Director Epidemiology GSK West Drayton, Uxbridge, United Kingdom
Background: Moderate asthma exacerbations are symptom-driven events which are associated with poor outcomes. However, they are difficult to identify in retrospective real-world data sources, such as administrative claims data.
Objectives: The aim of this study was to develop claims-based algorithm for identifying moderate asthma exacerbations in administrative claims data and assess its performance using electronic medical records (EMRs).
Methods: This was a retrospective, longitudinal cohort study using an administrative claims + EMRs linked dataset from Reliant Medical Group (Reliant) in Massachusetts, USA. Patients were included if they had ≥1 prescriptions for an asthma medication on or after 10/01/2016 (first prescription was defined as “the index date”), were ≥18 years old as of the index date, had ≥12 months of continuous health insurance coverage prior to index date and had ≥1 medical claims with a primary or secondary diagnosis code for asthma.
A definitional algorithm was developed based on claims operationalization of an international consensus clinical definition of moderate exacerbations: ≥1 outpatient visit, emergency department visit, or inpatient hospital stay of ≤1 day with asthma or asthma symptoms as the primary diagnosis and a pharmacy claim for a respiratory medication (except systemic steroids) and the definition of severe exacerbation (e.g., exacerbation requiring long-term hospitalization or treatment with steroids) not met.
Based on this algorithm, potential moderate asthma exacerbation events were identified in the health insurance claims database from Reliant and were flagged as ‘potential exacerbation events’. A random sample of 150 of these potential events was selected for chart abstraction among unique patients using a random number generation method. Abstracted charts were then reviewed and confirmed as ‘true exacerbations’ (i.e., true positive) or ‘non-events’ (i.e., false positive) by two independent clinical experts. Non-events were further classified as a true non-event (no exacerbation) or as a severe exacerbation.
To assess the performance of the algorithm, the positive predictive value (PPV) was computed based on the number of true positive and false positive events.
Results: Of the 150 events, 63 were validated as likely moderate exacerbations, resulting in a PPV of 42.0%. 42 (28.0%) events were validated as severe exacerbations and 42 (28.0%) were validated as non-events.
Conclusions: The relatively modest performance of the algorithm underscores limitations of using claims data to identify moderate exacerbations of asthma. The algorithm may show increased performance in identifying moderate and severe events combined together.
