Background: In oncology, Patterns of Care (PoC) provide a detailed overview of all cancer-treatment interventions and recapitulate the entire patient’s journey. While manual-chart PoC reviewing is time-consuming, the use of Electronic Health Records (EHRs) enables the creation of data-mining solutions to automatically reconstruct the whole cancer trajectory with different granularity.

Methods: We tested the ability of the i2b2 (Informatics for Integrating Biology and Bedside) solution to support the automatic reconstruction of the PoC for consecutive and unselected HER2+ and TNBC breast cancer patients through a retrospective EHRs analysis over a decade of observations.

Results: From 2008 to 2017, 561 HER2+ and 412 TNBC patients were retrospectively identified by i2b2 platform at the Papa Giovanni XXIII Hospital in Bergamo. Most patients, 74.3% in the HER2 group and 71.8% in the TNBC group, received a (neo) adjuvant chemotherapy, with anti-HER2 drugs whenever indicated. Among the HER2 cohort, the 5-year Time to Treatment Change (TTC) and Overall Survival (OS) were 69.4% and 77.4% respectively, with 25% of patients receiving up to 3 lines of treatment in metastatic setting. Among the TNBC cohort, the 5-year TTC and OS were 59.3% and 69.4% respectively, with only 2% of patients receiving active treatment as third line of therapy. The consistency of the automated PoC reconstruction had to be reviewed in 1/3 of cases to clean conflicting data.

Conclusion: The i2b2 solution has the potential to provide a retrospective, automated reconstruction of the different PoC, with limited manual-chart review refinement and might contribute to support investigations in the field of real-world data.