🔗 Share this article

Artificial intelligence has the ability to estimate people's health problems well into the future down the line, say scientists.

The technology has been trained to identify sequences in individual health data to calculate their risk of more than 1,000 diseases.

Experts describe it as a weather forecast that predicts a significant likelihood of precipitation – however applied to individual medical outcomes.

The objective is to apply the algorithm to spot high-risk patients to avoid medical conditions and to assist medical facilities anticipate needs in specific regions, well in advance.

The Mechanism

This system – named Delphi-2M – employs comparable methods to well-known AI chatbots like ChatGPT.

Language models are educated to comprehend communication sequences so they can predict the sequence of words in a sentence.

The predictive system has been programmed to identify trends in deidentified health data so it can anticipate subsequent developments and at what point.

It doesn't predict precise timelines, including medical emergencies on October 1, but instead determines chances of 1,231 diseases.

"Comparable to climate forecasting, where we could have a high probability of showers, we can apply this for wellness management," commented the main investigator.

"The system enables not just for one disease but all diseases at the same time - this represents a breakthrough to accomplish such forecasting."

Development and Confirmation

The AI model was originally designed using confidential records - covering medical admissions, physician documentation and personal behaviors like nicotine consumption - gathered from over four hundred thousand individuals.

The model was then tested to see if its predictions stacked up using data from other participants, and then with 1.9 million people's health data from Scandinavian sources.

"The performance is strong, highly accurate in Denmark," stated the main scientist.

"When the system predicts a specific likelihood, it really does seem that it manifests to be the predicted rate."

The algorithm is most accurate with conditions such as adult-onset diabetes, myocardial infarctions and systemic inflammation that have a defined development pattern, as opposed to unpredictable occurrences such as bacterial illnesses.

Implementation Scenarios

Patients sometimes get cardiovascular drugs through probability estimation of their risk of cardiac events or cerebral incidents.

This technology is not ready for healthcare implementation, but the plan is to use it in a similar way, to detect at-risk cases while there is a chance to take action prior to health deterioration.

Possible uses encompass medicines or tailored wellness recommendations - for example those likely to develop certain hepatic conditions showing improvement with reducing beverage consumption exceeding typical guidelines.

The algorithm could also help inform disease-screening programmes and process comprehensive medical data in an area to forecast requirements - including the number of cardiovascular incidents a year there will be within particular areas in 2030, to support preparation efforts.

"This is the beginning of an innovative approach to grasp medical outcomes and health deterioration," commented a prominent researcher in artificial intelligence for cancer.

"Predictive systems including these approaches could one day help tailor treatments and anticipate healthcare needs at scale."

Future Developments

This technology needs refining and testing before it is applied in healthcare.

Furthermore present inherent constraints as it was built from information primarily obtained from people aged 40 to 70, instead of comprehensive demographics.

The model is now receiving enhancements to account for more medical data like radiographic studies, hereditary data and laboratory results.

"We must highlight that this is research – everything needs to be examined and properly controlled and carefully considered ahead of application," explained the main investigator.

He anticipates it will follow a similar path to the use of genomics in clinical settings where it took a decade to go from scientists being confident toward clinical application to employ it regularly.

A different researcher stated: "This research looks to be a major progression aimed at widespread, comprehensible, and – essentially – morally accountable form of predictive modelling in healthcare."