When algorithms infer gender: revisiting computational phenotyping with electronic health records data

Area of Science:

• Biomedical Informatics
• Health Informatics
• Computational Epidemiology

Background:

• Electronic health records (EHRs) often lack comprehensive gender data, hindering research on specific populations.
• Computational phenotyping offers a method to infer gender using existing EHR data.
• This approach aims to enhance the inclusion of transgender and gender-expansive individuals in research.

Purpose of the Study:

• To conduct a narrative review of computational phenotyping for gender in EHRs.
• To critically examine the methodological and ethical challenges associated with this approach.
• To propose future research priorities and highlight existing recommendations.

Main Methods:

• Narrative literature review.
• Critical analysis of computational phenotyping methodologies for gender inference.
• Examination of ethical implications and potential misuse of algorithm outputs.

Main Results:

• Computational phenotyping can address data gaps in EHRs regarding gender.
• Significant methodological challenges exist in algorithm development and validation.
• Ethical concerns include potential for misgendering and misuse of inferred data.

Conclusions:

• Computational phenotyping is a valuable tool but requires careful implementation.
• Addressing ethical and methodological concerns is crucial for responsible use.
• Further research is needed to ensure equitable and accurate application in biomedical research.