Evaluating bias in electronic health record data: using agent-based models to examine whether geographic disparities in community-acquired methicillin-resistant Staphylococcus aureus are due to differential health care–seeking behaviors - Scorecard - MDSpire

Evaluating bias in electronic health record data: using agent-based models to examine whether geographic disparities in community-acquired methicillin-resistant Staphylococcus aureus are due to differential health care–seeking behaviors

  • By

  • Brittany L Morgan Bustamante

  • Jose Pablo Gomez-Vazquez

  • Carlos Gonzalez Crespo

  • Larissa May

  • Laura Fejerman

  • Beatriz Martínez-López

  • January 6, 2025

  • 0 min

Share

Clinical Scorecard: Assessing Bias in Electronic Health Record Data: An Agent-Based Model Approach to Investigate Geographic Disparities in Community-Acquired Methicillin-Resistant Staphylococcus aureus Linked to Variations in Health Care-Seeking Behavior

At a Glance

CategoryDetail
ConditionCommunity-acquired methicillin-resistant Staphylococcus aureus (CA-MRSA) infection
Key MechanismsHealth care–seeking behavior influencing observed geographic disparities in EHR data
Target PopulationPatients with CA-MRSA infections across geographic regions in California
Care SettingEmergency departments (EDs) and primary care settings captured in electronic health records

Key Highlights

  • EHR data may reflect bias due to differential health care–seeking behavior, potentially inflating geographic disparities in CA-MRSA prevalence.
  • Agent-based modeling (ABM) simulating homogeneous disease prevalence and varying health care–seeking behaviors partially reproduced observed geographic disparities.
  • Geographic disparities in CA-MRSA prevalence likely arise from factors beyond health care access and seeking behaviors alone.

Guideline-Based Recommendations

Diagnosis

  • Consider potential bias in EHR-based CA-MRSA prevalence estimates due to differential health care–seeking behavior.

Management

  • Use comprehensive data sources beyond EHRs to assess true disease burden and geographic disparities.

Monitoring & Follow-up

  • Incorporate modeling approaches like ABM to understand the impact of health care–seeking behavior on observed data.

Risks

  • Risk of overestimating disease burden in low-income or rural areas due to differential access and utilization of health care services.

Patient & Prescribing Data

Individuals with community-acquired MRSA infections presenting to emergency departments in California

Approximately 50% of individuals with CA-MRSA may not seek medical care, leading to underrepresentation in EHR data.

Clinical Best Practices

  • Interpret EHR-derived infection prevalence with caution, accounting for potential biases from health care–seeking behavior.
  • Employ agent-based modeling to simulate and adjust for health care access and utilization patterns in epidemiologic studies.
  • Recognize that geographic disparities in CA-MRSA may reflect complex social and environmental determinants beyond health care access.

References

Original Source(s)

Related Content