Predicting neutropenia risk in patients with cancer using electronic data

Area of Science:

• Oncology
• Health Informatics
• Clinical Decision Support

Background:

• Current guidelines for myeloid growth factors like granulocyte-colony stimulating factor (G-CSF) rely on chemotherapy regimens, lacking individualized patient risk quantification.
• Electronic Health Record (EHR) data offers a potential source for patient-specific information to estimate neutropenia risk.
• An evidence-based algorithm for individualized neutropenia risk estimation has been developed.

Purpose of the Study:

• To evaluate the automated extraction of chemotherapy treatment data from EHRs.
• To externally validate an individualized neutropenia risk prediction model using EHR data.
• To assess the feasibility of using EHR data for personalized G-CSF prophylaxis decisions.

Main Methods:

• A retrospective cohort of adult cancer patients receiving their first cytotoxic chemotherapy cycle (2008-2013) was analyzed.
• Chemotherapy treatment data was electronically extracted from EHRs and validated via chart review.
• Neutropenia risk stratification was performed, and the model's calibration and discrimination were assessed.

Main Results:

• Electronic extraction of chemotherapy data from EHRs was successfully verified.
• The risk prediction tool categorized patients into low (57%), intermediate (20%), and high (23%) risk groups for febrile neutropenia.
• The model demonstrated good calibration (Hosmer-Lemeshow p=0.24) and adequate discrimination (c-statistic=0.75).

Conclusions:

• Automated extraction of chemotherapy treatment data from EHRs is feasible and accurate.
• The individualized neutropenia risk prediction model showed robust performance in an external validation cohort.
• This tool can support personalized risk assessment for neutropenia and guide G-CSF use in oncology practice.