AI-Driven Health Reform in Kenya Accused of Deepening Poverty

A joint investigation by Africa Uncensored, Lighthouse Reports, and The Guardian has uncovered serious flaws in Kenya's AI-powered healthcare system. The system, promoted by President William Ruto as a key pillar of his election campaign, was rolled out in October 2024 to replace the decades-old national health insurance scheme.

According to promotional material, the new system would 'drive digital transformation' to expand access to healthcare for Kenya's vast informal economy – including seasonal workers, street vendors, and farmers, who make up 83% of the workforce. President Ruto once declared: 'No Kenyan will be left behind.'

However, implementation has sparked fierce backlash. At the core of the system is a predictive machine-learning algorithm that performs a 'proxy means test' (PMT), estimating household income based on criteria such as toilet type, roof material, and asset ownership to calculate insurance premiums.

The investigation found that the system systematically operates unfairly: it overestimates the income of the poor, forcing them to pay premiums amounting to 10-20% of their meager earnings; conversely, it underestimates the income of the wealthy, allowing them to pay less than they can afford.

Grace Amani, a social worker registering the poorest households in the capital Nairobi, recounted: 'Many families cannot afford the insurance premiums; they have to choose between having money for food or paying rent. I see seriously ill people not receiving treatment because they haven't paid. People are dying, suffering.'

Health economist David Khaoya, a former advisor to Kenya's Ministry of Health, said policymakers were aware of the algorithm's flaws but deliberately prioritized setting correct premiums for the wealthy, despite knowing this would overcharge the poor.

The PMT algorithm has long existed and was once strongly backed by the World Bank. It has been applied in many Asian and African countries. However, independent studies show very high error rates: one program in Indonesia mis-identified up to 82% of those in need of assistance; in Rwanda, the figure was 90%.

The Kenya investigation found that over half of poor households are overcharged beyond their ability to pay. Of the more than 20 million people enrolled in the SHA system, only 5 million regularly pay premiums. Many hospitals are heavily underfunded due to delayed reimbursements from SHA.

Dr. Brian Lishenga, Chair of the Kenya Rural and Urban Private Hospitals Association, criticized: 'This is a failed experiment. It is a poor tool for identifying poor households, but an excellent tool for the government to evade responsibility.'