Evaluating Long-Term Health Disparity Impacts of Clinical Algorithms Using a Patient-Level Simulation Framework.

Area of Science:

• Oncology
• Health Disparities Research
• Health Informatics

Background:

• Adjuvant chemotherapy decisions for colon cancer significantly impact patient outcomes.
• Racial disparities persist in cancer care and outcomes, necessitating careful algorithm design.
• The role of race in clinical decision algorithms is a critical area for health equity research.

Purpose of the Study:

• To evaluate the long-term effects of omitting race from a colon cancer adjuvant chemotherapy decision algorithm.
• To assess the impacts on health outcomes, costs, and racial disparities using a simulation framework.
• To account for measurement errors and racial bias in cancer recurrence ascertainment.

Main Methods:

• A patient-level state-transition model was developed using electronic health records.
• Simulations projected 30-year quality-adjusted life-years (QALYs) and costs for 4839 adults with stage II/III colon cancer.
• Three scenarios were compared: current practice, algorithm with race, and algorithm without race, with probabilistic sensitivity analysis (PSA).

Main Results:

• An algorithm including race improved average QALYs by 0.048 and reduced racial disparity by 0.20 QALYs compared to current practice.
• Omitting race led to 13% fewer Black patients receiving treatment, decreasing their QALYs by 0.07 and widening disparity by 0.13 QALY.
• Health disparity increased in 94% of PSA iterations when race was omitted from the algorithm.

Conclusions:

• Colon cancer decision algorithms can enhance population health and decrease disparities.
• Omitting race from algorithms may disadvantage certain groups and limit disparity reduction.
• Patient-level simulations are valuable for assessing algorithm impacts on health disparities prior to implementation.