Predictive AI may revolutionize waterborne disease control in Tanzania, but digital divide persist

A new study assessing the role of artificial intelligence in managing public health crises finds that machine learning (ML) holds strong potential to enhance prediction and response strategies for weather-sensitive waterborne diseases (WSWDs) in Tanzania. However, the research warns that infrastructure gaps, limited awareness, and data quality issues may delay real-world implementation if left unaddressed.

The study, titled “Assessing the Potential for Application of Machine Learning in Predicting Weather-Sensitive Waterborne Diseases in Selected Districts of Tanzania”, was published in Frontiers in Artificial Intelligence. It surveyed 76 Environmental Health Officers (EHOs) across three Tanzanian district councils, Morogoro, Dodoma, and Ilala, to evaluate their digital literacy, perceptions of AI/ML, and readiness to integrate predictive technologies into frontline disease control efforts.

How prepared are frontline health workers to use AI for disease prediction?

The study found that while most EHOs are familiar with general information and communication technologies (ICT), their understanding of artificial intelligence and machine learning remains limited. Only 6% of respondents described themselves as very familiar with AI/ML, while 64% were somewhat familiar, and 30% not familiar at all.

Despite these knowledge gaps, 70% of respondents expressed confidence that AI and ML could improve the accuracy and timeliness of disease outbreak prediction. Notably, 40% believed AI could replace traditional methods of predicting WSWDs, while 26% viewed it as a complementary tool.

Gender emerged as a statistically significant factor influencing exposure to and familiarity with AI. Male respondents were more likely to have prior exposure to AI/ML terminology and report higher levels of understanding. This aligns with broader research showing persistent gender gaps in digital literacy across low- and middle-income countries.

Age and education levels showed no significant correlation with familiarity or trust in AI/ML, although individuals with higher education were more likely to use digital tools in their daily public health tasks.

What challenges limit the adoption of machine learning in Tanzanian public health?

The study highlights several systemic barriers to the adoption of AI/ML technologies in Tanzania's public health system. Chief among them is a lack of technical infrastructure and skilled personnel, cited by 58.3% of respondents. Data quality and availability were also flagged as major constraints by 54.2%, alongside concerns about the reliability and compatibility of existing health information systems.

Only 14% of EHOs reported using meteorological data in disease surveillance, and a mere 6% had integrated remote sensing data, despite these being critical inputs for accurate predictive models. Instead, most respondents relied on manual reviews of historical disease records (66%) and sanitation data (60%), both of which are susceptible to reporting lags and inconsistencies.

Additional challenges included low trust in AI-driven decision-making, difficulties integrating AI systems into existing workflows, ethical concerns, and weak regulatory frameworks. About 33% of EHOs cited data privacy and ethics as major concerns, while 22.9% highlighted integration challenges.

Despite these hurdles, 41.3% of respondents supported the idea of using AI to build early-warning systems, while 30.4% endorsed real-time anomaly detection through data stream analysis. These figures suggest a readiness to explore AI tools, provided the right support systems are in place.

What needs to be done to translate AI potential into actionable public health tools?

The authors recommend a multi-pronged strategy to bridge the gap between AI potential and public health impact. First, they emphasize targeted training for EHOs to build digital and analytical literacy. The lack of exposure to AI among frontline workers makes it difficult to design and adopt locally relevant solutions. Education programs must prioritize capacity building across gender and geographic lines to ensure inclusive adoption.

Second, improving infrastructure is a prerequisite. Reliable internet access, electricity, and computational power are essential for deploying cloud-based or real-time ML tools. Without these, AI applications may remain limited to academic prototypes rather than operational decision-support systems.

Third, the study advocates for the integration of diverse datasets, meteorological, water quality, disease incidence, and satellite imagery, into centralized platforms. This would improve forecasting accuracy and support the development of AI models optimized for resource-limited settings.

Lastly, stronger governance mechanisms are necessary to ensure ethical AI deployment. These include regulatory frameworks, data privacy safeguards, and institutional accountability structures to manage risks associated with automation and algorithmic bias.