Related Experiment Video For Systematized NOmenclature of MEDicine codes
Updated: Apr 16, 2026
Comparison of Predictive Performance of Three Lymph Node Staging Systems in Colorectal Signet Ring Cell Carcinoma Based on Machine Learning Model
Published on: April 18, 2025
Development and Validation of an Algorithm to Achieve Complete Registration of Danish Patients with Carcinoma In Situ
Martin Kjær Simonsen1,2,3, Trine Lønbo Grantzau4, Maj-Britt Jensen5
1Department of Pathology, Aarhus University Hospital, Aarhus, Denmark.
Purpose:
Registration of carcinoma in situ (CIS) lesions in the national Danish Breast Cancer Group (DBCG) database is incomplete. We aimed to develop and validate a computer-based algorithm designed to identify patients with pure CIS of the breast, defined as CIS without concurrent breast cancer (BC), and with no history of BC or ipsilateral CIS. A secondary aim was to assess the completeness of CIS registration in the DBCG database.
Patients And Methods:
We developed an algorithm to identify patients diagnosed with pure CIS between 2008 and 2017 using Systematized NOmenclature of MEDicine codes from the Danish Pathology Registry (DPR). We manually reviewed pathology records of a subcohort from Aarhus University Hospital (AUH) to serve as reference standard. To test reproducibility, the algorithm was reapplied to DPR data spanning 2008 to 2023.
Results:
Between 2008 and 2017, 13,720 patients diagnosed with CIS of any kind were identified in the DPR. Of these, 1,581 patients were diagnosed at AUH, and 572 of those were manually confirmed as cases of pure CIS. The algorithm correctly classified 560 of these cases, corresponding to a sensitivity of 97.9% (95% CI: 96.3-99.3%). The positive predictive value of the algorithm for identifying pure CIS was 100.0% (95% CI: 99.3-100.0%). On a national level, the algorithm identified 4,302 patients with pure CIS between 2008 and 2017. Of those, 1,002 (23%) were not registered in the DBCG database. In the extended study population, 7,206 patients with pure CIS were identified, and 4,266 (99.2%) cases from the original study population were reclassified as pure CIS. Of 7,206 total patients, 1,827 (25%) were not registered in the DBCG database.
Conclusion:
Demonstrating high accuracy and reproducibility, the algorithm represents an optimal method for future identification of patients with pure CIS who are not registered in the DBCG database.
