Improving Immunization Logistics and Uptake with AI

Public health agencies often manage the ordering, distribution, and monitoring of recommended childhood vaccine and some adult vaccines, such as influenza. The COVID-19 pandemic exposed some important weaknesses in these systems, including patient registration and eligibility assessments, recall reminders, and inventory management. Many immunization information systems have gaps in their ability to integrate with clinical record keeping systems, and, in some states, legal barriers restrict the completeness of recording and reporting. AI can help clean and standardize these records, filling in missing data and identifying children who have missed routine vaccines.

Several examples of using AI for immunization systems have been already documented outside of the U.S.. In DataKind’s collaboration with Gavi, an AI-supported system used natural language processing to scan paper-based health records and extract vaccine histories, enabling outreach workers to locate unvaccinated children. It also used machine learning to predict where vaccine stocks would run low and which areas were likely to have higher dropout rates. These insights help immunization programs pre-position supplies, organize mobile clinics, and target community education campaigns. They also help prevent both wastage and shortages of vaccines.

Strengthening HIV, STI, and TB Prevention and Treatment Programs with Artificial Intelligence

Public health departments operate some of the longest-running, most comprehensive programs to screen and treat TB, sexually transmitted infections (STI), and HIV. Despite tremendous advances in biomedicine—from rapid testing systems to improved prevention and treatment regimens for all three conditions—many health departments still rely on antiquated systems for other aspects of disease control. Registering, tracing, and tracking patients and exposed persons often requires the use of paper forms, faxed results, and manual data entry.

Digital chatbots, already tested in HIV and STI programs in South Africa, can improve outreach by answering common questions, booking appointments, and conducting eligibility screening. AI can also analyze surveillance data to identify geographic clusters of new cases or detect unusual patterns that might indicate a gap in prevention services.

For TB, computer vision algorithms can now interpret chest X-rays more accurately than non-specialist clinicians. While these tools do not replace clinicians or case managers, they could be used to triage patients faster, support decision-making, and flag anomalies in real time. That, in turn, allows skilled public health workers to focus on activities that require human interaction, empathy, and judgement, such as interviewing patients about highly sensitive topics, counseling them on treatment options and social services, and supporting them emotionally.

Strengthening Services to Pregnant Patients and Infants with Artificial Intelligence

Public health systems have long prioritized linking people to care as soon as they become pregnant and supporting them through pregnancy and the first weeks of a baby’s life. Delays in accessing care, poor follow-up, and low-quality care can lead to complications, disability, or death, as demonstrated by persistently high infant mortality rates and a growing epidemic of congenital syphilis.

Digital assistants powered by AI can help frontline health workers deliver timely information to pregnant people and new parents, including personalized messages about antenatal visits, warning signs, and infant feeding. When integrated with electronic health records, these systems can alert providers when patients miss appointments or when risk factors suggest the need for more frequent monitoring.

In Rwanda, an AI algorithm used data from routine prenatal visits to predict which pregnancies were most likely to result in adverse outcomes. When applied, this could help nurses and midwives prioritize specific patients for follow-up care, reducing their workload and improving outcomes. Similarly, AI-supported triage tools can help differentiate normal from high-risk deliveries, ensuring that women are routed to the most appropriate facility for their pregnancy.

Postnatal care may offer the most immediate opportunity to improve outcomes. Many deaths among mothers and newborns occur in the weeks immediately after birth, when contact with the health system becomes sporadic. AI-powered mobile apps or voice-based tools can help monitor symptoms, guide self-care, and prompt clinic visits when needed.

Expanding Reach and Reducing Inequity for Chronic Diseases with Artificial Intelligence

While screening for hypertension, diabetes, and cancer is primarily carried out by clinics and hospitals, public health agencies often organize mass screening events, promulgate guidelines, and work to connect underserved populations to care. AI can help improve the efficiency and equity of these efforts.

Algorithms can analyze electronic health records and insurance claims to identify people who are overdue for screening. AI tools can also support mobile screening units by rapidly interpreting test results and flagging high-risk individuals for follow-up. In India, researchers developed a smartphone-based retinal scanner paired with an AI system that screens for diabetic retinopathy, allowing trained technicians, not ophthalmologists, to provide service in rural areas.

For cancer screening, AI can support interpretation of mammograms, Pap smears, or stool-based tests. Chatbots, phone-based reminders, and predictive risk models can help ensure that people complete initial testing and follow through with diagnostic evaluation and treatment.

Implementing Artificial Intelligence into Public Health Practice Requires Funding, Leadership, and Oversight

One of the most critical unanswered questions is whether AI applications will improve services and disease control in populations who distrust health institutions or face barriers to care. AI evangelists argue that AI can help healthcare workers personalize their outreach messages, identify the highest priority persons or populations for intervention, and offload some routine tasks to increase the amount of direct human contact with persons in need of services.

The risks, however, remain substantial. AI in public health raises concerns about data privacy, surveillance, bias, and accountability. Systems trained on incomplete or biased data can reinforce inequities. Tools designed without community input may be rejected by those they aim to serve. And commercial AI systems may be unaffordable for public agencies unless open-source models or shared infrastructure are developed.

Realizing the benefits of AI, while limiting the harms, will require public health agencies to have a combination of funding to modernize data systems, leadership willing to take chances, and ongoing technical, ethical, legal review about patient rights and equitable application.