Artificial intelligence able to predict premature death ‘very accurately’

Researchers have made AI capable enough of predicting diseases prior to doctors, but a new research suggests that now AI might also be able to predict premature death in people and that too very accurately.

Experts from University of Nottingham recently trained an AI system to evaluate a decade of general health data submitted by over half a million people in the UK. The AI was then tasked with predicting if individuals were at risk of dying prematurely from chronic disease.

The predictions of early death made by the AI algorithm were ‘significantly more accurate’ than the predictions delivered by a model that did not use machine learning, said study’s lead author Stephen Weng, as per Live Science.

In order to evaluate the patients’ premature death, the team tested two types of AI: ‘deep learning’ AI in which layered information-processing networks help a computer to learn from examples, and a simpler ‘random forest’ AI that combines multiple, tree-like models to consider possible outcomes.

Artificial Intelligence outshines doctors at predicting heart disease deaths

Then, the team compared the AI models’ conclusions to results from a standard algorithm, known as the Cox model. Based on these three models, the team evaluated the database of genetic, physical and health data submitted by over 500,000 people between 2006 and 2016. During that time, around 14,500 of the participants died, primarily from heart disease, cancer, and respiratory diseases.

All the three models determined that factors like age, gender, smoking history and prior cancer diagnosis were top variables for assessing the likeliness of a person’s early death. But, the researchers found that the models diverged over other key factors.

Where the Cox model leaned greatly on ethnicity and physical activity, the machine-learning models did not. The random forest model focused more on body fat percentage, waist circumference, amount of fruit and vegetables people ate, and skin tone, whereas the deep-learning model emphasized on job-related hazards and air pollution, alcohol intake and the use of certain medicines.

After everything, the deep-learning algorithm gave the most accurate predictions, accurately identifying 76% of subjects who died during the study period. The random forest model, in comparison, predicted about 64% of premature deaths, whereas the Cox model identified only 44%.

The study co-author Joe Kai said that machine learning can be used to successfully predict mortality outcomes over time. Using AI this way ‘could help with scientific verification and future development of this exciting field’.