Initial Development of a Computer Algorithm to Identify Patients With Breast and Lung Cancer Having Poor Prognosis in...

Area of Science:

• Oncology
• Medical Informatics
• Palliative Care

Background:

• Physicians face challenges in prognostication and identifying cancer patients needing end-of-life care discussions.
• Effective identification is crucial for advance care planning, palliative care, and hospice counseling.
• Early intervention ensures patients receive timely and appropriate end-of-life care support.

Purpose of the Study:

• To describe the initial development of a computerized algorithm for identifying breast and lung cancer patients needing end-of-life care counseling.
• To create an electronic algorithm (e-EOL) to flag advanced cancer patients for discussions on care options.
• To improve physician's ability to offer timely palliative and hospice care consultations.

Main Methods:

• Extracted clinical and non-clinical data from electronic medical records of breast and lung cancer patients from 2010.
• Developed an electronic algorithm (e-EOL) using ICD-9 codes, metastatic disease presence, and albumin levels.
• Validated the algorithm against physician expert chart review for sensitivity, specificity, and predictive value.

Main Results:

• Identified 369 eligible breast and lung cancer patients; the e-EOL algorithm flagged 53 (14%) patients.
• Algorithm demonstrated 21% sensitivity, 96% specificity, and 91% positive predictive value compared to expert review.
• Survival analysis revealed significantly lower 6-month and 1-year survival rates for algorithm-positive cases.

Conclusions:

• The initial e-EOL algorithm shows promise in identifying advanced cancer patients.
• Further refinement by incorporating additional markers of advanced illness is planned.
• The goal is to enhance the algorithm's operating characteristics for real-time identification of patients with poor prognosis.