Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Added section about Special considerations for AI projects and deployment #7

Open
wants to merge 1 commit into
base: main
Choose a base branch
from
Open

Added section about Special considerations for AI projects and deployment #7

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
40 changes: 40 additions & 0 deletions README.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -401,3 +401,43 @@ Over time, as more features are added and bugs are fixed, your codebase can beco

### End-of-life
Finally, there might come a time when your project has served its purpose or has been replaced by better alternatives, and you decide to stop maintaining it. In such cases, it's good practice to clearly mark the project as no longer maintained, and if possible, direct users towards other alternatives. This helps set clear expectations for users and contributors.

## Special Considerations for AI Projects and Deployment in Public Health

Implementing AI in public health demands attention to specific challenges and best practices:

### 1. Data Privacy and Security
- **Compliance**: Adhere to data protection regulations (e.g., PHIPA, HIPAA) and public health standards.
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I may leave HIPAA out since it is for the US.

The public health standards IIRC do not outline data privacy or security explicitly, but instead discuss data-driven decision making within various standards.

We could add a reference however to adhere to any agency-specific policies on data management, privacy and security.

- **Data Handling**: Prioritize the use of de-identified or aggregated data to safeguard individual privacy.
- **Security Measures**: Implement robust encryption and access controls to protect sensitive health information.

### 2. Ethical AI Implementation
- **Bias Mitigation**: Assess and address potential biases that could affect vulnerable populations, ensuring equitable health outcomes.
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Potential reference:
Raza S. Connecting Fairness in Machine Learning with Public Health Equity. arXiv preprint arXiv:2304.04761. 2023 Apr 8.

"The connection between fair ML and public health equity is critical, as the use of unbiased ML models can aid in the reduction of existing health disparities and promote equal access to healthcare services. We can effectively analyse and address the underlying factors that contribute to health inequities in various population groups by ensuring that ML algorithms are fair and unbiased. Fair ML models can be used to identify at-risk populations, optimise resource allocation, and tailor public health interventions to specific community needs." (p5)

- **Transparency**: Develop interpretable models to foster trust among health practitioners and the public.
- **Alignment with Public Health Principles**: Ensure AI systems promote community well-being and adhere to ethical standards.

### 3. Model Lifecycle Management
- **Version Control**: Maintain systematic versioning of datasets, code, and models to track changes and updates.
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I don't disagree, but note that data versioning is not something discussed in this documentation (yet).

- **Validation**: Use diverse datasets representing target populations to validate models, ensuring accuracy and generalizability.
- **Continuous Improvement**: Plan for regular updates to accommodate evolving health data and public health priorities.

### 4. Infrastructure and Scalability
- **Resource Assessment**: Evaluate infrastructure needs to support large datasets and complex models.
- **Deployment Tools**: Utilize containerization (e.g., Docker) and orchestration frameworks (e.g., Kubernetes) to facilitate scalable deployment.
- **Adaptability**: Design systems capable of deployment across various regions and adaptable to different public health contexts.

### 5. Monitoring and Maintenance
- **Performance Monitoring**: Continuously monitor models for performance degradation, especially in dynamic health environments.
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Potential reference:
World Health Organization (WHO). Ethics and governance of artificial intelligence for health [Internet]. 2021 June 28; ISBN 9789240029200. Available from: https://iris.who.int/bitstream/handle/10665/341996/9789240029200-eng.pdf

"Even after an AI technology has been introduced into a health-care system, its impact should be evaluated continuously during its real-world use, as should the performance of an algorithm if it learns from data that are different from its training data." (p33)

- **Anomaly Detection**: Establish real-time alert systems to identify and address unexpected trends or anomalies.
- **Regular Audits**: Conduct periodic audits to ensure models remain aligned with public health objectives and ethical standards.

### 6. Stakeholder Engagement and Collaboration
- **Inclusive Development**: Engage public health experts, community representatives, and end-users throughout the development process.
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Potential reference:
Gurevich E, El Hassan B, El Morr C. Equity within AI systems: What can health leaders expect?. InHealthcare Management Forum 2023 Mar (Vol. 36, No. 2, pp. 119-124). Sage CA: Los Angeles, CA: SAGE Publications.

"The implementation of AI in projects must be a collaborative effort. It should include physicians, patients, and communities from diverse backgrounds of social, cultural, and economic contexts."

- **Communication**: Maintain transparency with stakeholders to explain AI decisions and build trust.
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Check out:

Information and Privacy Commissioner of Ontario (IPC). Artificial Intelligence in the public sector: Building trust now and for the future [Internet]. 2024 Feb 1; Available from: https://www.ipc.on.ca/artificial-intelligence-in-the-public-sector-building-trust-now-and-for-the-future/

- **Training and Education**: Provide training for public health practitioners to effectively utilize AI tools, enhancing their impact.
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Reference:
Kueper JK, Rosella LC, Booth RG, Davis BD, Nayani S, Smith MJ, Lizotte D. Inaugural Artificial Intelligence for Public Health Practice (AI4PHP) Retreat: Ontario, Canada.

"Despite excitement and perceived opportunities for AI to support public health, a key barrier is that many local public health units do not currently have the capacity or resources for dedicated AI-related work. The retreat ended with a discussion about future steps and how to foster public health leadership and presence in AI. "


### 7. Legal and Regulatory Compliance
- **Regulatory Adherence**: Stay informed about and comply with relevant laws and regulations governing AI use in public health.
- **Ethical Guidelines**: Follow established ethical guidelines, such as those from the World Health Organization, to ensure responsible AI deployment.
Comment on lines +439 to +441
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This can perhaps be removed and integrated into section 1 and 2