A Microsimulation-Based Approach for Mitigating Societal Bias in Chronic Kidney Disease Data.

Area of Science:

• Health Informatics
• Biostatistics
• Public Health

Background:

• Societal biases in healthcare data, particularly race-based criteria, can perpetuate inequities.
• Reassessing race-based criteria in chronic kidney disease (CKD) necessitates understanding their impact on disease progression.
• Existing data-debiasing methods may not fully capture the nuances of these criteria.

Purpose of the Study:

• To develop a microsimulation model to attenuate societal bias in primary care CKD data.
• To evaluate the effect of removing race-based diagnostic and treatment criteria on CKD progression.
• To generate counterfactual outcome distributions in the absence of race adjustments.

Main Methods:

• Developed a continuous-time, discrete-event individual-level simulation model for kidney function decline (eGFR).
• Simulated eGFR trajectories, incorporating factors like hypertension, diabetes, and CKD stage.
• Applied Bayesian calibration to estimate eGFR decline rates and compared scenarios with and without race adjustment.

Main Results:

• Removing race adjustment led to earlier CKD diagnosis for Black individuals and later diagnosis for non-Black individuals.
• The timing differences ranged from 0.6 to 9.6 years, with greater disparities in earlier CKD stages.
• No significant differences in life expectancy were observed between the race-adjusted and unadjusted scenarios (within 2 months).

Conclusions:

• The microsimulation model provides an alternative to existing data-debiasing approaches for CKD.
• Simulated data can inform policy decisions and the development of future interventions.
• Explicitly modeling the data-generation process helps anticipate the impact of policy changes on clinical data.