P225 Identifying asthma patients in Wales using latent class analysis of routine data. (12th November 2015)

• Record Type: Journal Article
• Authors: Al Sallakh, M; Rodgers, SE; Lyons, RA; Sheikh, A; Davies, GA
• Is Part Of: Thorax. Volume 70(2015)Supplement 3
• Page Start: A190
• Page End: A191
• Publication Date: 2015-11-12
• Subjects: Chest -- Diseases -- Periodicals; Thorax; Chest -- Diseases; Periodicals; Periodicals; 617.54
• DOI: 10.1136/thoraxjnl-2015-207770.361
• Languages: English
• ISSNs: 0040-6376

Abstract

Abstract : Background: The Wales Asthma Observatory aims to produce current estimates of asthma prevalence and disease burden using routine data. In the absence of a feasible gold standard to validate case definitions, latent class analysis (LCA) can be employed. Objectives: To estimate the prevalence of treated asthma in Wales using LCA of routine health data. Methods: We performed LCA using observed variables of asthma-related healthcare diagnostics and utilisation in the fiscal year 2011–2012 for a random sample of 98, 042 individuals in the Secure Anonymised Information Linkage (SAIL) databank. The observed variables were chosen if they exhibited expected distributions. Diagnostic performance of each of the observed variables was calculated. The model was tested for stability over multiple time windows and small area configurations. Since COPD can be misdiagnosed as asthma, a separate LCA was performed to identify COPD patients and cross-validate the asthma model. Results: Our LCA model estimated the prevalence of treated asthma in Wales in 2011–2012 as 8.9% (95% CI: 8.7%–9.1%), which was higher than estimates from the Quality and Outcome Framework (6.9%), but lower than both the prevalence of self-reported treated asthma estimated by the Welsh Health Surveys in 2011 (11.0%) and 2012 (10.0%) and the prevalence of 'GP reported and treated asthma' from the 'True Costs of Asthma in the UK' project (13.0%). In our model, prescription of any asthma medication had the highest accuracy among other observed variables (sens. = 99%; spec. = PPV = NPV = 100%), while asthma diagnosis variable had a lower accuracy (sens. = 66%; spec. = 94%; PPV = 51%; NPV = 97%). In the same sample, COPD prevalence was 2.0% (95% CI: 1.9%–2.1%) with only 2.8% of those classified as asthmatics were also classified as having COPD. Conclusion: Our LCA model provides a reasonable, data-driven, reference identification of people with treated asthma in Wales. Further work is needed to explore potential reasons for the observed differences in the estimates from other sources.