Objective. Although the efficacy of inhaled antiinflammatory therapy in improving symptoms and lung function in childhood asthma has been shown in clinical trials, the effectiveness of these medications in real-world practice settings in reducing acute health care use has not been well-evaluated. This study examined the effect of inhaled antiinflammatory therapy on hospitalizations and emergency department (ED) visits by children for asthma.
Design. Defined population cohort study over 1 year.
Setting. Three managed care organizations (MCOs) in Seattle, Boston, and Chicago participating in the Pediatric Asthma Care–Patient Outcome Research and Treatment II trial.
Participants. All 11 195 children, between 3 to 15 years old, with a diagnosis of asthma who were enrolled in the 3 MCOs between July 1996 and June 1997.
Outcome Measures. We identified children with 1 or more asthma diagnoses using automated encounter data. Medication dispensings were identified from automated pharmacy data. Multivariate logistic regression analysis was used to calculate effects of inhaled antiinflammatory therapy on the adjusted relative risk (RR) for hospitalization and ED visits for asthma.
Results. Over 12 months, 217 (1.9%) of children had an asthma hospitalization, and 757 (6.8%) had an ED visit. After adjustment for age, gender, MCO, and reliever dispensing, compared with children who did not receive controllers, the adjusted RRs for an ED visit were: children with any (≥1) dispensing of cromolyn, 0.4 (95% confidence interval [CI]: 0.3, 0.5); any inhaled corticosteroid (ICS), 0.5 (95% CI: 0.4, 0.6); any cromolyn or ICS combined (any controller), 0.4 (95% CI: 0.3, 0.5). For hospitalization, the adjusted RR for cromolyn was 0.6 (95% CI: 0.4, 0.9), for ICS 0.4 (95% CI: 0.3, 0.7), and for any controller 0.4 (95% CI: 0.3, 0.6). A significant protective effect for both events was seen among children with 1 to 5 and with >5 antiinflammatory dispensings. When the analysis was stratified by frequency of reliever dispensing, there was a significant protective effect for controllers on ED visits for children with 1 to 5 and with >5 reliever dispensings and on the risk of hospitalization for children with >5 reliever dispensings.
Conclusions. Inhaled antiinflammatory therapy is associated with a significant protective effect on the risk for hospitalization and ED visits in children with asthma. Cromolyn and ICSs were associated with similar effects on risks.asthma drug therapy, inhaled antiinflammatory agents, health maintenance organizations, hospitalization, emergency department.
- ED =
- emergency department •
- ICS =
- inhaled corticosteroid •
- MCO =
- managed care organization •
- PAC–PORT II =
- Pediatric Asthma Care–Patient Outcome Research and Treatment trial •
- NAEPP =
- National Asthma Education and Prevention Program •
- OR =
- odds ratio •
- RR =
- relative risk •
- CI =
- confidence interval
In children <15 years old, asthma is the most common cause of hospitalization other than infections1 and is responsible for 159 000 hospitalizations per year, with an average length of stay of 3.4 days.2 The rate of hospitalization for asthma among children 1 to 4 years old increased from 38.3 to 60.1 per 10 000 population between 1980 and 1992.3 Annual productivity losses caused by lost school days alone were estimated 10 years ago in 1990 at $1 billion.4 One of the national health objectives set in 1990 for the year 2000 was the reduction of asthma morbidity, as measured by a decrease in hospitalizations for asthma.5 This reduction has not been achieved.6
Childhood asthma may be largely preventable with current therapies.7 There is a substantial body of literature on the efficacy of inhaled antiinflammatory therapy in asthma using parameters such as lung function and symptom scores.8–10However, the clinical effectiveness of antiinflammatory medication in reducing hospitalizations and emergency department (ED) visits caused by asthma in children is not well-documented. These outcomes occur infrequently, so that the cost of a prospective randomized, controlled trial of adequate size makes it unlikely that a study of this type will be done.
In a retrospective observational study, Donahue et al11demonstrated a beneficial effect of inhaled corticosteroids (ICSs) and cromolyn on asthma hospitalizations in children and adults. A limitation of that study was that participants came from only 1 staff model managed care organization (MCO) and were residents of 1 geographic location. In addition, specific differences between health care use effects in children and adults were not addressed. In a Canadian study, Blais et al12 found a beneficial effect of regular ICSs on hospital admission rates in Saskatchewan but did not specifically consider the effect on children. This study also did not control for participants' area of residence, which may be associated with the number of available hospital beds and hence admission rates.13 Thus, the beneficial effect of ICSs may have been overestimated. Wennergren et al14 concluded that decreases in asthma-related hospitalization among children in Sweden between 1985 and 1993 probably resulted from increases in use of ICSs. These findings are limited by the use of an ecological rather than a cohort study design. None of these studies addressed the question of the effect of antiinflammatory medication on ED visits by children with asthma.
Therefore, although the efficacy of ICSs in children has been demonstrated in clinical trials, there are no effectiveness studies assessing the role of controllers on health care use specifically by children in the United States. An observational study design is a practical alternative to an experimental design in the setting of clinical practice, where it might otherwise not be feasible to study the effect of real-world medication use on hospitalization and ED treatment.15,16 This article describes a study in defined populations of children cared for in 3 geographically diverse MCOs participating in the multicenter Pediatric Asthma Care–Patient Outcome Research and Treatment (PAC–PORT II) trial of pediatric asthma outcomes. In this study the impact of asthma antiinflammatory medication on hospitalizations and ED visits for asthma was assessed in children over a 12-month period.
Data were collected as part of a larger study, the PAC–PORT II multicenter trial, which examines implementation strategies for the National Asthma Education and Prevention Program (NAEPP) guidelines for the diagnosis and management of moderate to severe asthma in children. All participants were members of 1 of 3 MCOs in 3 US metropolitan areas (Boston, Chicago, and Seattle) that collectively cover the care of 2 million people. Approximately 90% of all members have prepaid drug coverage that provides up to a month's supply of medicine for a nominal payment.
All 3 MCOs maintain computerized information systems that capture basic demographic data and claims files for all hospitalizations and ED visits. Automated pharmacy records contain detailed information on all prescriptions dispensed at outpatient pharmacies.
The study population consisted of all children enrolled in the study MCOs ages 3 to 15 years old with at least 1 diagnosis of asthma (International Classification of Diseases, Ninth Revision, Clinical Modification codes 493.00–493.99) listed for a hospitalization, ED visit, or ambulatory encounter (ie, provider-diagnosed asthma) from July 1996 through June 1997. Only children continuously enrolled in the plans for the 12-month study period who had prepaid prescription drug coverage were eligible for analysis. In addition, to avoid unstable or extreme rates, we excluded children with fewer than 30 days of person-time preceding their first hospitalization or ED visit during the study period.
For each type of drug, frequency of dispensing was calculated for each child by summing the number of canisters or containers of drug dispensed. Asthma antiinflammatory therapy was defined to include ICSs (referred to as inhaled steroids) and inhaled cromolyn and nedocromil (referred to as cromolyn). Oral antileukotriene preparations rarely were dispensed and were omitted from the analysis. Relievers included inhaled or pediatric oral β-agonist preparations (and anticholinergics) but excluded long-acting β-agonists such as salmeterol.11 Pharmacy data included initial dispensings and refills of all prescription medications.
Separate analyses were performed for each of the main outcomes of interest (ie, the first hospitalization or the first ED visit for asthma). The data were censored at the occurrence of the first hospitalization for that analysis and at the first ED visit for the ED analysis. Dispensings of medication or events occurring after the first events were not included in the respective analyses. ED visits were included in the analysis only if they occurred as primary events, not if immediately followed on the same or next calendar date by a hospitalization for asthma. The frequency of β-agonist dispensing served as a surrogate for disease severity and was used as the main variable for stratification in the analysis. This technique has been used successfully to control for disease severity and confounding by indication in previous studies.11,17 Information on race was unavailable for a substantial proportion of the population and therefore was not considered in this analysis. We have included all events and pharmacy dispensings that were billed to the MCOs in any way. This would not include items completely paid for by another insurer. Previous work from within these systems indicates the numbers of such occurrences to be very small.18
Differences in the proportion of children with hospitalizations or ED visits in each stratum were assessed for significance by χ2 tests and Mantel–Haenszel methods for analysis of 2 × k tables. Separate analyses were performed where reliever and antiinflammatory medication canister dispensings per year were collapsed into 2 (yes/no) and 3 strata (0, 1–5, >5). Age was also divided into 4 groups (3–5, 6–8, 9–11, and 12–15 years). All dispensings of antiinflammatory medications were weighted equally in these analyses (ie, no adjustment was made for different potency of ICSs). Multiple logistic regression was used to assess independent effects in models for hospitalizations and ED visits. Separate models were also developed for ICSs and cromolyn. Because previous analyses demonstrated that medication use is affected by age, gender, and MCO, all models included these variables. Because of concern that children who receive oral corticosteroids at an ED visit may be a source of confounding on the hospitalization analysis, the analysis was also performed with the data censored at the first occurrence of any event (ie, a hospitalization or an ED visit). We refer to odds ratio (OR) estimates from logistic regression models as relative risks (RRs) because the 2 measures approximate each other if the probability of the outcome is small.19 Effect modification was evaluated by stratified analysis and by inclusion of interaction terms in the logistic model. We also examined the effect of a drug-by-age interaction term in the models.
Approval for this study was obtained from institutional review boards at each participating institution.
During the 12-month study period, 11 195 eligible children 3 to 15 years old (mean age: 9.4 years; standard deviation: 3.6) were assigned an asthma diagnosis; 59% were male. Age and gender distribution were similar across MCOs. Overall, 39% of children received at least 1 dispensing of an antiinflammatory medication during the year; this increased to 60% in children also dispensed ≥3 relievers and to 77% in those also dispensed ≥6 relievers. Of all children, 12% were dispensed ≥5 antiinflammatory medications, which increased to 39% among children dispensed ≥6 relievers. During the 12-month observation period, 217 (1.9%) of children had an asthma hospitalization and 757 (6.8%) had an ED visit. Generally, hospitals did not make claims for ED visits that resulted in immediate hospitalization; therefore, <3% of ED visits in the automated databases were followed by a hospitalization on the same or next calendar date, and these ED visits were not included in the analysis. The frequency of events varied with frequency of reliever dispensing (Table 1). Both types of events were more common among children with no record of receiving any reliever and in children with frequent (>5) reliever dispensings than in midrange users (1–5 dispensings). Preschool-aged children (3–5 years old) were significantly more likely to have an ED visit (OR 1.6; 95% confidence interval [CI]: 1.3, 2.0; P = .0001) or a hospitalization (OR: 2.9; 95% CI: 2.0, 4.3; P = .0001) than older children.
Asthma drug use was strongly associated with the risk of ED visits and hospitalizations. After simultaneous adjustment for age, gender, MCO, and reliever dispensing, antiinflammatory use was associated with a significantly lower risk of both ED visits and hospitalization. The adjusted RR for an ED visit for those with any dispensing of cromolyn was 0.4 (95% CI: 0.3, 0.5), for any ICS it was 0.5 (95% CI: 0.4, 0.6), and for ICS or cromolyn (any controller) it was 0.4 (95% CI: 0.3, 0.5). For hospitalization, the adjusted RR for cromolyn was 0.6 (95% CI: 0.4, 0.9), for ICS it was 0.4 (95% CI: 0.3, 0.7), and for any controller it was 0.4 (95% CI: 0.3, 0.6).
Table 2 shows the adjusted RRs of events for cromolyn, ICS, and any controller within various strata of reliever dispensing. For ED visits (all P < .01) a significant protective effect was seen for any antiinflammatory therapy and for cromolyn among children with 0, 1 to 5, and >5 reliever dispensings. ICS also had a significant protective effect for those with 0 and >5 reliever dispensings. For hospitalizations, cromolyn, ICS, and any controller were associated with a significantly lower risk among those dispensed >5 relievers. In the no-reliever stratum, dispensing of any controller medication (ICS or cromolyn) was associated with a significantly lower risk for hospitalization, but the CIs for ICS and cromolyn when analyzed separately crossed unity. The effect of controllers on hospitalizations in the 1- to 5-reliever dispensing stratum was indeterminate.
Table 3 shows the RRs for events stratified by the frequency of antiinflammatory therapy dispensing, adjusted for age, gender, MCO, and frequency of reliever dispensing. For both events, a protective effect was seen for any controllers combined across the range of dispensing frequencies. For ED visits, there was a significantly lower risk for both 1 to 5 and >5 dispensings of cromolyn and for 1 to 5 dispensings of ICS. ICS dispensed on >5 occasions was associated with a lower risk, with CIs crossing 1. For both cromolyn and ICS, 1 to 5 dispensings were associated with a lower risk of hospitalization. The effect of >5 dispensings of cromolyn or ICS on risk of hospitalization was indeterminate. When the analysis was performed on the data censored at the first occurrence of a hospitalization or an ED visit, nearly identical results were obtained.
In these practice settings, children with asthma who are dispensed any antiinflammatory medication have a significantly lower risk of hospitalization and ED visits after adjustment for differences in the frequency of reliever dispensing, age, gender, and health care organization. This protective effect was seen across a range of frequencies of antiinflammatory dispensing and across disease severity (as measured by frequency of reliever dispensing).
Because we did not have an independent measure of severity, the frequency of reliever dispensing served as a surrogate. The risk of events was significantly higher for those with frequent (>5) dispensings of relievers, consistent with the notion that greater use of relievers is a marker of increasing severity.11,17 The adjusted analysis showed any antiinflammatory therapy to be protective for children across a wide range of reliever use.
Cromolyn and ICS had similar effects on the RRs of health care use. For all strata of reliever dispensing, any dispensing of cromolyn or ICSs had similar effects on health service use. When analyzed by frequency of antiinflammatory dispensing, a similar general reduction in risk for ED visits was seen for both cromolyn and ICS. For hospitalizations, a protective effect was seen with less frequent use of either therapy. Our results suggested that ICSs may be more protective with more frequent (>5) dispensings, although for both drugs the estimate crossed unity.
Our results support current NAEPP guidelines.7 Those needing only intermittent rescue medications may have mild, intermittent asthma. These children may benefit less from antiinflammatory medication in terms of reducing serious acute exacerbations leading to hospitalization than children with more severe illness, yet they derive significant benefit from antiinflammatory therapy in terms of reducing risk of ED visits. Children who needed reliever medications more regularly or who needed them intensively during certain periods such as particular seasons derived great protective benefit from antiinflammatory medications.
The indeterminate effect of more frequent use of cromolyn or ICS on hospitalizations when analyzed separately may have a number of explanations. Adjustment by reliever use only incompletely accounts for asthma severity, particularly in children. It is likely that some confounding by severity (or indication) will be present in the analysis because children with more severe asthma who are at higher risk for greater morbidity also are more likely to receive controllers and because we have incompletely adjusted for this factor. Another possible explanation is inadequate use of controllers by children who need more regular use or higher dosages,20 something that will be influenced by a number of modifying factors. These include the expertise of the physician in accurately classifying severity and consequently in prescribing medication appropriately;20the quality of the interaction between physician, parents, and children, which affects self-management skills and treatment adherence; and other patient-related factors such as underrecognition of symptoms. It should be noted that the estimate for both drugs combined (ie, any controller), with larger numbers contained in the analysis, showed a significant protective effect for both ED visits and hospitalizations.
Determining the optimal management approach to children with asthma remains difficult. The degree of inflammation is not well-correlated with clinical grading.21 König and Schaffer22 showed that children classified as having mild disease and treated with only β-agonists have lower eventual lung function than children with moderate illness managed with cromolyn. The logic of studies showing that earlier use of ICS is associated with better lung function outcomes is that, in persistent asthma of any severity, treatment with controllers should begin as early as possible.8 If the goal of treatment is to control inflammation so that eventual function is improved, then early treatment for those with mild persistent disease should be considered. The forthcoming results of a large multicenter trial examining the long-term effects of inhaled antiinflammatory therapy on children with mild to moderate asthma may offer additional guidance to clinicians in this area.23
Clinical trials comparing antiinflammatory efficacy have favored ICSs over cromolyn in reducing symptoms and improving lung function in children.20 Unlike with ICSs, there are no prospective studies of the beneficial long-term effects of early cromolyn therapy. However, in 2 recent retrospective studies, a delay in starting cromolyn had a negative effect on clinical outcomes and lung function in children with mild asthma.22,24 It is possible that different patient selection criteria are used by clinicians when selecting a particular patient for cromolyn or ICSs and that ICSs are more commonly used in children whose asthma is more difficult to control and are consequently at greater risk for morbidity. Cromolyn has an excellent safety profile, may be used in children as young as 2 years, and can be administered easily to young children through a nebulizer.25 Given the risks of long-term adverse effects and the uncertainty in young children as to whether asthma will persist, physicians and parents may find initial ICS use hard to justify. Our results encourage the use of cromolyn as first-line therapy, particularly in those classified with milder disease. Similarly, for physicians concerned about the evidence on the risks of delaying antiinflammatory treatment,8 early cromolyn use may be a good compromise. Although this approach is consistent with the current guidelines for asthma management, it is not current practice. In a previous analysis (submitted for publication) we have shown that all types of controllers are dispensed significantly less often to children 3 to 5 years old than to older children. In addition, only small numbers of children at any age are receiving any form of antiinflammatory medication sufficiently often to be considered regular users.
The use of computerized databases eliminates the risk of recall bias for drug exposure. However, we have had to rely on medication dispensing as a surrogate for actual medication use. Dispensing overestimates use of controllers and underestimates their effect. We do not have any measure of disease severity or control other than medication dispensing. As discussed earlier, there is also likely to be some residual confounding by asthma severity (or indication). Because more frequent dispensing of controllers probably is a marker of disease severity, this analysis may underestimate their protective effect. That is, it is unlikely that our proxy measure compensates completely for disease severity among recipients of antiinflammatory medications. Relying on dispensing records does not capture medications obtained by other means (eg, physician samples or from family members).
Children with no β-agonist dispensings probably are a heterogeneous group. For a number, this is a marker of incident disease for which the diagnosis is made at the time of a hospitalization or ED visit. Some of these children have at least moderate disease well-controlled by preventive therapy and need rescue medication only rarely. Others obtain rescue medications from other sources such as siblings and therefore in reality are part of the infrequent users group. The effects of these factors on the estimates in the analysis suggest that controllers may be more effective in reducing hospitalizations in the infrequent (1–5) reliever-dispensing group than they appear. This probable confounding does not substantially alter the general conclusion about the effectiveness of controllers.
The fact that most study participants received care from staff model MCOs may limit the generalizability of the study's findings to other care settings. The rate of hospitalizations among children in our study is substantially lower than that reported for children from low socioeconomic backgrounds26 and also lower than some reports for children in MCO settings, although it is similar to that of children in general.27 This is likely to reflect the greater risk of morbidity in socioeconomically disadvantaged children living in the inner-city or from minority backgrounds, compared with a general pediatric population with mostly mild asthma, all of whom have comprehensive health insurance. These differences in risk are likely to reflect disparities in affordability and access to care28,29 as well as other less well-defined features of the impact of socioeconomic disadvantage on health30 and differences in the organization and delivery of asthma care.31 The observational nature of the study design and the short period of the study follow-up (1 year) also limit the ability to draw conclusions about the effect of long-term antiinflammatory use. Since the study period, use of newer medications including higher-potency corticosteroids and oral antileukotriene preparations has become more widespread, and we have not been able to examine the effect of these medications on acute asthma events in children.
Children with frequent reliever dispensings are at high risk for acute use of health services, as are children with a diagnosis of asthma not receiving any therapy. The current NAEPP guidelines recommend that persistent asthma is most effectively controlled with regular antiinflammatory therapy.7 Our results show that antiinflammatory medications, both ICS and cromolyn, are effective in the practice setting in protecting against asthma exacerbations in children that lead to hospitalizations and emergency treatment.
The study was funded by Pediatric Asthma Care–Patient Outcome Research and Treatment Trial II Grant HS08368-01; the Agency for Health Care Policy and Research; the National Heart, Lung and Blood Institute; and Rhone Poulenc Rhorer Pharmaceuticals.
Dr Adams is a recipient of the Thoracic Society of Australia & New Zealand/Allen & Hanbury's Respiratory Research Fellowship.
Dr Fuhlbrigge is supported by a Mentored Clinical Scientist Development Award (1 KO8 HL03919-01) from the National Heart, Lung, and Blood Institute.
- Received March 17, 2000.
- Accepted August 3, 2000.
Reprint requests to (A.F.) Channing Laboratory, Brigham & Women's Hospital, 181 Longwood Ave, Boston, MA 02115. E-mail:
- Centers for Disease Control and Prevention. Vital and Health Statistics, National Hospital Discharge Survey. Annual Summary, 1993. DHHS Publ. No. PHS 95-1782. US Department of Health and Human Services, Public Health Service, National Center for Health Statistics; 1995
- Centers for Disease Control and Prevention. Asthma mortality and hospitalization among children and young adults, 1980–1993. MMWR Morb Mortal Wkly Rep. 1996:45:350–353
- Weiss KB,
- Gergen PJ,
- Hodgson TA
- Public Health Service. Healthy People 2000: National Health Promotion and Disease Prevention Objectives: Midcourse Review and 1995 Revisions. Washington, DC: US Department of Health and Human Services, Public Health Service; 1995
- Public Health Service. Healthy People 2000: Review–1998–1999. Washington, DC: US Department of Health and Human Services, Public Health Service; 1999
- National Asthma Education and Prevention Program. Expert Panel Report 2: Guidelines for the Diagnosis and Management of Asthma.Bethesda, MD: NIH/National Heart, Lung, and Blood Institute; 1997. Publ. No. 97-4051
- van Essen-Zandvliet E,
- Hughes MD,
- Waalkens HJ,
- Duiverman EJ,
- Pocock SJ,
- Kerrebijn KF
- Blais L,
- Suissa S,
- Boivon JF,
- Ernst P
- Goodman DC,
- Fisher ES,
- Gittelsohn A,
- Chang CH,
- Fleming C
- Suissa S,
- Blais L,
- Ernst P
- Rosner B. Fundamentals of Biostatistics. Boston, MA: PWS-Kent; 1990
- Halfon N,
- Newacheck PW
- Lozano P,
- Fishman P,
- Vonkorff M,
- Hecht J
- Copyright © 2001 American Academy of Pediatrics