Back to top

Using machine learning methods and Annual Population Survey data to predict job loss

Posted by Ann Caluori | Fri, 14/03/2025 - 17:38

Dr Lara Shemtob, SOM Honorary Policy Adviser, introduces new SOM guidance: 

The relationship between work and health has never been more relevant to the UK population. The number of working age people who are economically inactive due to ill health is approaching three million. Meanwhile, the population is getting sicker and placing higher demands on public services. This polarisation is creating a tension between government revenues and expenditure that is unlikely to be sustainable long term. The new government has defined its first mission: to achieve sustained economic growth.

To drive productivity, the working age population must be in work. The Get Britain Working White Paper was published in November 2024, setting out the government’s plans to achieve an 80% employment rate. One aspect of this is achieving retention, including preventing people who are in work from falling out of work due to ill health. In our current healthcare landscape, with the absence of universal access to occupational health, government spending on resources must be targeted to those in need and most likely to benefit from support.

In this innovative research report, Michael Oldridge and fellow students have built and assessed four machine learning models for predicting job loss 12 months down the line of an initial sickness absence. This technology has the potential to target resource to demand when it comes to investment of taxpayer money in services at the interface of work and health. This work is incredibly salient in the current climate, where efficiency is essential in any public spending aimed at growth.

The machine learning models developed by Oldridge and fellow students are a starting point. Future pilot programmes could build on this work as a springboard for testing human vs technological accuracy in triaging those who stand to benefit from work and health support, following service user journeys through to outcomes.

Beyond the specific research findings, this report opens a new direction for research, policy and practice in occupational health. We must leverage technology to deliver more efficient care at the interface of work and health. Researchers and clinicians must upskill in technological competencies, work with technology experts and be open to implementing technological change in the work that we do, and evaluate its impact to iterate and improve on the services and care we deliver.

Download 'Using machine learning methods and Annual Population Survey data to predict job loss amongst workers in the early stages of sickness absence' here.