Validación del diagnóstico de cáncer gástrico en las bases de datos administrativas del sistema de salud colombiano

Objective: To validate the diagnostic coding of stomach cancer in large Colombian administrative health databases.Methods: We used the capitation payment unit database (BDUPC, for its initials in Spanish) of the Colombian health system and the database of the Hospital Universitario San Ignacio as a comparator. A matching algorithm was developed to identify anonymized records in the BDUPC. Three validation methods were used for diagnostic coding: exploratory analysis of five diagnostic algorithms combining ICD-10, procedures, and medications; estimation of sensitivity, specificity, and predictive values for ICD-10 coding; and comparison of gastric cancer prevalence in Bucaramanga (Colombia) according to the BDUPC and the population-based cancer registry.Results: A total of 67,253 records with gastric cancer were identified in the BDUPC using the ICD-10 codification. The gastric cancer-specific ICD-10 code showed better performance than the combination of codes. The sensitivity of ICD-10 coding was 68.7% (95% CI=57.6%-78.4%), specificity was 99.9% (95% CI=99.91%-99.98%), and positive predictive value (PPV) was 86.4%(95% CI=76.4%-92.5%), whereas the PPV for anemia diagnosis was 54.1%. The ICD-10 code allowed for an accurate estimation of gastric cancer prevalence.Conclusion: ICD-10 coding for cancer has adequate specificity; however, if a small number of expected cases is obtained, it would be required to combine it with procedure codes to increase the number of cases. On the other hand, the difference in the performance of diagnostic coding for anemia indicates the need for independent validation for each disease in administrative databases.