multiple chronic conditions of any complexity, the use of ≥5 CMs increased from 0.7% for those with 1 chronic condition to 34.1% for those with ≥6 chronic conditions.

Most (72.4%) of the subset of children identified on the basis of claims data with a CCC had no (55.6%) or only 1 (16.8%) CM. For some CCCs, the majority of patients used CMs and frequently used multiple CMs (eg, transplant, technology assistance, respiratory, gastrointestinal, and neuromuscular), whereas for other CCCs (eg, cardiac and congenital and/or genetic), the use of multiple CMs was less common (Fig 2). For children with technology assistance, 15.1% used 1 CM, 27.2% used 2 to 4 CMs, 19.3% used 5 to 9 CMs, and 5.7% used ≥10 CMs. In comparison, of children with a congenital or genetic CCC, 15.9% used 1 CM, 17.8% used 2 to 4 CMs, 6.5% used 5 to 9 CMs, and 1.2% used ≥10 CMs. For subjects with multiple CCCs, the use of ≥5 CMs rose from 5.3% for those with 1 chronic condition to 36.9% for those with ≥4 CCCs.

Of the total $12.1 billion spent by Medicaid on the entire study population, pharmacy comprised the costliest health service category ($2.3 billion; 18.7%), which exceeded spending on inpatient ($2.0 billion; 16.7%), ED ($0.9 billion; 7.1%), and primary care services ($1.2 billion; 10.2%). The majority (59.3%) of pharmacy spending was attributable to the 7.6% of subjects with 2 to 4 CMs (38.6% of the pharmacy spending) and the 1.2% of total subjects with ≥5 CMs (20.7% of the pharmacy spending; Fig 1). Median annual pharmacy PMPY spending increased significantly with higher numbers of CMs, from $21 for no CM (IQR: $0–$107) to $9892 (IQR: $6081–$17 357) for ≥10 CMs (Table 3). For those with any CM use, 73.4% of pharmacy spending was attributed to chronic versus acute medications.

The median annual total PMPY spending across the care continuum per subject increased significantly with higher numbers of CMs, from $824 for no CMs (IQR: $351–$1616) to $32 460 (IQR: $15 090–$86 696) for ≥10 CMs (Table 3). The percentages of subjects using each type of evaluated health service increased significantly with higher numbers of CMs. Significant trends were observed from 0 to ≥10 CMs for increased use of ED care (32.1% to 56.2%), inpatient care (2.3% to 36.7%), home health (0.3% to 15.9%), and durable medical equipment (10.4% to 51.2%; P < .001 for all).

With nearly 1 in 5 children using a CM, of whom approximately half use multiple CMs, medication expenditures comprised ∼20% of total Medicaid spending, and the use of multiple CMs correlated with increased health care resource use. The most prevalent CMs across all children were central nervous system drugs, which are often implicated in studies of potential drug-drug interactions.²⁴ Additionally, the use of other medications with known potential for adverse effects (eg, hypotensive agents) increased in children with higher numbers of CMs.²³ Children with the highest number of CMs included those with immunologic, renal, and neurologic disorders; organ transplant; and/or assistance with medical technology. Health care resource use, including ED and hospital visits, and total health care spending increased substantially with higher numbers of CMs.

Our findings on the prevalence and types of CM use compare and contrast with the limited but emerging literature on CM use in children. Comparable with our results, authors of a recent national study reported that 22% of US children receive any medication annually, and 9% receive ≥2 medications; they did not distinguish between acute medications and CMs.² Previous studies of children have also revealed that amphetamine stimulants and antidepressants are among the most commonly used type of CM in children.^25,26 However, a substantially higher prevalence of children used those medications in our study. In a study of commercially insured children, the prevalence of stimulant medication use was 3% compared with 17% in our study of children insured by Medicaid.²⁵ Relatedly, compared with medication findings from the NHANES, we found threefold and fourfold increases in the use of antipsychotic and anticonvulsant medications in children with Medicaid, respectively.²⁶ These differences could indicate that children insured by Medicaid may have a higher penetrance and/or severity of chronic disease that necessitates greater use of medications.

To that end, our findings suggest that counting the number of CMs may be a simple, straightforward, and helpful method to identify and describe populations of children and to understand the scope of their health care needs and resource use. To our knowledge, most pediatric population health methods and efforts do not typically use CMs for those activities.^{20,21,27,–29} Rather, they rely predominately on information about chronic diagnoses. The use of CMs is important to consider, especially because many children in the current study with CCC diagnoses did not use any CMs. For example, 40% of children with a neuromuscular CCC, including cerebral palsy, did not use a CM. These findings suggest that there is substantial variation in CM use within the most complex pediatric populations that are typically targeted for case management interventions. CM use could be leveraged as an independent case-mix adjuster of medical complexity, increased health resource use, and pharmacovigilance needs in children.

For children with identified CM use, to maximize medication efficacy and safety and minimize associated costs, it is imperative to understand the appropriateness of prescribing practices for specific CMs. Is a child receiving the right medication for the right reason? Although administrative claims data are not designed to answer this question, they provide a population-level perspective to identify patterns of drug use requiring further investigation. In this study, clonidine use was substantial across all levels of CM use. Although clonidine is approved by the Food and Drug Administration for the treatment of attention-deficit/hyperactivity disorder, clonidine is frequently used off label to treat neuroirritability and behavioral and sleep issues in children with neurodevelopmental disorders.³⁰ State Medicaid programs may find it useful to partner with local medical and pharmacy-based clinicians to assess prescribing practices for specific medications, such as clonidine, that are used off label or without significant pediatric-specific evidence.^31,32 This type of assessment is critical to conduct ahead of implementing prescribing policies (eg, restricted formularies or limitations on the number of refills) that could interfere with children’s access to essential medications and adversely affect the health of children.

Finally, improved pediatric pharmacovigilance is needed to identify downstream consequences of CM use, especially for children with the highest numbers of CMs. Although a small population, these children accounted for a substantial percentage of pharmacy spending and total health care spending. Moreover, they had the highest rates of ED and inpatient hospital use of all children. In our clinical experience, these children often receive insufficient medication management characterized by uncoordinated, fragmented prescribing practices across multiple providers (eg, primary care and specialists) and settings of care. Additionally, providers without long-term continuity of care (eg, some ED clinicians and hospitalists) may prescribe new medications for these children. It is important to optimize the organization, roles, and responsibilities of clinicians who prescribe medications to children who are already on high numbers of CMs. Models of comanaged and integrated care between primary care providers, specialists, and pharmacists have shown promise in delivering high-quality prescribing and monitoring of CMs for adults.³³ It remains unknown how much these models have penetrated pediatric health care systems.

Our results must be considered in the context of several limitations. First, because some Medicaid enrollees may have supplemental commercial insurance or pay for certain prescriptions and over-the-counter medications out of pocket, we may have underestimated CM use. Accordingly, our findings generalize best to children with Medicaid-only insurance and may not generalize as well to commercially insured children. Second, a pharmacy claim indicates that a prescription was dispensed but does not specify actual use. However, the definition of a CM requires ≥2 refills, indicating that a CM was likely being used. Third, we studied annual cumulative exposure to CMs and not concurrent use.³⁴ This limited our ability to (1) determine if medication switching (eg, cetirizine changed to loratadine) versus add-on therapy contributed to increased CM counts and (2) perform analyses of potential drug-drug interactions. Fourth, chronic conditions and CCCs were identified from claims-based diagnosis codes without validation of actual disease, and these classification systems may have a wide range of sensitivity and specificity across different data sources. However, our percentages of CCCs are consistent with previous Medicaid studies using these classification systems.³⁵ Finally, the reported rates and trends of therapeutic medication groupings may be influenced by off-label use (eg, propranolol for migraine prophylaxis).

Nevertheless, our findings advance knowledge about CM use in children. The frequency and associated costs of CM use in children with Medicaid are substantial. Not all children with chronic conditions (including multiple chronic conditions and CCCs) use CMs. Increased numbers of CM use correlate strongly with increased health care resource use and spending. To best inform the development of local, state, and federal pharmacovigilance efforts, we must advance several aspects of pharmacotherapy. First, we must understand the best medication choices, optimal dosing schedules, and monitoring parameters for children with CM use. This includes improving our limited knowledge about the interactions of multiple drugs and their risks, particularly for medications whose mechanisms of action have known or postulated effects on other medications or direct effects on pediatric physiology. The Food and Drug Administration should require sponsors to look at the use of multiple medications across the pediatric age span to discern drug-drug interactions and subsequent effects on, for example, children’s developing central nervous systems, immune systems, or even bone growth.³⁶ Second, we must carefully evaluate potential strategies to manage CM use in at-risk children. Finally, we must assess the impact of these strategies on clinical outcomes, health care use, and the families and caregivers who administer complex medication regimens.^5,37 Together, these efforts will ensure the safest and best practices for CM use in children.

One in 5 US children enrolled in Medicaid are exposed to ≥1 CMs annually, and a smaller but more vulnerable group of children are exposed to significant chronic polypharmacy with unclear risks. CM use in children should be highly important to health care systems, payers, and policy makers as they strategize how to provide safe, evidence-based, and cost-effective pharmaceutical care to children.

CCC

complex chronic condition

CM

chronic medication

ED

emergency department

ICD-9-CM

International Classification of Diseases, Ninth Revision, Clinical Modification

IQR

interquartile range

PMPY

per member per year