Every morning, hospital managers across England manually assess patient dependency on each ward to decide how many nurses to schedule for the coming shift. This matters because getting nurse numbers wrong has direct consequences: too few nurses increases the risk of patient harm, while too many wastes scarce NHS resources. The current gold-standard tool, the Safer Nursing Care Tool, requires staff to collect separate data specifically for staffing decisions—a labour-intensive process that pulls nurses away from patient care. This project will develop algorithms that predict nurse staffing needs using data already collected for patient care and hospital administration—admission records, care plans, and treatment codes. The researchers will build and test these algorithms across five NHS Trusts, validating them against real patient outcomes and nurses’ own perceptions of whether wards were adequately staffed. If successful, the approach could replace manual staffing assessments with automated predictions, freeing nursing time for direct patient care. It could also enable real-time monitoring of staffing adequacy, helping managers deploy nurses more efficiently across wards. The algorithms will be designed with input from nurses, managers, and patients to ensure they are practical and trustworthy.
View original technical description
Background: Having enough nurses on hospital wards is vital for patient safety but planning for varying numbers and needs of patients is hard. Almost all acute NHS Trusts in England use the NICE-endorsed Safer Nursing Care Tool (SNCT) to guide staffing decisions. However, this approach is labour-intensive and necessitates the collection of data specifically to measure staffing requirements, not informed by data gathered for administration or care management. Aim: Develop a method to measure demand for nursing staff on hospital wards using routine data to help plan establishments (number of ward employees), monitor staffing adequacy in real-time, and inform safe and efficient deployment of staff. Design: A retrospective longitudinal observational study across wards providing acute adult somatic (i.e. not mental health) inpatient care in 5 general hospital Trusts, predicting nurse staffing requirements from routinely collected data and validating these predictions against patient and staffing adequacy outcomes. Algorithms will be developed according to user-centred design and by engaging with patients to understand experiences of hospital nurse staffing and implications for developing algorithms. Workstream (WS) 1 Objective: understand what does/does not work for nurses and managers when using staffing tools, and incorporate this into algorithm design. Method: User-centred design approach comprising i) a national survey of staffing matrons and Chief Nursing Information Officers to find out how staffing tools are used and patient data availability/quality, ii) workshops with nurses and nursing managers to understand staffing decision support needs at different timepoints, iii) workshops with this group plus NHS IT managers and roster companies to discuss algorithm design considerations. WS2 Objective: develop statistical/machine learning algorithms to estimate nurse staffing requirements from routinely available patient data. Method: Since there is no "gold standard" for measuring nurse staffing requirements, we will first replicate measurements from the SNCT, a patient acuity/dependency classification tool. We will develop alternative algorithms including replicating individual patient acuity/dependency classifications and replicating the staffing requirements for a whole ward. We will consider staffing decisions at different timepoints. Our predictor variables will come from administrative and care plan data. WS3 Objective: assess the validity of algorithms. Method: We will fit regression models to investigate the associations between actual under/over-staffing relative to each candidate measure of staffing requirements and multiple outcomes. For this, we will use routine data extracted from hospital IT systems and a micro-survey of nurses to understand perceptions of staffing adequacy. We will test whether as staffing increases relative to a measure of staffing requirements, the risk of poor patient outcomes and perceptions that staffing is inadequate decreases. We will compare model fit against models with staffing requirements measured by the SNCT. Timelines: 2.5 years Anticipated impact: A better match between staffing and workload on hospital wards, more efficient deployment of scarce resources and less time-consuming staffing assessments Dissemination: Open-access journal articles, magazine articles for nurses and videos/posters for the public. We will share results with intended users through workshops and user groups.
Plain English summaries and category classifications on this site are generated by AI and may not perfectly reflect the original research.
Is something wrong? Let us know