Community Data Health Initiative

Cross-sector, cross-agency data harmonization has demonstrated enormous potential for population-level health outcomes, but the practice is not common in U.S States and cities, limiting the ability of healthcare providers to make the most effective decisions to protect the health of their populations. The global pandemic (COVID-19), followed by multiple disease outbreaks in communities across the US, including Polio which was thought to have been eradicated, highlighted the urgency with which targeted, hyper-local strategies are needed to prioritize resource deployment. Robust data systems and the use of advanced technologies like artificial intelligence (AI) can help in this process. Currently:

1. Cities have not fully explored the use of integrated, non-traditional data sources to get a complete landscape overview and which helps to identify communities with the most urgent needs. These data sources could include consumer wearables and IoT devices, sentinel data from satellites, environmental and ecological data, wastewater surveillance, cybersecurity, and data from grass roots organizations)
   

2. Local governments have not harmonized data across their several agencies (education departments, health departments, energy and the environment etc.) to establish ground truth. Accepted convention by practitioners is that this process is not possible, yet, through this initiative, we are demonstrating that it can be done. 

3. While some cities are exploring the use of technologies (IoT devices), the haphazard approaches undertaken may provide little utility because of non-standardized approaches
   

4. Cities are not effectively building and leveraging relationships between the public, private, non-profit, and academic communities. Historical approaches have been public-private or public-academic, not the trifecta
   

5. Cities are not leveraging the power of technologies like Artificial Intelligence and Machine Learning in health, to parse multi-source, big data, to provide relevant insights for health decision and policy making

6. Cities are not investing sufficient resources to better understand the ethical implications for the use of novel and advanced technologies in public health

Pooling data across agencies and community programs helps paint a more nuanced and holistic picture of health determinants which helps to formulate targeted interventions that addresses the root causes of health disparities. 

Cities need to become more efficient with resource allocation, early detection and intervention, improved care coordination, and enhanced research and innovation, all of which can be enabled through the applications of artificial intelligence in secure, privacy-preserving ways. 

Our solution has two parts: 

1) The Community Data Health Initiative is currently working with cities to build internal capacity for data discovery and analytics, identify and address data and personnel siloes, and build community engagement infrastructure. Each pilot city first established working groups comprising city and county officials (mayors offices), chief data officers, public health and scientific experts, community organizations - including grassroots groups and activists, and researchers from academic institutions to a) Chart a path for data informed decision and policy making b) Understand and map risks c) Establish variables as indicators of outcomes of interest d) Outline data needs and their owners e) Establish interagency and organizational data commitments, data sharing mechanisms, and data safety, including ethical analyses f) Develop and establish evaluation methodologies and analytics, g) Create visual data systems. 

We are breaking down historically accepted conventions that it is impossible to bridge interagency data systems and establish hyperlocal (city level) data sharing mechanisms. Five cities are currently piloting our initiative: Washington DC, Baton Rouge, Denver, and Baltimore.

2) The use of AI: As we set up data infrastructure, we are working with cities to address the questions of what data are needed, who owns it, how can it be shared, legal and regulatory issues, standardization and formatting issues, ethical concerns, storage and security concerns, who will have access, and how AI will be used to improve efficiencies and health outcomes. 

Our cities will be using AI in the following ways:

Micro-Population Health Interventions:

We are better able to micro target subgroups within populations, who need urgent help. By analyzing comprehensive data sets, we are better able to optimize, down to street levels, who is most in need of resources. Traditionally, health interventions and programs relied almost exclusively on the combination of social factors and health records for targeting. AI makes it possible to analyze very large datasets and include other confounding factors in it's analyses including climate change effects data, traffic and pollution data, critical service gaps like housing, insurance, and transportation, and proximity to environmental hazards, to identify specific areas within communities with the most need. By omitting these other data points in historical analyses, health programs often missed the individuals who needed the most help. 

Predictive Analytics for the identification and prevention of disease: 

In collaboration with our partners, we are using AI to analyze historical data to identify patterns and to predict potential outbreaks within communities. Using the combination of cross-agency and shared community data sources provides a more comprehensive overview of ground conditions and the algorithms are able to predict outcomes with a higher degree of accuracy.

Over time and with more data, the solution will be able to help with:

• Earlier detection of risks to population health
• Discovering behavioral patterns that affects health outcomes and prescribing interventions
• Monitor community health indicators
• Generate more comprehensive health insights

CDHI is currently piloting in 4 cities in the US, Including: 

Baltimore, Maryland:

• Baltimore: Population - 576, 498, roughly 70% minority, life expectancy differs by up to 20 years between neighborhoods, and 23% living below the poverty level. 

Washington DC:

• Population - 712, 816, roughly 60% minority, 27 year difference in life expectancy between neighborhoods, about 30% of black population living below poverty

Baton Rouge, Louisiana:

• With a population of 222, 185, the city of Baton Rouge does not have a health department. 25% of the total population live below the poverty line, however, within minority communities, that rate increases to about 56%.

By sharing cross agency data and data from devices such as air pollution sensors, then visualizing the data by using mapping technologies, each city-level working group can get a clear overview of ground conditions, including which communities are most at risk. For example, using AI to analyze multi agency datasets and confounding factors such as areas with lowest rates of health insurance, PM 2.5 levels, life expectancy, and heart health, points to neighborhoods and specifically, which streets, are most in need and that are often overlooked in city budgets and resource prioritization.

The initiative in each pilot city is led by the active collaboration between small, local community organizations, city agencies, and governments, to identify health challenges, establish potential solutions, and work together to include all relevant stakeholders. This includes:

• Students and local academic institutions 

• Grass roots and community organizations 

• Government agencies and staff

• Domain experts across both the public and private sectors

The working groups convene regularly (weekly) to evaluate progress and iterate on processes to ensure compliance and progress towards established goals. 

ITGH has been leading the topic group on “AI for Outbreaks'' within the joint World Health Organization and International Telecommunications Union working group on “AI for Health,” over the past three years to benchmark the use of AI and ML in disease outbreak and to develop standards for the use of the technologies in that context. The institute has also been conducting research on the use of novel technologies to improve population health outcomes and has published academic papers on related topics including policy and ethics. Lessons from these activities helps to inform the design of this initiative.

The team is international and boasts a diverse background across race, gender, nationality, and experiences including working on solutions in global and undeveloped contexts. The ITGH team has broad experience with implementing digital health solutions, working with governments across all levels - from hyperlocal to federal, and both within the US and globally. Importantly, the team is composed of domain experts who are interested in both research and practice and has multiple active collaborations with leading academic institutions, research centers, local community groups, and multilateral organizations. 

There are several distinguishing factors that make this solution innovative: 

• A departure from traditional approaches. Health practitioners typically rely on "health data" which includes electronic health records, clinical data, hospital admissions, insurance claims, and medical histories, along with some social determinants data, in making decisions. Many more factors affect health outcomes and by including these data points in our analyses, we can get a much more comprehensive overview of what is actually taking place within communities. 
• Truly integrated data ecosystem: cross-agency data sharing has historically been a major problem for both policy-makers and health practitioners because there have either been limited or no mechanisms for doing so. By starting at the hyperlocal levels, we are facilitating a path to bridging the divide and with representatives from all relevant agencies and groups.
• Data Fusion and Analysis: The machine learning algorithms enables identifying complex patterns, correlations, and trends that otherwise would not be apparent through traditional methods of analyses
• Analytics for Proactive Interventions: The predictive capacity of AI enables the identification of health risks and trends before they escalate, giving health professionals sufficient time to devise containment strategies. 
• True cross-sector collaborations outside of emergencies: many health departments, agencies, and community groups work in silos. This initiative establishes new pathways for active collaborations and engagements across these groups, which could lead to more cohesive and comprehensive strategies and interventions targeting health disparities. 

The solution is comprehensive in scope, addressing the challenge of multi-stakeholder inclusion in the decision making process - from small, grass roots organizations all they way up to mayors offices, and the sharing of the data that they collect. 

Through the initiative, the combination of these stakeholders decide together on which health challenges to focus on within their communities and the relevant data needed to address them. Importantly, the support of the mayors offices helps to facilitate the mechanisms through which data sharing is done.

Incorporating technologies like AI and ML in multi-sourced city-level data and using visualizations from mapping technologies to generate unique insights is novel. We expect this model to be replicated by other cities and indeed, it is already generating strong interest.

Our solution contributes to the UN Sustainable Development Goal 3 in three different ways:

Universal Health Coverage: 

By using AI to identify health disparities and areas or groups with the most need, we can optimize resource allocation and develop targeted interventions, to move closer to achieving universal health coverage i.e. ensuring that all members of a given population have equal access to the healthcare that they need. 

Health Security and reducing inequality: 

The capability AI provides for early threat detection, which improves response, helps in the prevention of disease. Further, by analyzing larger, yet community-specific datasets, helps identify and address health inequities which ultimately reduces health disparities. 

Data-Driven Decision Making and Capacity Building: 

Designing informed policies and interventions helps to enhance the effectiveness of health systems. By building teams across sectors and agencies, we are helping to address efficiency issues and establish broader support in the delivery of health services, building capacity in the process. 

For a comprehensive system that enables AI to accurately analyze, predict, and enhance health outcomes, we are using the following features and elements:

Big Data and Sources: Combination of data across multiple agencies, departments, organizations, groups, and individuals, including behavior, environmental, and social determinants data.

Integration and Interoperability: facilitating aggregation and integration of the diverse datasets

Machine Learning Algorithms: Used for predictive analytics, classification and clustering, and Natural Language Processing for extracting insights from the unstructured data. 

Deep-Learning Models: Including convoluted neural networks and recurrent neural networks with image recognition and time-series analyses, while feature learning helps extract relevant features from the complex datasets

Continuous Learning and Improvement Mechanisms: Feedback loops for incorporating feedback mechanisms for continuous learning. 

We are looking into how differential privacy and federated learning can be incorporated into these systems for privacy and protection. 

Data governance and compliance, security, transparency and explainability, community engagement and participation, and bias mitigation are all key features of this bold undertaking.  

Teams across each pilot city are diverse and representative, adopt ethical guidelines and adhere to standards. Anonymization and deidentification are central to all data collection practices and robust practices are in place to do so. 

ITGH has published extensively on the ethics of AI and it's applications to health and is well positioned to guide implementations and develop best-practice guidelines with partners. These include:

• Using Data and Digital Tools in Public Health: Decision Making in a Health Crisis
  

• Ethics Principles for Artificial Intelligence-based Telemedicine for Public Health

• The AI Ethics Principles of Autonomy in Health Recommender Systems

• Do No Harm: Importance of Interoperability for Deploying Digital Solutions in a Humanitarian Context

Goals for the next year with four pilot cities include: 

1. Create models for cities and counties to address health issues in a cross-agency manner, including housing, code enforcement, infrastructure investment and public health.
2. Identify means of testing and sensing, that produce the most deleterious effects on residents.
3. Suggest intervention strategies that combine analytics, IoT information and community generated data.
4. Design for solutions for local officials, and in particular members of the African American Mayors Association (partners)
5. Develop visualized spatial narratives that bring urgent attention to the problem and show where interventions will make the most difference.
6. Develop communications strategies that incorporate community sentiments in identifying neighborhood concerns related to environment and health.

Through the achievement of these goals, we anticipate members of the communities most impacted through omission in resource distribution, to begin experiencing change.

Goals for the next five years include significant expansion of pilot cities, while continuing to iterate on existing processes and achieving growth in initial pilot locations

ITGH leadership is composed of six individuals and approximately 10 graduate students. Project partners include five from academic institutions, six from contractors, and a varied number of city leadership staff. The total composition of the team is dependent on the pilot locations and city staff involvement.

The initiative launched in April, 2023.

ITGH and all partners are committed to diversity which is reflected in the composition of the teams. It is important to highlight the active participation of the African American Mayors Association as a project partner on this initiative and HBCUs - black and brown communities are historically deprioritized in decision making processes and often shoulder the brunt of negative health outcomes, - community organizations and groups which are often composed of minority residents, and women in public health.

The initiative is being designed and implemented with a core diverse group of partners - the Institute for Technology and Global Health (ITGH), the Data-Smart City Solutions lab at the Harvard Kennedy School, the Environmental Defense Fund (EDF), ESRI, the African American Mayor's Association, and public health researchers from Johns Hopkins and Howard University. Pilot cities identify and bring to the table all relevant agencies, community organizations, individuals, and organizations whose inclusion are vital. 

Operationally, ITGH's role in the initiative is solution design, facilitation, analytics, and evaluation. Further, we structure research to assess the impact of the initiative and prepare cases for relevant stakeholders. The Environmental Defense Fund manages the inclusion and support from subject matter experts. The Data Smart City Solutions team helps to identify key stakeholders and establishes relationships. The African American Mayors Association establishes and manages relationships with mayors and the inclusion of HBCUs. ESRI helps with data, mapping, and visualization needs, technologists and researchers help with the development and structure of the technological tools. Implementations are managed by city leadership

Project managers and associates lead weekly meetings with pilot cities to facilitate the movement of the initiative forward. Convenings are held in the initial phase of the initiative to: 

• Conduct needs assessments and defining goals
• Stakeholder engagement
• Agree on data needs, collection methodologies, and integration
• Data governance and security
• Technology infrastructure needs
• Community engagement and education
• Implementation planning
• Oversight and governance
• Public advocacy
• Scaling and replicability
• Monitoring and evaluation
• Capacity building

Because buy-in is at all levels of the city, from advocacy groups to the mayors' offices, access to tools and resources needed to implement the solution is more easily available. 

Once the solution has been fully implemented and pilot cities take full ownership, our role will evolve to consultants to ensure continued progress and to help shape future iterations.

The initial phases of the initiative in each city is currently supported by grants and funding Robert Wood Johnson Foundation.

Because we want cities to take full ownership of the program after it has been launched and fully integrated into their ecosystem, further funding is built into their operational budgets. Project partners take on a different role beyond that point. 

We anticipate that funding to scale the initiative to other cities (and there is a pipeline of cities that have expressed interest), will continue to come from grants. However, we also anticipate long-term maintenance and consulting contracts with local-government to continue evolving the solution within their cities. 

ITGH's portion of operational costs primarily consists of travel and lodging to the pilot cities (approximately $27,000) and human capital in the form of fellowships ($35,000), for an approximate total of $62,000.

We expect this figure to remain the same over the next year. 

ITGH is seeking funding of $62, 000k for this effort, which will include support for :

- Travel for team members to pilot cities

- Continuous convening and engagements with communities

- human capital in the form of fellowships

The 1 year residency within an ecosystem of other practitioners and experts who are aligned on a shared mission to improve population level health outcomes is incredibly exciting. Novel and innovative solutions do not often emerge from silos and cross pollination of work can lead to powerful breakthroughs in solving complex challenges.

ITGH is a team of junior practitioners and researchers and the opportunity for funding to execute our mission and mentorship from experienced experts and executives would be invaluable. Because we are now actively engaged with New York City's health authority, having a residency in the city would be exciting and help with efficiency.