Original title: LLM4CKD: Large Language Models for Early Stage Chronic Kidney Disease Screening
Muhammad Ashad Kabir, Sirajam Munira · September 2026
Catching chronic kidney disease early is critical, but standard computer models require lots of labeled training data. This makes them hard to use in real-world clinics. Researchers wanted to see if large language models, or LLMs, could help instead. They built a system that takes clinical patient data, arranges it into structured text templates, and gives it directly to the model. The model reads these prompts to screen for the disease without needing any task-specific training. The team found that language models perform competitively using just a few examples. In these low-data situations, they often match or beat traditional models. However, language models become less stable when inputs grow more complex, while standard systems improve more consistently with large datasets. Still, language models offer a flexible backup option.
Imagine you're trying to recognize different types of fruit at a market, but you've never been trained on fruit names. Normally, you'd need someone to show you many examples first. But if someone gives you just two or three labeled examples and you look closely at their features, you might guess correctly for new fruits. In this paper, the large language models are given a few examples of kidney disease cases and use clinical features to make screening decisions without prior training.
Chronic kidney disease affects millions, and early detection saves lives. This work shows that AI models can help screen for it even with very little labeled data, which could expand screening in resource-limited settings.
The study does not establish that LLMs are universally superior; their performance depends on the specific model used and degrades with input complexity. It also does not provide long-term validation or real-world deployment results, focusing only on controlled evaluations.
Ask anything about this paper. Answers come from its abstract, and say so when the abstract doesn't cover your question.
LLM4CKD: Large Language Models for Early Stage Chronic Kidney Disease Screening — Muhammad Ashad Kabir, Sirajam Munira
Abstract licence: CC0
Explanation generated by Synthesis from the abstract — it is not the original paper. This is a preprint, so it hasn't necessarily been peer reviewed.