AI-driven health system in Kenya under fire for raising costs for the poor

A flagship healthcare reform in Kenya, pitched as a path to affordable, universal access, is now under growing scrutiny after an investigation found it may be placing a heavier financial burden on the very households it was meant to support.

Rolled out in October 2024, the new system replaced Kenya’s longstanding national insurance scheme and stood at the centre of President William Ruto’s campaign agenda. Built to extend coverage to the country’s vast informal workforce, day labourers, farmers and small traders who make up nearly 83 per cent of all workers, the programme introduced a contribution model calculated using artificial intelligence.

But early findings suggest the reform has had unintended consequences, with poorer households reportedly facing higher costs under the new structure.

“No Kenyan will be left behind,” Ruto had said during his 2023 presidential campaign in Kericho.

Concerns over pricing model

The system calculates contributions using a predictive machine learning algorithm rather than newer AI models. Through a means-testing process, it estimates how much households can afford to pay.

More from World

However, the formula has been described as “flawed” and lacking transparency, leading to protests and public anger.

Findings from investigation

An investigation by Africa Uncensored and Lighthouse Reports found that the system has consistently overestimated the incomes of poorer households while underestimating those of wealthier individuals, reported The Guardian.

The analysis showed that this has resulted in higher charges for the poorest and lower contributions for the better-off.

How the system works on the ground

Field workers such as Grace Amani visit households to collect data through questionnaires. Residents are asked about living conditions, including the type of toilet they use, their roofing materials and whether they own items such as radios.

The information is entered into a digital system, which calculates the household’s contribution for public health insurance. Many residents are left confused by the process, and some fear they are under investigation.

Amani said many of the people she registers in Nairobi are among the poorest, yet are being assigned contributions they cannot afford. Some households face charges amounting to between 10 per cent and 20 per cent of their incomes.

She also reported cases where critically ill individuals were unable to access treatment due to unpaid contributions.

“People are dying, people are suffering,” she said.

Since its launch, the Social Health Authority has faced criticism over misclassification and unclear pricing. People who fail to pay risk being denied treatment or facing high medical bills.

Online, many Kenyans have shared their experiences, including sharp increases in contributions. “From struggling to pay 500 Kenyan shillings [£2.90] previously to being billed 1,030 Kenyan shillings,” one user wrote.

Health economist David Khaoya said the system’s design involved a trade-off. According to him, it could either accurately assess poorer households or wealthier ones, and the decision was made to prioritise evaluating the wealthy.

“If you identify a richer person as poor and therefore ask him to pay less, this person will never own up and say, ‘I’m actually supposed to be paying more,’” he said.

The system is based on proxy means testing, a method used in several World Bank-backed programmes to estimate income using indicators such as living conditions and household assets. Experts note that this approach has been applied across multiple regions.

An audit comparing the system’s estimates with real household data found repeated cases of overestimation, including instances where incomes were calculated at twice the actual level based on factors such as home ownership and access to electricity.