Pioneering Methods Developed to Investigate the Burden of Acute Rheumatic Fever and Rheumatic Heart Disease Using Multiple Linked Data Sources

• Main Authors: Rebecca Seth, Daniela Bond-Smith, Judith Katzenellenbogan
• Format: Article
• Language: English
• Published: Swansea University 2020-12-01
• Series: International Journal of Population Data Science
• Online Access: https://ijpds.org/article/view/1632

Introduction Acute Rheumatic Fever (ARF) and Rheumatic Heart Disease (RHD) remain a major public health concern in Australia. Government action requires reliable burden estimates, however data from single or unlinked sources are only partial and likely to be skewed, exacerbated by systemic problems with ICD-10 codes for RHD. Linked data provide an opportunity to address these shortcomings. Objectives and Approach Objectives: to develop a methodology using harmonised linked data across five Australian jurisdictions to determine the burden of ARF and RHD <55 years, in particular robust case definitions for calculating incidence and prevalence. For identifying RHD in hospital-only patients, validated case and non-cases from non-hospital sources were used with linked inpatient hospital admissions to develop a RHD prediction model. Additional data sources (register and surgery databases) were used to identify cases for reporting RHD prevalence. A unique ARF episode was defined as an ARF record >90 days from the previous one across both register and hospital data. For first-ever episodes we applied a lookback to mid-2001 for both ARF and RHD. For Western Australia, we evaluated the effect of look-back period on prevalence pooling. Results For total ARF incidence over 3 years (2015-2017), there was 1425 episodes compared to 1027 episodes for first-ever ARF. There was an annual average of 5241 cases of RHD identified using our new methods (0-54yrs) – substantially higher than 2634 and 4255 RHD cases reported by Global Burden of Disease Study and Australian Institute of Welfare estimates respectively for 2017. Increased lookback had no effect on first-ever ARF but increased RHD prevalence >25 years. Conclusion / Implications By using multiple sources and cross-jurisdictional data we were able to provide contemporary and robust estimates for the burden of ARF and RHD in Australia. The prediction algorithm we developed can also be used in other countries, where only hospital data is available, to quantify RHD burden.