Introduction
Asthma is one of the most common chronic diseases, diagnosed in about 10 % of children and 4–5 % of the adult population in Germany [
1]. The economic burden for the German Statutory Health insurance has increased gradually from 2002 to 2008 up to €1,789,000,000 for the year 2008 [
2]. The treatment of asthma varies based on the severity of symptoms and disease manifestation. An insufficiently treated asthma patient can suffer from life-threatening asthma attacks with the need for emergency hospitalisation. It is generally accepted that both asthma burden, i.e. for patients in terms of quality of life etc., and treatment costs increase with asthma severity and insufficient control [
3]. Substantial economic evaluation of asthma costs requires knowledge of asthma severity, which is generally assessed by using clinical information from the patient. Asthma is a heterogeneous disease whose symptoms can vary over time, and that can change rapidly from day to day. Given that the disease is well-characterized in some patients, the relationship between the underlying disease processes and their clinical manifestations may not be strong. This issue poses a challenge regarding how patients with asthma should be diagnosed and assessed, and how treatment should be adjusted [
4]. The concept of asthma severity itself has evolved substantially over the years. Previous Global Initiative for Asthma (GINA) guidelines have differentiated asthma severity into four categories: intermittent, mild persistent, moderate persistent, and severe persistent, referring to the clinical characteristics before treatment and the magnitude of disease features such as the severity of airway obstruction [
5]. A patient’s treatment is decided based on this severity classification. As the clinical perspective of asthma has been refined over the years, now focussing more on asthma control rather than on severity, the assessment of severity from a health economic perspective is still of importance given the possibilities of disease management [
3]. In general, severity reflects the underlying disease manifestations and thus helps targeted treatments. Furthermore, maintaining a concept of asthma severity includes the option of referring to patients with whom asthma management is challenging either due to poor adherence or, although being adherent, requiring high-intensity treatment [
4]. These patients absorb a high proportion of asthma health resources, which is relevant from a health economic perspective.
Hence, not only is the level of asthma control important in terms of the treatment required to achieve adequate asthma treatment, but also the corresponding asthma severity.
Claims data offer important advantages for economic evaluations by providing observational information for a large number of patients, which reflect decisions made both by health care providers in routine clinical practice and by patients with regard to prescription fills and use of inpatient and outpatient care [
6]. German claims data include information on an individual patient level such as: biographic data (e.g. age, gender, etc.), healthcare resource utilisation and direct healthcare costs for outpatient and inpatient procedures, drugs, devices and aids, occupational therapies, sick leave payments (with reason) and early retirement. German healthcare insurances cover the most health care services, resulting in only marginal patient co-payments. Healthcare provider payments on the expense of sickness funds (hospital, physician, or pharmacist) represent almost the complete direct health care costs on an individual basis. Due to federal data protection laws, claims data do not include direct clinical data input, such as measures of lung function, forced expiratory volume (FEV) or peak expiratory flow (PEF). Considering that no direct clinical data is captured in claims data, methods to identify different disease severities and disease worsening are needed in order to be able to use this data source for economic evaluation [
7]. A variety of algorithms has been developed over the past two decades to fill this gap. Healthcare Effectiveness Data and Information Set (HEDIS) is a quality measurement program from the National Committee for Quality Assurance developed on a claims data based definition of persistent asthma. This definition relies on asthma-coded medical visits and asthma-related pharmacy claims. According to this definition, a population can be identified for whom asthma controller therapy is indicated [
8]. In order to be identified as a persistent asthma patient, one or more of the following criteria must be met for the current year: at least one emergency department (ED) visit with asthma as the principal diagnosis, or at least one acute inpatient claim/encounter with asthma as the principal diagnosis, or at least four outpatient asthma visits with asthma as one of the listed diagnoses and at least two asthma medication dispensing events, or at least four asthma medication dispensing events [
6,
9,
10]. Recent publications have modified the HEDIS criteria to a 2-year timeframe for the assessment of the above described criteria [
11‐
17]. Although the HEDIS criteria was first used for claims data studies by Berger et al. [
18], a validation of the criteria was lacking until 2010. Schatz et al. [
8] used survey data including medication use, asthma symptoms and the presence of exacerbations to validate the HEDIS criteria.
The Leidy method [
19] determines mild persistent asthma based on the frequency of claims for β
2-agonist combined with the frequency of claims for oral corticosteroid prescriptions (OCS). Mild persistent asthma is defined by four to six short-acting β
2-agonist (SABA) refills and zero oral OCS prescriptions per year, or two to three SABA refills and less than two OCS prescriptions per year. Furthermore, one (or less) SABA refill and one oral OCS prescription per year can also account for mild persistent asthma. Moderate persistent asthma includes more than six SABA refills and less than two OCS prescriptions per year, or four to six SABA refills and one to two OCS prescriptions per year. Patients with severe persistent asthma are required to have more than six SABA refills per year and the number of OCS prescriptions per year is greater than or equal to two. Moreover, zero to six SABA refills and three or more SABA prescriptions per year also constitute severe persistent asthma. Clinical validation of the Leidy criteria is warranted [
19].
The current GINA guideline provides recommendations for categorizing levels of asthma control. However, previous GINA documents have subdivided asthma by severity based on the level of symptoms, airflow limitation, and lung function variability. Four categories were included: intermittent, mild persistent, moderate persistent, and severe persistent [
20]. The daily dose of inhaled corticosteroids (ICS) and long-acting β
2-agonist LABA were divided into low and high intensity treatment. Mild persistent asthma was defined by either using low-dose ICS consistently, or using ICS inconsistently, including zero to two claims. Patients with moderate persistent asthma were defined as such if they received low-dose ICSs and either a LABA, a leukotriene modifier, theophylline or medium- or high-dose ICSs. Severe persistent asthma was defined by the use of medium- or high-dose ICS plus a LABA along with other controllers [
20]. A validation of the GINA based claims data algorithms is still lacking.
The Canadian Asthma Consensus Guideline (CACQ)-based database indexes were developed and validated by Firoozi et al. [
21]. The severity index defines three levels of asthma severity by assessing asthma medication and the presence of moderate/severe asthma exacerbations over a period of 1 year. Patients in the mild asthma category are supposed to show no presence of moderate/severe asthma exacerbations over a period of 1 year, receive ICS doses of 0–500 µg/day with no additional controller therapy or, for patients with additional controller therapy, a dosage of 0–250 µg ICS per day. Moderate asthma is classified by ICS doses of >500 µg/day for patients without additional controller therapy, and doses of >250 µg/day for those with additional controller therapy. Patients with high use of SABA and moderate or severe asthma exacerbations are also classified as moderate asthma. The category of severe asthma consists of individuals receiving ICS doses of >1000 µg/day, or >10 doses of SABA per week, with moderate/severe exacerbations. The CACQ database indexes were validated against pulmonary function test results of a sample of 71 randomly selected asthma patients. Patients were recruited from two asthma clinics and medical chart reviews were used to validate the CACQ database indexes against FEV
1 values [
21].
The aim of this study was to systematically review the international literature to assess if the already existing algorithms are applied for the stratification of asthma patients according to disease severity based on available information in claims data. Furthermore, potential best practice standards are identified and their transferability to the German setting was discussed.
Discussion
The systematic literature search yielded 54 publications that evaluated asthma severity based on claims data, despite the fact that clinical data is missing in this data source [
40]. Different approaches have been developed over the last two decades to overcome this limitation. Previous work has shown that claims-data-based instruments are feasible to assess quality-of-care [
33], and that algorithm-based severity categorisation is possible [
6]. Claims data analyses provide relevant observational information for a large number of patients, reflecting real-life treatment patterns [
40,
66]. The reviewed literature suggests that previously described algorithms such as HEDIS, Leidy and CACG are used widely but no best practice for the identification of disease severity in asthma patients using claims data has been established so far. Also, the HEDIS criteria was applied in 31 publications, but a more differentiated look at the most recent publications indicates that alternatives are still of interest. In the timeframe of the most recent 5 years (2011–2015), six publications used the HEDIS criteria whereas five publications used other algorithms. Expanding the timeframe to the most recent 6 years shifts the result in favor of other algorithms than HEDIS (11 other vs. 10 HEDIS). HEDIS relies on asthma claims coded at ED visits, hospitalisations, outpatient visits, or SABA prescription fills, which is a commonality also found in Leidy’s algorithm. As Birnbaum et al. [
6] state, this medication-derived method can categorise patients as having more severe asthma than the symptom-derived methods based on clinical data. An analysis in children with asthma suggested that HEDIS criteria for persistent asthma is very sensitive, but has relatively low specificity; hence, it might misclassify patients with intermittent asthma as having persistent asthma [
67]. To avoid this possible misclassification, Leidy’s criteria was applied to exclude patients who might have intermittent asthma, by incorporating minimal requirements for the number of SABA claims to be identified as persistent. Thus, Leidy’s algorithm is commonly used as an additional secondary screen to the HEDIS criteria, when classifying patients with mild persistent asthma [
10]. The claims-data-based GINA criteria—an approach used in combination with HEDIS and Leidy—provides recommendations based on the daily dose of ICS and LABA, but is less specific than HEDIS and Leidy when comparing the requirements for asthma medication use based on claims data. However, the GINA guideline is considered the gold standard in clinical practice for the assessment of disease control. In contrast, Leidy’s criteria refers only to SABA use, which does not include inhaled corticosteroids, present in the former GINA guideline [
20]. The CACQ database indexes were used as standalone classification for asthma severity, also incorporating asthma control. These indexes use a comprehensive matrix of criteria, including daily ICS dose, weekly SABA dose, other controller medication and markers of moderate/severe exacerbations to assess asthma severity [
21]. Thus, they are more complex then HEDIS, LEIDY and GINA. So far a validation for HEDIS, Leidy and CACQ is warranted.
One objective of this review was the evaluation of a potential replication of the identified algorithms to the German setting. Most of the studies presented here were conducted in the United States, where a different health care system and coding system is in place, which makes the assessment of a possible transfer to a German setting even more important. Claims data from the German Statutory Health Insurance are collected primarily for the purpose of reimbursement and documentation. Clinical data, and also information about the intention of the physician, e.g. prescribed dosage and frequency, is missing. This limits the potential of dosage-based classifications such as the GINA-based approach.
In Germany, information on diagnosis in the outpatient sector is given only on a quarterly basis. In contrast, medical services are recorded on a daily basis. Therefore, it is not possible to exactly match a diagnosis with a specific outpatient visit. The limitation also takes effect on the identification of outpatient emergency cases [
7]. To apply the HEDIS criteria to German claims data is not completely possible due to the quarterly documentation of outpatient diagnoses. Each diagnosis is only recorded once a quarter for every physician the patient consulted. Especially, the identification of at least four outpatient visits with asthma listed as a diagnosis poses a challenge, since a patient would be required to have an outpatient claim in each quarter or with different physicians to amount to four claims for asthma. Furthermore, the analysis of an emergency case with an asthma diagnosis is possible only for the inpatient setting. Emergency cases in the outpatient setting might be misclassified in terms of multi-morbid patients, due to the quarterly documentation of outpatient diagnoses.
Leidy’s criteria assess asthma based on requirements for the amount of asthma-specific prescriptions per year. Mild persistent asthma is defined by four to six SABA refills and zero oral OCS prescriptions per year or two to three SABA refills and less than two OCS prescriptions per year. This specific algorithm can be applied to German claims data, as the medication prescriptions are documented and can be assessed.
The former GINA guideline provides recommendations for categorising mild or severe persistent asthma based on the daily dose of inhaled corticosteroids and at least a second controller, i.e. LABA, LTRA, Theophylline, and OCS. These criteria cannot be transferred to German claims data without inaccuracy, as the daily dose can only be estimated. The data does not include prescribed dosage information, which might modify the use of ICS and salmeterol in a specific case [
7].
The studies that did not refer to the algorithms mentioned above were categorized based on the severity assessed. In total, 16 publications were stratified to mild, mild intermittent, intermittent, moderate, persistent, mild persistent, moderate persistent, severe persistent, severe and low/high-risk asthma. The basis for the inclusion of persistent asthma patients were mostly asthma service claims. The publications varied especially in specificity concerning the amount of prescriptions for asthma specific medication and where asthma claims needed to be coded, i.e. inpatient or outpatient sector, or ED visit [
43,
55,
61]. Publications assessing mild asthma with various criteria excluded asthma exacerbations, mostly defined as an asthma episode that required hospitalisation, an emergency department visit, or an outpatient visit in which patients received nebulised medication or a prescription for OCS. Moreover, similar to the partly medication-derived algorithms from HEDIS and Leidy, asthma-specific medication, or an overreliance on SABA, were considered an indication for a higher asthma severity [
49‐
51,
55]. Publications determining moderate to severe persistent asthma focussed on asthma-specific medication, such as fluticasone propionate/salmeterol, albuterol, and levalbuterol, which are β
2-agonists. Furthermore, the number of prescription fills for inhaled corticosteroids was considered an important identification criterion for more severe asthma [
43,
55].
The evaluation of methods applied suggests that asthma severity in administrative data is connected with claims for asthma and asthma-specific medication, varying by the type of therapy received. Claims for oral or inhalative corticosteroids are associated with higher disease severity, whereas mild asthma is associated mostly with restricted use of short-acting β2-agonists.
Due to the fact that the identified algorithms have commonalities with the specific algorithms from HEDIS and Leidy, transfer to the German context is possible with a few restrictions. As already mentioned, physician contacts and emergency cases in the outpatient setting cannot be accurately connected to a specific ICD-10-GM diagnosis code. Furthermore, the severity categorisation based on medication use can be applied if the use does not refer to daily doses but instead to the number of prescriptions. It should be noted that claims for prescriptions dispensed can be imprecise as the data identifies only that a canister was dispensed by a pharmacy—regardless of whether the medication was actually used by the patient [
27].
Conclusion
The results of this systematic review suggest that there is no best practice method for the categorisation of asthma severity grades with claims data. Also, although HEDIS is used in the majority of studies, this is more heterogeneous for the most recent publications (2010–2015). Rather, a combination of the specific algorithms seems to be a pragmatic approach. Furthermore, it should be noted that, by the date of the systematic search, only one study was identified that used a study design similar to the German context. The analysis of the specific algorithms indicates some limitations, which might lead to a misclassification of asthma severity if only a single algorithm is applied. A factor common to the assessed algorithms, both specific and unspecific, is that they refer to either asthma-specific medication and/or claims in the inpatient or outpatient sector. It should be noted that the studies vary in the amount of necessary prescriptions for asthma specific-medication and claims in the inpatient or outpatient sector. The transfer to a German context is not entirely possible without considering particular conditions associated with German claims data, especially in the outpatient sector. Nevertheless, as claims data has important advantages based on the observational information for a large number of patients, which also accurately reflects the resource use and costs of a disease, these algorithms could be modified and applied to the German setting and provide an approach for a health economic evaluation.