Flaws in Kenya’s AI-driven health reforms push up costs for poorest households

Kenya’s AI-assisted healthcare financing system is under scrutiny after findings that it may be increasing costs for low-income households due to inaccuracies in its digital means-testing model. The system, introduced under Kenya’s Social Health Authority (SHA) reforms, uses algorithmic scoring and household data to estimate income levels and determine contributions to public health coverage as part of a broader transition to universal healthcare financing. According to a joint investigation by Africa Uncensored, Lighthouse Reports and The Guardian, the model used to classify households has in

Kenya’s AI-assisted healthcare financing system is under scrutiny after findings that it may be increasing costs for low-income households due to inaccuracies in its digital means-testing model.

The system, introduced under Kenya’s Social Health Authority (SHA) reforms, uses algorithmic scoring and household data to estimate income levels and determine contributions to public health coverage as part of a broader transition to universal healthcare financing.

According to a joint investigation by Africa Uncensored, Lighthouse Reports and The Guardian, the model used to classify households has in some cases overestimated the incomes of poorer families while underestimating those of wealthier households, leading to uneven contribution burdens.

The investigation found that some low-income households may be paying a disproportionate share of income toward health premiums, while others have experienced inconsistent access to services due to classification outcomes.

Household assessments are based on indicators such as housing quality, assets and access to basic services, collected through field surveys and digital tools as part of the reform design.

Critics have described the system as lacking transparency, arguing that the algorithm operates as a “black box” in determining eligibility and contribution levels.

Concerns about fairness and accuracy were reportedly raised prior to rollout by individuals involved in the system’s development, according to the investigation, though the specific contents of those assessments were not made public.

Kenya’s health authorities have defended the broader digitisation of healthcare financing as necessary to improve efficiency and expand coverage, while acknowledging ongoing adjustments to the system’s implementation.

The controversy has renewed debate over the use of artificial intelligence in public service delivery, particularly in contexts where informal employment and incomplete income data make automated classification more complex.

Analysts say the case reflects broader global concerns about algorithmic decision-making in welfare systems, where data limitations can risk reinforcing existing inequalities if not carefully calibrated.