This large, retrospective, observational study identified a significant clinical and economic burden in patients with mild-to-moderate AD treated at the secondary care level. Economically, direct medical costs were €1663 higher per patient over 5 years. At a population level in Sweden, a country with 10 million inhabitants, this translates to an aggregate (excess) societal economic burden of €118.9 million. Given the high prevalence of AD worldwide, this study provides important data on the clinical and economic burden of this common skin disease [
5,
18,
33].
A decline in secondary healthcare resource use following index date was observed in the present study and may indicate that patients’ symptoms improved over time, potentially through successful treatment regimens or spontaneous resolution of the disease in some patients. However, a recent study [
34] shows that disease persistence is high and that the decline in healthcare resource use may be a consequence of patients being returned to primary care for continuous management (and thus are not captured in the present study). The evolution of costs over time suggests that healthcare resource use is the most intense in the short term after onset of the disease or at relapse of symptoms.
Comparison with Previous Studies
The cost burden of pM2M AD patients reported across other studies varies substantially (between €49 and €1519), likely due to differences in study design (retrospective, prospective, and survey), study setting (inclusion of reference cohort), and type of cost (drugs, primary and secondary care visits, out-of-pocket expenses, indirect costs). These differences make comparisons between studies difficult. The estimated cost burden of pM2M AD in the present study appears to be within the range of cost burden reported in the published literature. A German study that included children used a similar study design as the present study (retrospective with 8-year follow-up), and reported an average annual cost for AD of €194 in mild patients and €417 in moderate patients [
20]. The average annual cost per patient (over 5 years) in the present study was €224. Another retrospective study in a French setting reported the average cost of medications per patient used during the first year of follow-up to be €175 in the eczema cohort compared with €99 in the reference individuals. The average cost of dispensed drugs in this study was €225 and €67 per patient in the pM2M AD cohort and reference cohort, respectively [
35]. Two Nordic studies (survey and prospective study) included cost of primary and secondary care visits together with cost of medications [
36] and indirect cost [
37], but neither differentiated results by severity. The survey study estimated the incremental cost of patients experiencing eczema symptoms at €410, while the prospective study estimated the incremental cost at €291. Finally, an Italian survey reported an annual average cost of €900 in mild patients and €1519 in moderate patients [
19].
Strengths and Limitations
A major strength of this study is the use of two large national databases with complete population-level coverage to identify the study population. This study, therefore, provides a complete account of healthcare resource use in secondary care and for dispensed drugs originating from both primary and secondary care.
Further, through the use of these national databases with long follow-up, it was possible to identify a large study population of pM2M AD patients for inclusion in this study (N = 87,721). At the same time, we used a look-back period of at least 3.5 years which increases the likelihood that the index date represents the initiation of newly clinically managed AD rather than a relapse of this chronic disease. Additionally, a matched reference cohort (non-AD) was identified from the national population register to allow for comparisons.
Three key factors contribute to possible bias in the study results. First, initial contact with the healthcare services in Sweden for mild, non-acute diseases is often in primary care. The present study collected primary care data for dispensed medications but not for diagnoses. Therefore, the complete economic burden of mild-to-moderate AD is likely higher than estimated in this study since this study does not include the cost of primary care visits and most patients with mild-to-moderate atopic dermatitis are treated by primary care physicians.
Second, the date for first dispensation of a TCS/TCI or a diagnosis in secondary care was used as the index date rather than the date of primary care diagnosis. This means that patients treated solely with emollients or lowest-potency TCS medications were not included in this study since these agents are available as over-the-counter drugs in Sweden and hence not captured by the PDR. From this perspective, the results in this study may therefore overestimate the economic burden since these “mildest” mild-to-moderate patients are likely to incur less than average healthcare resource use compared with the average mild-to-moderate AD patient. On the other hand, emollients are the basis of treatment for AD in all severities and the cost of these is not captured by this study.
Third, severity based on clinical assessment was not available in the administrative registers used in this study. Therefore, the definition of mild-moderate and severe AD was based on treatment class (topical, systemic, phototherapy) where information on potency and treatment concentrations was not available. This may have led to misclassification as we could not distinguish drugs with respect to potency within the class of TCS or concentrations within the class of TCIs since we only had access to the first three digits of the ATC codes in the database. Additionally, misclassification may have occured as mild-to-moderate patients with a flare may have been treated temporarily with systemic treatments or may have received phototherapy (this was not available in the database), or severe patients may not have received systemic treatment for unknown reasons. Although it was not possible to validate the severity classification, it follows treatment guidelines and the percentage of severe patients in this study (11.7%) is in line with other studies [
2].
In addition, date of death was not included in this study, and therefore patients were not censored at death. Since the study population consisted of pediatric patients, this likely had a minor impact on the study results. In addition, this study only accounted for direct health care costs and did not include other types of costs, such as indirect costs associated with loss of school days. Finally, another limitation in this study was the inability to verify the sensitivity and specificity of the inclusion and exclusion criteria employed to identify AD cases from prescription data. However, the algorithm used in this study was developed by specialists in dermatology, pharmacoepidemiology, and biostatistics and has been validated in a Danish setting, yielding a sensitivity and specificity of 74.1% and 73.0%, respectively [
38].