Governing data for better health

It starts with a vaccine: in rural Tanzania, Saida receives her fifth dose of the DTaP vaccine. Her electronic immunization record is updated by a community health worker and a new data point is entered into the health system—adding to the billions of health data points collected every year.

Leveraging this single piece of data, Saida’s father will know what vaccines remain for her to be fully vaccinated. A health worker can report on the stock of DTaP vaccine in the community. The Minister of Health can celebrate the increased immunization coverage rates in the country. Each of these events makes use of that single data point; this is the power of data within health systems.

Data systems are a complex structure of inputs, users, and information flows. These systems contain immense amounts of information about the health of individuals, communities, and countries. Data governance—a broad term covering many different types of policies, processes, and protections—is the framework that defines how data flow through these systems. It can also help coordinate between systems, ensuring that data from multiple sources can be used together.

What makes a strong, thoughtful data governance structure? Here are four characteristics of strong data governance:

Data governance must prioritize the protection of individual-level data through data protection laws, ethical guidance on data collection and use, and technical regulations about data storage and processing. As data systems become more integrated and analytics tools become more sophisticated, data governance must account for how individuals may be tracked across data systems.

Even data that does not feel personal can still pose a risk if not properly protected. Data about the fifth dose of DTaP vaccine is in the same system that includes Saida’s name, details about her health, and her parents’ names and address. If “non-sensitive” data are improperly accessed, this could lead to data that are clearly “sensitive.” Artificial intelligence, machine learning or even simple cross-referencing can lead to unintentional data breaches if the proper precautions are not in place.

Data governance establishes a landscape in which information can be safely and securely shared between systems. Data architectures and policies for new data systems within a country can encourage best practices for data use and reinforce the data culture within a health system.

The Tanzania Data Use Partnership, an initiative led by the Government of Tanzania with support from PATH, is a prime example of how strong data governance policies are supporting the information needs of the health system and simultaneously protecting individual privacy.

Tanzania’s recently launched National Digital Health Strategy includes a new data governance framework that guides how data flows between the levels of the health system, so that national program managers can deploy resources where needed. The strategy also sets the standards for who can access and use health data. The Data Use Partnership incorporates security and privacy into all components of its work, from a national enterprise architecture to the personas used when ideating how to digitalize primary health care systems.

Technology reflects the biases of the people who create it. The data sets used to develop health care algorithms can be flawed or incomplete—creating an environment where marginalized communities may be mis- or under-represented. Data governance is one mechanism that can help ensure that marginalized populations are represented within health systems.

In the United States, an algorithm used by hospitals to allocate health care to patients was found to systematically discriminate against Black people. Black patients were less likely to be referred to the specialized care they needed. A different artificial intelligence–enabled diagnostic tool that is used to assess mammograms was trained using data from Europe, where breast cancer is detected in early stages in older women. This same tool may be less effective in sub-Saharan Africa where breast cancer is more often diagnosed in younger women.

When data laws, policies, and structures are grounded in equity, they advance the availability of diverse, representative data to inform health services. Digital technologies are increasingly using people-centered design to inform their products. These same principles can be used in the development of people-centered data systems and policies. Incorporating data from marginalized populations will require strict guidance on protection—but must be prioritized to ensure data represents the diversity of countries and communities.

Every sector balances concerns of privacy and the need for accessible data. Non-health data sources are being incorporated into health systems to better understand how things like the movement of populations, the trends in populations over time, and even weather patterns might impact health outcomes.

In the COVID-19 pandemic, countries began exploring Bluetooth-enabled contact tracing using smartphones. Trends in online searches identified by search engines can be used to detect increases in influenza cases based on what information individuals are searching for. These data sources—relevant for public health but outside the control of health systems—open opportunities to uncover novel approaches to health promotion and planning. Data governance systems, both within and without health systems, must be prepared to address this additional layer of complexity.

As digital technologies continue to expand the data available and how data are connected, data governance must also continue to evolve. Its role in providing individual protections, unlocking benefits for public health, and ensuring the representation of underserved populations will only become more important.