Administrative hospital and mortality data can potentially address the paucity of high-quality RHD data globally. We identified systematic problems with RHD codes (ICD-10-AM I05-I09) where some cardiac valve conditions are coded as RHD even when medical charts do not specify rheumatic origin. We aim to develop approaches to improve identification of RHD in administrative data.
Phase 1 involved a chart review validating RHD-coded hospital admissions. Phase 2 included national consultation with government managers, coders, clinicians and epidemiologists to develop an algorithm to increase the specificity of codes used for identifying RHD. Phase 3 reviewed the impact of this algorithm on identification of RHD admissions from three tertiary hospitals.
A third of RHD-coded patients were identified as having RHD in the chart review, varying by age and high-risk status. Particular ICD-codes (I07/I08) were more likely to be false positives. The validation data, coding directives and clinical knowledge informed the consultation process, resulting in categorisation of codes as likely, possible or not RHD. When the new algorithm was tested on unlinked hospital data, RHD-coded admissions were 26.7% lower than using traditional RHD codes in 15-59year olds, 84.7% lower in ≥60yrs, and 21.9% lower in Aboriginal people. However, a review of linked hospital data showed that false positives were still likely.
Based on the findings, a more quantitative approach is being developed to improve case ascertainment of RHD using linked hospital/death data. Reliable RHD data are essential for monitoring progress towards goals to eliminate RHD in Australia.