Validation of expert–algorithm agreement for initial risk stratification and dynamic response assessment in differentiated thyroid cancer
Article Sidebar
Main Article Content
Abstract
Risk stratification for recurrence and dynamic response assessment are central components of differentiated thyroid cancer (DTC) management. Their application requires the integration of multiple clinical, biochemical, pathological, and imaging variables, which may lead to interobserver variability. In this context, digital tools based on prespecified clinical algorithms could facilitate standardized and reproducible classification. The objective of this study was to evaluate the agreement between a digital clinical algorithm and independent expert classification in DTC. Materials and methods: this methodological validation study used 50 synthetic clinical cases independently assessed by 9 experts, all of whom had participated in the Argentine intersociety Delphi consensus on DTC. Overall, 450 expert-case evaluations were analyzed. The primary domains were initial recurrence risk and dynamic response to treatment; prognostic stage was a secondary domain. Agreement was assessed using observed agreement, unweighted kappa, linearly weighted kappa, and Gwet's AC1 coefficient. Ninety-five percent confidence intervals were estimated through cluster bootstrap resampling at the clinical-case level. Results: agreement for initial risk across four categories was 88.4% (kappa 0.840) and increased to 91.8% when grouped into three categories (kappa 0.876). Agreement was 87.7% for dynamic response (kappa 0.830) and 93.3% for prognostic stage (kappa 0.872). Linearly weighted kappa values ranged from 0.812 to 0.875. Disagreements occurred predominantly between adjacent categories, with no direct discordance between low- and high-risk classifications. Conclusions: The algorithm showed high agreement with expert classification. These findings support the feasibility of implementing complex clinical criteria through a structured digital tool and justify prospective evaluation in real-world clinical practice.
Article Details
Section
This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License.
EndoAr adopta una política clara y transparente respecto a los derechos de autor, alineada con los principios del acceso abierto y las buenas prácticas editoriales internacionales.
- Licencia Creative Commons CC-BY-NC (NonCommercial). Todos los artículos aceptados y publicados en la revista se distribuirán bajo la licencia Creative Commons Atribución-NoComercial 4.0 Internacional (CC-BY-NC 4.0). Esta licencia permite a cualquier persona: Compartir: copiar y redistribuir el material en cualquier medio o formato. Adaptar: remezclar, transformar y construir a partir del material. Siempre que se cumpla con las siguientes condiciones: Atribución: se debe dar crédito adecuado al autor original, proporcionar un enlace a la licencia e indicar si se realizaron cambios. No comercial: el material no puede ser usado con fines comerciales sin un acuerdo previo por escrito.
Esta política busca maximizar la difusión y el impacto de la investigación publicada, garantizando que los contenidos puedan ser utilizados y citados libremente en la comunidad científica, respetando los derechos de los autores.