Creating a Learning Health System in a Cancer Center: Generalizability of an Electronic Health Record Phenotype for...

Area of Science:

• Oncology
• Health Informatics
• Palliative Care Research

Background:

• Electronic Health Records (EHRs) are crucial for patient data management.
• Developing accurate methods to identify patients with advanced cancer is essential for targeted care.
• Previous EHR phenotypes for advanced cancer were validated in single institutions.

Purpose of the Study:

• To assess the generalizability of a previously developed EHR phenotype for advanced solid cancer patients.
• To validate the EHR phenotype in diverse healthcare settings, including the Veterans Health Administration (VA) and an academic cancer center.

Main Methods:

• A comparative study was conducted between January 1, 2016, and December 31, 2019.
• An EHR algorithm for identifying advanced solid cancer patients was compared against a human-coded reference standard.
• The study included a random sample of patients with active cancer from the VA and an academic cancer center.

Main Results:

• The EHR algorithm demonstrated high specificity (93% in VA, 97% in academic center) and reasonable sensitivity (85% in VA, 87% in academic center).
• Patients with advanced cancer exhibited significantly higher 6-month mortality rates (29.2% in VA, 17.0% in academic center) compared to non-advanced cancer patients (6.8% in VA, 3.5% in academic center).

Conclusions:

• The validated EHR phenotype is effective for identifying advanced cancer patients across different healthcare systems.
• This EHR phenotype can be utilized to measure and enhance the quality of palliative care for advanced cancer patients.
• The findings support the use of EHR data for improving cancer care quality in various healthcare settings.