Discover the Science Behind Blood Test Analysis with IduScore

What it is

 

IduScore is a research prototype. It uses machine learning to find long-term wellbeing patterns in routine blood and lifestyle data. It shows how people with similar profiles tended to fare over time in a large research cohort. It does not provide medical advice or diagnosis.

 

​

Where the data comes from

 

We trained and evaluated the models with the United Kingdom Biobank (UKB), a large prospective study in England, Wales, and Scotland. About 500,000 participants were recruited in 2006–2010 and followed through linked health records. For this demo, we used ~8 years of follow-up for respiratory wellbeing and ~10 years for cognitive, cardiovascular and renal wellbeing.

Source: https://www.ukbiobank.ac.uk/

 

 

What information we use

 

Inputs are items available from routine care and simple history:

 

• Blood and urine measurements

• Lifestyle factors such as smoking, activity, sleep, alcohol

• Basic clinical factors such as age, sex, BMI, selected history

 

 

How IduScore was built

 

We train on historical data where baseline inputs are known and later wellbeing patterns are recorded. We split data into training, validation, and test sets to limit optimism and check performance stability. These are research quality controls, not clinical validation.

 

 

What “population-profile” means

 

For a person’s inputs, we find broadly similar profiles in the training data. We then show how those groups evolved, for example, the percentage in each group that showed poor wellbeing with the area over 8–10 years.

​

Important limitations

​

• UK Biobank participants are mostly of white European background. This affects generalisability.

• Follow-up time, coding, and missing data can shift estimates.

• The prototype is not clinically validated and is not a medical device.

• Personal medical decisions should be made with a clinician using standard care, and not based on IduScore.