Developing machine learning-based models to help identify child abuse and neglect: key ethical challenges and...

Area of Science:

• Public Health
• Health Informatics
• Medical Ethics

Background:

• Child abuse and neglect are significant public health concerns in the U.S.
• Electronic health records (EHR) offer potential for developing machine learning (ML) models to identify child abuse and neglect.
• Utilizing EHR data for child abuse and neglect detection presents critical ethical considerations.

Purpose of the Study:

• To discuss key ethical issues in the development and evaluation of ML-based risk models for child abuse and neglect detection.
• To provide recommendations for addressing identified ethical challenges.
• To highlight areas for future policy and research.

Main Methods:

• Phenomenological approach to analyze ethical issues.
• Identification and discussion of seven core ethical challenges.
• Development of recommended solutions for each challenge.

Main Results:

• Identified ethical challenges include data bias, EHR system design, lack of evidence base, absence of a diagnostic gold standard, and difficulties in model evaluation, practical testing, and clinical presentation.
• Provided specific recommendations to mitigate these ethical concerns.
• Outlined directions for future policy and research.

Conclusions:

• Addressing ethical considerations is crucial for the responsible development and implementation of ML models in child abuse and neglect detection.
• Collaborative efforts in policy and research are needed to ensure the safe and effective use of these technologies.
• Recommendations aim to guide ethical practice and improve outcomes for child welfare.