Depression, Anxiety, and Emergency Department Use for Asthma

We used administrative data from the Massachusetts All-Payer Claims Database from 2014 to 2015. This comprehensive database provides data from all insurers: commercial, Medicaid, and Medicare. The Massachusetts All-Payer Claims Database was developed in 2009 and has been used in peer-reviewed literature and for population-based reports on health.^17,18  In addition, the Massachusetts Health Quality Partners (MHQP) database attributes individual national provider identifiers (NPIs) to clinics where the providers work.¹⁹  These data are unavailable in most states and allow for analyses clustered at the clinic level. The MHQP database is reviewed and updated annually by practicing physicians (J. Courtemanche, Vice President of Programs and Analytics, MHQP, personal communication, 2017). For those patients whose primary care provider’s NPI was not in the MHQP database (∼30% of the sample), we used patient 5-digit zip codes as the cluster.

Because our main predictors were anxiety or depression, we restricted the population to patients aged 6 to 21 years in 2015. This age range is consistent with previous claims-based studies of anxiety and depression, reflecting the low prevalence and greater challenges and variability in diagnosing mental health conditions in young children.^20,21 

Inclusion criteria for the CAPQuaM measure include the following^15,16 : (1) 3 months of consecutive enrollment in the same insurance plan (the measurement month and the 2 months before) and (2) evidence of claims for identifiable asthma (below) during a look-back period, including the measurement month, all previous months in the measurement year (2015), and the year (2014) before the measurement year (Supplemental Fig 2).

We identified eligible asthma patients using the CAPQuaM measure denominator definition.^15,16  Briefly, “identifiable asthma” was defined as administrative claims evidence of any of the following:

1. previous hospitalization with an asthma code as the first or second diagnosis per the International Classification of Diseases (ICD) Ninth or 10th Revision;

2. other qualifying events after the fifth birthday, including the following:

   • 1 or more previous ambulatory visit with asthma as the primary diagnosis,

   • 2 or more ambulatory visits with asthma as a diagnosis, or

   • 1 ambulatory visit with asthma as a diagnosis and at least 1 asthma-related prescription; or

3. other qualifying events, at any age, including the following:

   • 3 or more ambulatory visits with a diagnosis of asthma or

   • 2 or more ambulatory visits with a diagnosis of asthma and 1 or more asthma-related prescription.

Asthma-related prescriptions included long-acting β-agonist or inhaled corticosteroid, antiasthmatic combinations, methylxanthines, and/or mast-cell stabilizers. Patients meeting the sole criterion of using short-acting β-agonists were not included because these patients likely have less severe asthma, making asthma-related ED use potentially more difficult to prevent. Patients with any diagnoses of cystic fibrosis or emphysema were excluded.

Detailed measure definitions and codes are available at http://chipper.ucsf.edu/studies/implement/documents and www.capquam.org.

Patients with anxiety or depression were identified with at least 1 diagnosis-related ICD 9 or 10 code in inpatient, outpatient, or ED claims. Using a single code has demonstrated good negative (0.76) and positive (0.77) predictive values in administrative claims.²²  We did not use medications to identify patients because evidence-based therapy is often the recommended first line of treatment for pediatric depression or anxiety.^23,24 The time period used was the same rolling look-back period used for identifiable asthma criteria. The ICD codes for depression and anxiety reflect previous research²⁵  with some refinement by coauthors to align with Diagnostic and Statistical Manual of Mental Disorders, 5th Edition main diagnoses groups (Supplemental Table 3). We did not restrict our definition to patients with a depression or anxiety diagnosis before the index ED visit because there can be a lag between symptom onset and formal diagnosis.^23,24 

Our primary outcome used the CAPQuaM quality measure specifications, which measure an epidemiological rate: the number of asthma ED visits per 100 child-years of patients with identifiable asthma.^15,16  Quantitatively, it is interpretable as the number of ED visits that occur in 100 children with asthma who are in the plan for 1 year.

Eligible events for the outcome included an asthma-related ED visit or inpatient hospitalization. This reflects CAPQuaM team analyses that showed that the measure was more accurate when including hospitalizations because claims are often not submitted for ED care that leads to hospitalization.²⁶  To avoid double counting, an ED visit was excluded from the numerator if a hospitalization occurred within the same or the next (to allow for ED visits that cross midnight) calendar day. Per the recommendations of the CAPQuaM expert panel, an asthma diagnosis had to be the first or second diagnosis in the ED or hospitalization claim. We included claims with a second diagnosis of asthma because the primary diagnosis was often a related symptom (eg, fever or wheezing) or a known asthma trigger (eg, upper respiratory tract infection, pneumonia, or influenza).

To assess the independent association between the mental health diagnoses and asthma-related ED visits, we adjusted for age in 2015, sex, comorbid chronic disease status, and insurance type, which could vary each month.

To determine chronic disease status, we used the Pediatric Medical Complexity Algorithm (PMCA), categorizing patients into 3 categories using ICD codes: no chronic conditions, noncomplex chronic conditions, or complex chronic conditions.^27,28 To avoid collinearity in our modeling, we excluded from the PMCA diagnoses that overlapped with our populations of interest: anxiety and depression from the mental health category and the respiratory category, which included almost exclusively patients with asthma. Insurance type categorizations were provided in the Massachusetts All-Payer Claims Database. Commercial insurance was subcategorized as preferred provider organizations (PPOs) and health maintenance organizations (HMOs), and Medicaid was subcategorized as fee for service (FFS) and managed care.

Patients were attributed to their primary care provider by using a previously published algorithm.²⁹  The algorithm attributed a child to the NPI number for the claims on the majority (≥50%) of their well-child visits in a year. In case such a majority was not present, the child was attributed to the NPI number for the claim on their most recent well-child visit in that year. For children without any well-child visits for a given year, children were attributed on the basis of evaluation and management codes for other preventive visits.²⁹ 

In our multivariable statistical testing, we used negative binomial regression to accommodate the overdispersed distribution of the outcome. We controlled for age, sex, presence of chronic illness, and insurance type using robust SEs clustered at the clinic level, which accommodate correlation at the clinic or provider level. To assess bivariate associations, we used the same regression modeling approach but included only 1 predictor in each model. Adjusted rates for the multivariable analyses were obtained by using the postestimation margins command in Stata (Stata Corp, College Station, TX; Supplemental Information).³⁰  We tested for potential interactions with age, sex, and insurance and performed stratified analyses for statistically significant interactions.

We performed several sensitivity analyses. To assess whether insurance churn might drive the relationship between ED use and anxiety or depression, we analyzed a subpopulation of patients with 12 months of continuous insurance coverage. In addition, because ICD-coded data do not reliably identify asthma severity, we performed an analysis using short-acting β-agonist use alone (without controller use) as a marker of potentially milder asthma. To perform this, we defined the larger denominator using pharmacy claims, which increased our population ∼10% and decreased the ED visit rate (indicating a less severely ill population). We conducted multivariate modeling using the larger population, including (as a marker of asthma severity) a variable indicating patient inclusion based only on albuterol usage, and assessed for changes in results. In addition, because numerator events were identified by using asthma as the first or second diagnosis, we tested 2 alternative numerator definitions. We excluded all ED visits with depression or anxiety as a first or second diagnosis in case those diagnoses were driving increased use. We also separately reran the analyses using only ED visits with asthma as a first diagnosis in the numerator.

We used SAS version 9.4 (SAS Institute, Inc, Cary, NC) for data management and to calculate ED visits per member-month. All other calculations used Stata 13. The University of California, San Francisco’s institutional review board approved this study.

In the sample of 65 342 children with identifiable asthma, there were 6385 asthma-related ED visits from 2014 to 2015. Of our sample, 24.7% had a diagnosis of anxiety or depression or both, with 11.2% having a diagnosis of anxiety alone, 5.8% having a diagnosis of depression alone, and 7.7% having comorbid diagnoses of anxiety and depression (Table 1). There were 1750 patient clusters in the analysis, representing patients seen at an individual clinic or within a zip code (for patients whose clinicians were not affiliated with a clinic in the MHQP database).

The overall rate of asthma-related ED use was 17.1 visits per 100 child-years (95% confidence interval [CI]: 16.7–17.5). In unadjusted analyses, ED rates (number of visits per 100 child-years) were higher for children with a diagnosis of anxiety only (18.6; 95% CI: 16.6–20.6), children with a diagnosis of depression only (24.8; 95% CI: 20.7–28.8), and children with both anxiety and depression (30.5; 95% CI: 27.5–33.5) compared with children who had no diagnosis of anxiety or depression (15.2; 95% CI: 14.1–16.3; P < .001 for all comparisons; Table 2). Children who had depression, anxiety, or both had unadjusted ED visit rates of 23.6 (95% CI: 21.7–25.6).

These relationships persisted after controlling for age, sex, insurance status, and presence of chronic illness (Table 2, Fig 1). The rate of asthma-related ED use in multivariate analyses for patients with anxiety and depression (27.6; 95% CI: 25.0–30.5) was substantially higher than that of other vulnerable patients, including children with Medicaid managed care (21.7; 95% CI: 20.9–24.2) or Medicaid FFS (24.5; 95% CI: 22.6–27.8) and patients who had other complex chronic medical conditions (21.8; 95% CI: 20.1–24.7; Fig 1). The adjusted rate for patients with depression or anxiety or both was 22.1 (95% CI: 20.5–23.8). Interaction analyses found a statistically significant interaction by age group. However, stratified analyses by age group were qualitatively similar to the primary analysis (Supplemental Tables 4 through 6).

The relationships we found in the primary analysis between asthma-related ED visits and anxiety, depression, or comorbid anxiety and depression were robust to all sensitivity analyses except for the analysis defining ED events only by using asthma as a first diagnosis. For this sensitivity analysis, overall rates were lower. Relationships were similar, but with decreased sample size and power, statistical significance only remained for patients with both anxiety and depression and was borderline for those with depression (Supplemental Tables 7 through 10).

In this large, administrative claims–based analysis of children with asthma, we found that children with anxiety or depression had a substantially higher (1.2 to 1.8 times higher) rate of asthma-related ED visits compared with children who did not have anxiety or depression even after controlling for relevant covariates. This analysis is the first study to demonstrate an association between anxiety, depression, and asthma-related ED use in pediatric patients.

Mental health conditions, such as anxiety and depression, may have a particular impact on individuals’ or families’ abilities to manage asthma. Asthma self-management is complex, requiring recognition of symptoms, adherence to medication regimens, and avoidance of triggers. Likewise, depression and anxiety may require careful attention to manage.³¹  The cognitive load of managing multiple conditions can degrade the capacity to successfully manage 1 or both. The specific symptomatology of anxiety and depression may also make adherence more challenging, thus leading to more ED visits.²  In either instance, there may be a greater tendency to use the ED as a supportive resource even in the absence of a serious asthma exacerbation. Furthermore, symptoms of these mental health conditions may overlap with both asthma symptoms and side effects of asthma medications.^32,33 Thus, pediatric patients with these conditions may seek care not only for the exacerbations that result from nonadherence but for symptoms such as shortness of breath, rapid heartbeat, or chest pain, whose etiology may be ambiguous.^34,35 These factors likely explain the increased likelihood of ED visits we found in children with asthma and depression or anxiety.

We also found a high prevalence of anxiety and depression in children with asthma, with nearly 25% of our population having at least 1 of these. This is consistent with research showing that adults with chronic health conditions, and asthma specifically, have an increased risk for anxiety and depression.³⁶  In addition, our findings are consistent with adult literature showing that anxiety alone, depression alone, and anxiety and depression are associated with higher health care use.^10,11  This study highlights a population of children who may benefit from more intensive care coordination or case management for both asthma and a comorbid mental health diagnosis.

We found that the association of anxiety and depression with asthma-related ED use is even higher for patients who had both mental health conditions. Although anxiety alone or depression alone was independently associated with considerably higher asthma-related ED use, anxiety increased ED use less than depression did, and having both was associated with the highest ED use. This suggests a “dose-response” type of relationship between these diagnoses and asthma-related ED use. This finding is of particular importance given that almost 8% of our population of pediatric patients with asthma had comorbid anxiety and depression, a higher number than the 6% of pediatric patients with asthma and depression alone.

Given the increased risk of asthma-related ED use in children with anxiety or depression, more proactive approaches for this at-risk population may lead to reduced ED use. Strategies may include more intensive counseling to improve chronic medication adherence and symptom recognition for asthma, close outpatient follow-up for chronic issues,³⁷  and improved mental health care for these children, in whom untreated depression or anxiety may hinder asthma self-management. For example, medication adherence may not improve if there is untreated depression or anxiety. Identification of these children could occur in the primary care setting and would be increased through adherence to recommendations of the American Academy of Pediatrics to screen for behavioral health conditions and integrate behavioral health care into primary care,^38,39 perhaps in the patient-centered medical home model.⁴⁰  Improved care coordination and population health management may impact ED use and health care costs. Finally, future studies could focus on the extent to which these increased asthma-related ED visits in pediatric patients with depression and anxiety represent pulmonary exacerbations, increased somatic symptoms of anxiety and depressive disorders, or both. This understanding would allow more tailored and appropriate treatment of these patients in the emergency setting as well as strategies to enhance self-management of chronic conditions to prevent ED visits.

Our study had several limitations. We used data from a single state database, so asthma-related ED visit rates may not be generalizable to a national population, although the database includes claims from all payers. ICD codes may be limited in their accuracy and sensitivity for correctly identifying all patients with mental health diagnoses because providers may be reluctant to document these potentially stigmatizing diagnoses. This would lead to an underestimate of the number of children with anxiety or depression, although it is more likely to identify the more severe cases. ICD codes in administrative claims databases often do not accurately represent asthma severity, so we could not control for disease severity, although our results were robust to the sensitivity analysis treating albuterol use alone to indicate lower severity. The Massachusetts data we used do not include patient-level race and ethnicity, so we were unable to assess disparities by race or ethnicity, although evidence from Medicaid data in 29 states shows that race and ethnicity were not independently associated with ED use for asthma (use of a controller medication and racial segregation were associated).⁴¹  The CAPQuaM measure includes ED visits with asthma as the second diagnosis because many of these visits had an asthma exacerbation-related diagnosis (eg, fever, pneumonia, or wheezing) as the primary diagnosis, which may have allowed ED visits not driven by asthma to be included. However, our findings were robust to sensitivity analyses when only ED visits with asthma as a first diagnosis were included. Finally, it is possible that the increase in asthma-related use for patients with anxiety or depression may reflect overall higher ED use rather than asthma-specific increased use. This would not negate the main findings or recommendations from our study and could prove to be an important avenue for future analyses.

In summary, children with asthma and anxiety, depression, or anxiety and depression have meaningfully higher rates of ED use for asthma. Nearly one-quarter of children with asthma have such comorbidity, and those with increasing mental health complexity are at increasing risk. Subsequent work can test whether more intensive management and care coordination for pediatric patients with asthma and comorbid anxiety and/or depression might reduce ED use.

CAPQuaM

Collaboration for Advancing Pediatric Quality Measurement

CI

confidence interval

ED

emergency department

FFS

fee for service

HMO

health maintenance organization

ICD

International Classification of Diseases

MHQP

Massachusetts Health Quality Partners

NPI

national provider identifier

PMCA

Pediatric Medical Complexity Algorithm

PPO

preferred provider organization