Prediction of premature all-cause mortality: prospective general population cohort study comparing machine-learning and standard epidemiological approaches

Talk Code: 
Stephen Weng
Luis Vaz, Nadeem Qureshi, Joe Kai
Author institutions: 
University of Nottingham


Prognostic modelling using standard methods is well-established, particularly for predicting risk of single diseases. Machine-learning may offer potential to explore outcomes of even greater complexity, such as premature death. To develop and report novel prognostic methods using machine-learning, in addition to standard survival modelling, to predict premature all-cause mortality.


This was a prospective cohort study to develop novel mortality prediction models in a large UK general population cohort. 502,628 participants aged 40-69 years were recruited to the UK Biobank cohort from 2006-2010 and followed-up until 2016. Participants were assessed on a range of demographic, biometrics, clinical and lifestyle factors at baseline. Algorithms for predicting premature all-cause mortality were developed and validated by randomly assigning 75% sample for training and 25% for testing. Two machine-learning algorithms were developed using deep learning and random forest using 10-fold grid-searches. Two standard models based on Cox regression were developed using stepwise modelling. A total of 60 potential predictor variables for inclusion. Calibration was assessed by comparing observed to predicted risks; and discrimination was assessed by area under the ‘receiver operating curve’ (AUC). The main outcome of mortality was provided by the Office of National Statistics.


A total of 14,418 deaths (2.9%) occurred over a total follow-up time of 3,508,454 person-years. A simple age and gender Cox model was the least predictive (AUC 0.689, 95% CI 0.681 – 0.699). A multivariate Cox regression model significantly improved discrimination by 6.2% (AUC 0.751, 95% CI 0.748 – 0.767). The application of machine-learning algorithms further improved discrimination by 3.2% using random forest (AUC 0.783, 95% CI 0.776 – 0.791) and 3.9% using deep learning (AUC 0.790, 95% CI 0.783 – 0.797). These ML algorithms improved discrimination by 9.4% and 10.1% respectively from a simple age and gender Cox regression model. Machine-learning algorithms were well-calibrated, while Cox regression models over-predicted risk.


Machine-learning significantly improved accuracy of prediction of premature all-cause mortality in this middle-aged general population, compared to standard methods. This study illustrates the value and exploitation of machine-learning for risk prediction within a traditional epidemiological study design, and how this approach might be reported to assist scientific verification.

Submitted by: 
Stephen Weng
Funding acknowledgement: 
The study was funded internally by the University of Nottingham, with the costs of data access provided by Road to Health Ltd.