Skip to main content
main-content

01.12.2017 | Research article | Ausgabe 1/2017 Open Access

BMC Medical Informatics and Decision Making 1/2017

Using medication utilization information to develop an asthma severity classification model

Zeitschrift:
BMC Medical Informatics and Decision Making > Ausgabe 1/2017
Autoren:
Tsung-Hsien Yu, Pin-Kuei Fu, Yu-Chi Tung

Abstract

Background

Claims data are currently widely used as source data in asthma studies. However, the insufficient information in claims data related to level of asthma severity may negatively impact study findings. The present study develops and validates an asthma severity classification model that uses medication utilization in Taiwan National Health Insurance claims data.

Methods

The National Health Insurance Research Database was used for the years 2006–2012 and included a total of 7221 patients newly diagnosed with asthma in 2007 for model development and in 2008 for model validation. The medication utilization of patients during the first year after the index date was used to classify level of severity, and the acute exacerbation of asthma during the second through fourth years after the index date was used as the outcome variable. Three models were developed, with subjects classified into four, three, and two groups, respectively. The area under the receiver operating characteristic curve (AUC) and the Kaplan-Meier survival curve were used to compare the performances of the classification models.

Results

In development data, the distribution of subjects and acute exacerbation rate among the stage 1 to stage 4 were: 62.71%, 5.54%, 22.79%, and 8.96%, and 8.17%, 9.55%, 11.97%, and 14.91%, respectively. The results also showed the higher severity groups to be more prone to being prescribed oral corticosteroids for asthma control, while lower severity groups were more likely to be prescribed short-acting medication and inhaled corticosteroid treatment. Furthermore, the results of survival analysis showed two-group classification was recommended and yield moderate performance (AUC = 0.671). In validation data, the distribution of subjects, acute exacerbation rates, and medication uses among stages were similar to those in development data, and the results of survival analysis were also the same.

Conclusions

Understanding asthma severity is critical to conducting effective, scholarly research on asthma, which currently uses claims data as a primary data source. The model developed in the present study not only overcomes a gap in the current literature but also provides an opportunity to improve the validity and quality of claims-data-based asthma studies.
Literatur
Über diesen Artikel

Weitere Artikel der Ausgabe 1/2017

BMC Medical Informatics and Decision Making 1/2017 Zur Ausgabe