Key result
Routine imaging and ECG models identify undiagnosed VHD with a C-statistic of ~0.86.
Why the study?
Valvular heart disease conveys high morbidity and mortality, making earlier diagnosis important given expanding therapeutic options, but the utility of routine healthcare data tools in community settings remained uncharacterized.
Do prediction models using routine healthcare data accurately identify undiagnosed valvular heart disease in community settings?
Meta-Analysis
Do prediction models using routine healthcare data accurately identify undiagnosed valvular heart disease in community settings?
Effect estimate: C-statistic 0.861 (95% CI 0.827-0.889)
Prediction models using routine healthcare data, particularly imaging and ECG-based tools, show strong potential for identifying undiagnosed valvular heart disease in the community, though prospective validation is needed before implementation.
Supports potential for routine imaging/ECG models to detect undiagnosed VHD; leaves open prospective validation before clinical adoption.
Background and Aims Valvular heart disease (VHD) conveys a high burden of morbidity and mortality and, given the increasing armamentarium of therapeutic interventions, earlier diagnosis is important. This meta-analysis aimed to evaluate the potential of tools in routine health data to identify individuals with undiagnosed VHD in community or population-based settings. Methods MEDLINE and Embase were searched from inception to October 6, 2025, for studies describing tools derived, validated or augmented to identify VHD in community populations. Potential measures (C-statistics) were pooled using Bayesian meta-analysis stratified by valve class and data modality with potential data from ≥3 cohorts, with heterogeneity assessed using 95% prediction intervals. Risk of bias was evaluated using the Prediction model Risk Of Bias Assessment Tool (PROBAST). Results Nineteen studies including 33 prediction tools were included; 18% had a high risk of bias. Imaging- and ECG-based models demonstrated excellent potential (C-statistic 0.861 [95% CI 0.827-0.889; I2 = 96%] and 0.830 [0.802-0.855; I2 = 100%]), while multivariable clinical models displayed good potential and biomarker-based models adequate potential (0.786 [0.663-0.872; I2 = 99%] and 0.699 [0.629-0.760; I2 = 0%]). Models targeting right-sided and mitral valve disease showed excellent potential (0.885 [0.837-0.920; I2 = 98%], 0.851 [0.819-0.879; I2 = 99%]). Models targeting aortic valve disease displayed good potential (0.797 [0.765-0.826; I2 = 100%]). External validation was limited, only one study assessed clinical utility, and no study prospectively validated a model. Conclusions Models using routine healthcare data to identify risk of undiagnosed VHD show potential but comprehensive external and prospective validation is required before consideration of implementation.
No takes yet. Share an insight, caveat, or question.
Helbitz et al. (2026) conducted a meta-analysis in Valvular heart disease (VHD). Prediction tools using routine healthcare data was evaluated on C-statistic for identifying VHD (imaging-based models) (C-statistic 0.861, 95% CI 0.827-0.889). Imaging- and ECG-based models using routine healthcare data demonstrated excellent potential for identifying undiagnosed valvular heart disease (C-statistics 0.861 and 0.830, respectively).
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: