An AI model analysing routine clinical data and blood fat levels may help identify people with Parkinson’s at greater risk of future cognitive decline.
The study found that machine learning could predict cognitive impairment, including problems with memory and attention, using information from standard medical assessments and blood tests.
Researchers believe the approach could help clinicians identify patients who may need closer monitoring, although further validation is required before it can be used in clinical practice.
Scientists in China developed the model using data from 91 people with Parkinson’s disease who received care at Beijing Tiantan Hospital between 2018 and 2025.
None had cognitive impairment at their initial assessment, but approximately half developed cognitive problems during several years of follow-up.
The researchers used a machine learning framework called Tabular Prior-data Fitted Network (TabPFN), which identifies patterns in structured information such as medical records and laboratory results.
The model analysed routinely collected clinical information, including patients’ age, education history, initial cognitive assessment scores and blood triglyceride levels.
Triglycerides are a type of fat found in the blood that can be measured through standard cholesterol testing.
In the original patient dataset, the model correctly identified approximately three-quarters of those who subsequently developed cognitive impairment and a similar proportion of those who did not.
Researchers then tested it using an external dataset from the Parkinson’s Progression Markers Initiative.
The model correctly identified 75 per cent of patients with cognitive impairment and 68.9 per cent of those without.
The researchers also examined whether triglyceride levels were associated with cognitive decline.
Higher levels were linked to a greater likelihood of cognitive impairment among people with Parkinson’s.
Removing triglyceride measurements from the model somewhat reduced its predictive accuracy, particularly its ability to identify patients who would not experience cognitive deterioration.
However, the association does not establish that higher triglyceride levels directly contribute to cognitive decline.
Further research is needed to understand how fat metabolism, the process by which the body processes and uses fats, relates to changes in cognitive function.
Cognitive impairment is a common non-motor symptom of Parkinson’s disease, affecting attention, memory and other thinking abilities.
Many people do not experience cognitive difficulties when first diagnosed, although some develop problems as the disease progresses.
Not everyone with Parkinson’s develops cognitive impairment, and clinicians currently lack reliable ways to predict which patients are most likely to experience these changes.
The researchers suggest that using routinely available clinical information could make cognitive risk assessment more accessible without requiring specialised testing.
They wrote: “The model may therefore be useful for identifying patients who warrant closer cognitive monitoring.”
Identifying people at increased risk earlier could help clinicians and families prepare for changes in cognitive function before they begin to interfere with daily activities.
However, the researchers acknowledged several limitations, particularly the small number of participants used to develop the model.
They said larger prospective studies involving multiple clinical centres are needed to confirm its performance across different settings.
Further research must also establish whether its predictions are sufficiently reliable and useful to support clinical decisions.
Although combining routine clinical assessments with blood triglyceride measurements showed potential for predicting cognitive impairment, the model requires further validation before it can be used in clinical practice.
Water pollution sources linked to higher risk of rare neuro condition
