Digital health and AI in public health: where they help, and where they don't

Digital health and artificial intelligence attract a great deal of enthusiasm, and some scepticism. Both reactions are understandable. The honest position is that these tools can help a great deal, but only when the foundations are in place.

Where digital tools help most

• Routine health information systems. Platforms such as DHIS2, used in many countries, give managers timely data from facilities and districts and make it possible to compare performance.
• Dashboards and reporting. Well-designed dashboards turn dense datasets into something decision-makers can read in minutes.
• Supply chain visibility. Knowing stock levels across facilities reduces stock-outs and waste.
• Surveillance. Faster reporting improves outbreak detection and response.
• Document-heavy work. AI tools can help summarise large collections of reports, classify information and draft first versions of documents, leaving people to check and decide.

Where they disappoint

Technology cannot fix a process that is broken. If data definitions are inconsistent, entry is delayed, or nobody reviews the results, a new system simply produces bad numbers faster. Tools also fail when they are built without the people who will use them, or when there is no plan for connectivity, training and maintenance.

Cautions specific to AI

• Quality in, quality out. AI trained or prompted on poor data gives confident but unreliable answers.
• Bias. Models can perform worse for groups that are under-represented in the data.
• Privacy and consent. Health data is sensitive. Collection, storage and sharing need clear rules and security.
• Accountability. A person, not a model, must be responsible for decisions that affect patients and public money.
• Dependence. Avoid locking a health system into a single vendor or closed format.

A sensible sequence

1. Define the use case and who will use it.
2. Fix the data process first: definitions, quality checks, feedback.
3. Pilot narrowly, with real users.
4. Evaluate honestly, including costs and unintended effects.
5. Scale what works, using open standards so systems can talk to each other.

Where we focus

At ACI, our interest in digital health and data systems is practical: assessing data quality, designing information systems that managers will use, and turning data into recommendations that people can act on. Technology serves the decision, not the other way round.