Electronic health record-based assessment of oral corticosteroid use in a population of primary care patients with asthma: an observational study

The prevention of exacerbations of asthma symptoms is recognized in treatment guidelines as a key component of controlling asthma [1]. An exacerbation is broadly defined as an acute worsening of asthma symptoms requiring a transient change in treatment. There is, however, neither a uniformly accepted definition of an exacerbation nor a consensus on how an exacerbation should be measured [2].

The American Thoracic Society and European Respiratory Society (ATS/ERS) recommended in an official statement that the definition of a severe asthma exacerbation should include the use of systemic corticosteroids [2]. However, this definition applies only in clinical trials [2], and subjects enrolled in major randomized clinical trials of asthma medications are not representative of asthma patients in the general population [3]. Asthma patients in the general population are often studied retrospectively by way of administrative data sets in which surrogate variables are used as a measure of exacerbations.

Observational studies of administrative data sets have used oral corticosteroid prescriptions, either alone or as part of a composite measure with asthma-related emergency department visits and inpatient stays, as a measure of asthma exacerbations [4‐10]. The definitions of an exacerbation in these types of studies vary extensively. Finkelstein et al. used the “dispensing of an oral steroid preparation as a proxy for the occurrence of an acute exacerbation” [11]. Friedman defined an asthma exacerbation as an episode requiring “hospitalization, treatment in an emergency room, or an outpatient visit” where the patient “received nebulized medication or a prescription for oral corticosteroids” [12]. More recent studies have been more explicit in terms of days’ supply of oral corticosteroid and the timing of a prescription relative to an outpatient visit [7, 13, 14]. However, the lack of a standardized definition of exacerbations limits the applicability of observational findings to clinical practice.

The NIH and the Agency for Healthcare Research and Quality have drafted recommendations for the standardization of asthma outcome measures for both clinical trials and observational studies [15]. The recommendations define an exacerbation as a “worsening of asthma requiring the use of systemic corticosteroids to prevent a serious outcome” [16]. For exacerbations in patients twelve years of age and older, the core outcome measures for observational studies are the use of systemic corticosteroids (oral, injected, or IV), hospitalization, and emergency department (or urgent care) visits for asthma [16].

One limitation of using oral corticosteroids as a measure of asthma exacerbations is that administrative claims data sets typically do not link a claim for a prescription to the diagnosis prompting the prescription, making it difficult to determine whether a given oral corticosteroid prescription was written for an asthma exacerbation or for a condition unrelated to the patient’s asthma. Oral corticosteroids can be prescribed for variety of conditions unrelated to asthma, including systemic inflammatory diseases and pain syndromes. Integrated health care delivery systems that maintain electronic health records of all patient encounters contain the information needed to link a prescription order to the diagnosis assigned at the time the order was placed. Thus, with an electronic health record database, it is possible to determine whether an oral corticosteroid prescription was ordered for an asthma exacerbation or for some other reason.

The objective of this study was to use the electronic health records of an integrated health care delivery system to determine the medical diagnoses associated with oral corticosteroid prescriptions ordered for asthma patients, and thus to characterize the use of oral corticosteroid prescriptions as a measure of asthma exacerbations in a primary care setting.

In this population of primary care patients, over half (61.5%) of the oral corticosteroid prescriptions written for patients with asthma were prescribed for conditions unrelated to asthma (Table 3). This result suggests that, without an associated diagnosis, a prescription for an oral corticosteroid is not by itself an adequate marker for an asthma exacerbation in patients with asthma (but without other chronic respiratory diseases). This is consistent with evidence that a systemic corticosteroid prescription is poorly predictive of a diagnosis of asthma [17].

Oral corticosteroid use meets three of the four requirements for a core outcome measure in the NIH draft recommendations: it is clinically relevant, feasible, and enables comparison across studies [15]. The fourth requirement, evidence of validity, requires further research. In particular, the factors that contribute to patient and clinician decisions to use oral corticosteroids need to be investigated [16]. This study is, to our knowledge, the first to report the frequency with which asthma patients receive oral corticosteroids for conditions unrelated to asthma.

Across all age groups in this sample, most asthma patients were acute users of oral corticosteroids. A minority (12.5% overall) were considered chronic users. However, frequent exacerbations are difficult to distinguish from poorly controlled persistent asthma. The chronic asthma-related use by some patients—the number of prescriptions ranged up to 103, i.e., 8 years of 30-day prescriptions—may represent poor control of persistent asthma due to inadequate use of long-term control medications (inhaled corticosteroids, etc.) rather than exacerbations. (‘Chronic’ use of systemic corticosteroids is an option to control severe asthma [1]). Thus, only the acute use (which was 87.5% of all use) can reliably be related to exacerbations.

The principal limitation of this analysis is uncertainty about the diagnosis associated with prescription orders. Physicians are required to enter a diagnosis code on each prescription order, but there is no quality check to ensure that the selected code is accurate. As a result, misclassification of asthma-related versus non-asthma-related prescription orders might occur. A small number (9.4%) of the oral corticosteroid orders in the EHR did not have an associated diagnosis, and these were categorized as being non-asthma-related in our analysis. Given the potential for orders classified as upper respiratory but not asthma-related to alter our results, we conducted a sensitivity analysis classifying bronchitis as an asthma-related condition. The resulting proportion of oral corticosteroid orders that were not asthma-related declined from 61.5% to 55.0%, which still supports our conclusions. In addition, the appearance of ‘asthma’ in a patient’s problem list in an EHR has not been validated as a means of diagnosing asthma. Hence, it is possible that patients identified as having asthma solely on the basis of an entry in the patient’s problem list did not actually have asthma. The validity of the use of health administrative databases to identify asthma patients, however, has been studied. In this study, patients were classified as having asthma if they had ≥2 office visits within a 12-month period with an associated ICD-9 code for asthma. Gershon et al. (2009) reported that a similar algorithm (≥2 ambulatory care visits and/or ≥1 hospitalization for asthma in a 2 year period) had a sensitivity and specificity of 83.8% and 76.5%, respectively, with expert chart review as the reference standard [18]. The positive predictive value was 61.5% (but 72.5% with the primary care physician chart diagnosis as reference standard) [18].

This analysis did not account for the chronology of the oral corticosteroid prescription with respect to exacerbations. Unless oral corticosteroid orders were recorded only after a diagnosis of asthma was entered into the EHR, it is possible that patients did not have asthma when the oral corticosteroid was prescribed. One way to improve this study design is to define the chronology of the prescription with respect to the asthma-related outpatient visit, as was done by Schatz et al., who noted that some oral corticosteroids may be prescribed prophylactically rather than for a current exacerbation [7]. Another way is to impose a limit on the time between an outpatient visit for asthma and a prescription order related to asthma (e.g., ≤5 days), as was done by Lee et al. [19]. A second level of stringency could include defining the days’ supply of oral corticosteroid, as was done in two recent studies that defined exacerbations by oral corticosteroid prescriptions for <22 days’ supply [13, 14]. These approaches combined would ensure that the prescription was truly for an acute exacerbation and not the result of inadequate control of persistent asthma symptoms.

In this population of primary care patients, over half of the oral corticosteroid prescriptions for patients with asthma were for conditions unrelated to asthma. This finding has implications for how oral corticosteroid prescriptions should be interpreted in future observational research, particularly retrospective studies. In studies utilizing administrative claims data, a prescription for oral corticosteroids may be an unreliable marker of asthma exacerbations. Investigators should consider co-morbid conditions for which oral corticosteroid use may also be indicated. The accuracy of an oral corticosteroid prescription as a marker of an exacerbation should be evaluated, and algorithms to identify oral corticosteroid prescriptions ordered for acute asthma episodes in administrative claims data should be developed and validated.