Predicting length of stay on NHS mental health adult inpatient wards using machine learning Sussex Partnership NHS Foundation Trust

Bed occupancy in acute mental health NHS facilities is often close to or above 100%. This results in patients being admitted far from their homes and being separated from family and carers. It impacts patient experience, quality and safety, and it is challenging to admit patients who are in a crisis.

Length of stay (LoS) is an important driver of these occupancy levels, with patient stays over two months contributing disproportionately. However, most trusts do not routinely or accurately predict LoS.  

Predicting LoS is challenging, as multiple factors need to be taken into account, including primary diagnosis; physical health and/or substance misuse co-morbidity; housing, relationship and employment status; functional and social impairment; multiple ward changes; and the cultural milieu of a ward.

In partnership with the University of Sussex, this Advancing Applied Analytics project team will develop an algorithm using machine learning to predict LoS within three days of admission to an adult ward. The model will be piloted in Sussex and then tested at Oxford Health NHS Foundation Trust.

Routinely and accurately predicting LoS will provide patients with better quality of care, shaping their journey to make every day count and giving tangible hope at a time when they are acutely unwell. It will also free up clinicians to focus on shared decision making with the patient and carer, and support operational staff in bed management.

The algorithm will use data collected as part of the Mental Health Service Data Set, meaning it could be used by other NHS mental health trusts.

Contact

For further information about this project, please contact Rachael Duke, Head of Charity, Sussex Partnership NHS Foundation Trust.