Machine Learning Can Accurately Predict Your Death Date
Would you want to know when you’re going to die? Okay, not my most upbeat of blog openings but it’s a question that always sparks debate. One that AI and Machine Learning could help us answer sooner than you might think, too.
A British based funeral planning service surveyed almost 3000 people in the UK asking if they’d want to know the date of their death. Out of all participants, 64% said they would want to know, with the top reasons being cited as wanting to try and avoid their deaths and wanting to make the most of their remaining time. Good enough reasons and to be honest, what I would also lean towards – although the prospect does naturally, fill you with unease.
This information can do more than fulfil a morbid curiosity however, and machine learning is now being used to predict early deaths. With an aim to help medical professionals in preventative healthcare, machine learning is able to predict this with better accuracy than doctors alone.
Researchers from the University of Nottingham have recently published research on how accurately machine learning models can predict death in comparison to current methods. They detail their success in this using deep learning models, random forest and cox regression methods.
Researchers gathered data from over 500,000 participants in the UK. Aged between 40 and 69, the subjects participated in the study between 2006 and 2010, with a follow up in 2016.
They processed a large sum of data to drive the outcomes. The algorithms analysed demographic data, biometrics and clinical data. Even life style and dietary information – such as their intake of fruit, vegetables and meats, were taken into account. This was compared to statistics on life expectancy and other information on diseases.
The accuracy of results improved from a simple age and gender cox model as the least predictive; they saw improvement using increased variables cox regression model then further enhanced with random forest and deep learning methods. The two latter identified correct outcomes with accuracy between 64-76%, whereas the first Cox model was 44% accurate. In a statement, Dr Stephen Weng said “We found machine learned algorithms were significantly more accurate in predicting death than the standard prediction models developed by a human expert.” This is a really exciting time for healthcare using high technologies like artificial intelligence and machine learning in preventative medicine.
This research can help improve preventative healthcare, tailor personalized medical care and enhance doctors’ approaches to risk management of disease progression. Predicting the risks of disease is nothing new to health tech. However, advances in the application of machine learning like this, we can hope to delve deeper into more complex health predictions for patients – prolonging life expectancy and treating disease earlier.
Other blog posts
Deep Learning, Guest Blogs
Python Frameworks for Your Upcoming Deep Learning and Data Science ProjectsRead More