LEAP PRIDE: An optimisation framework for resource allocation in palliative and end-of-life care

End-of-life care for frail and elderly patients is frequently characterised by high healthcare utilisation, fragmented service delivery, and limited coordination, resulting in variable quality and excess cost. The data supports this study which presents a proof-of-concept framework, tested using synthetic data to illustrate potential applications in strategic planning. Few planning approaches integrate patient-level pathways into operational models that balance efficiency with patient-centred outcomes.Optimisation models were developed to support strategic resource planning for frail, elderly, and palliative patients in the final year of life. Two formulations were explored: one minimising overall cost and another aligning demand with available capacity. Patients were stratified into ten representative categories and assigned to structured pathways with varying resource intensities across hospital beds, palliative beds, community nursing, and virtual wards.A synthetic dataset representing plausible twelve-month service trajectories was used to assess model performance. Both models produced feasible allocations that satisfied expected demand within capacity limits. Most patient groups were consistently assigned to dominant pathways, while some shifted depending on the optimisation objective, illustrating trade-offs between cost efficiency and balanced utilisation. Demand intensified in the final months of life but remained manageable under planning assumptions. The modelling framework demonstrates the feasibility of applying optimisation to anticipatory planning, enabling comparison of service configurations and supporting more coordinated, efficient, and patient-centred end-of-life care.Variable Clarification<b>Patient Type (1–10):</b><br>In this synthetic dataset, Patient Type is used to segment the population into subgroups in order to introduce heterogeneity. These groupings are stochastically generated as part of the optimisation model and are therefore not tied to predefined clinical categories within the dataset itself.In practical applications, this variable is intended to represent clinically meaningful cohorts used to distinguish patient groups, for example, underlying cause of death (e.g. cancer, cardiac, respiratory), frailty or age groups, or other relevant classifications.The specific mapping of Patient Type codes to real-world categories is therefore user-defined, depending on the context in which the model is applied.<b>Mode (1–2):</b><br>Mode represents alternative care pathways and is the primary driver of differences in healthcare utilisation within the model.<br><b>Mode 1:</b> More hospital-oriented pathway (higher inpatient utilisation)<b>Mode 2:</b> More community-oriented pathway (greater use of community, virtual ward, and palliative care services)<b>Interpretation:</b><br>The combination of Patient Type and Mode determines the resulting utilisation patterns. The dataset illustrates these differences and can be used to understand how different group-pathway combinations translate into resource use. The rest of the dataset has been generated around this structure to support the model in the manuscript.