Development and Evaluation of a Prediction Model for Ascertaining Rheumatic Heart Disease Status in Administrative...

Area of Science:

• Cardiology
• Public Health
• Health Informatics

Background:

• Concerns exist regarding the accuracy of International Statistical Classification of Diseases (ICD) codes for rheumatic heart disease (RHD).
• Misclassification of non-rheumatic valvular disease (non-rheumatic VHD) as RHD is a significant issue in administrative hospital data.
• A validated, quantitative method for RHD case ascertainment in hospital records is lacking.

Purpose of the Study:

• To develop and validate a predictive model for accurate RHD case ascertainment in administrative hospital data.
• To address and reduce false-positive rates caused by misclassified non-rheumatic VHD and acute rheumatic fever (ARF).

Main Methods:

• A dataset of validated Australian RHD cases was linked to inpatient hospital records (2000-2018).
• A generalized linear mixed model was developed using demographic and clinical variables.
• Internal and external validation was performed, calculating conditional optimal probability cutpoints.

Main Results:

• The model reduced the false-positive rate for non-rheumatic VHD misclassified as RHD from 0.77 to 0.22.
• It also reduced the false-positive rate for acute rheumatic fever (ARF) misclassified as RHD from 0.59 to 0.27.
• The model demonstrated strong discriminant capacity (AUC: 0.93 internally, 0.88 externally) and can function with basic data.

Conclusions:

• Misclassification of non-rheumatic VHD and ARF as RHD leads to substantial false-positive rates.
• The proposed model effectively addresses these biases, offering a reliable solution for RHD case ascertainment.
• This facilitates improved epidemiological disease monitoring and policy evaluation for RHD.