JMIR Publications

ABSTRACT

Background:

Palliative care aims to improve the quality of life of patients with life-limiting illnesses. However, global access to palliative care is severely limited; in 2020, only 14% of the 40 million patients requiring palliative care actually received it. In their daily practice, general practitioners (GPs) directly manage care pathways by identifying complex situations, organizing tailored care, treating pain or other refractory symptoms, adjusting their consultation schedules, and actively mobilizing territorial health resources. Despite this central role, proactive identification remains challenging due to clinical uncertainty, under-detected symptoms, and a lack of practical validation tools. Consequently, early screening is severely delayed, even though it is well established that timely identification significantly improves clinical outcomes. Leveraging Health Data Warehouses (HDWs) for the secondary use of routine health data has emerged as a promising strategy to bridge this gap and screen unmet palliative care needs.

Objective:

The objective of this study was to map the current use of data from HDWs for identifying unmet palliative care needs.

Methods:

Following PRISMA-ScR 2019 guidelines, this scoping review included quantitative studies in English or French language including adult populations (≥18 years) with life-limiting or incurable illnesses, a life expectancy <12 months, or palliative symptoms identified via secondary use of inpatient or outpatient HDW data. Purely technical studies without clinical results were excluded. A systematic search was performed on December 6, 2024, across PubMed, Web of Science Core Collection, Cochrane Library, and Lens.org. In line with scoping review standards, no formal risk-of-bias assessment was conducted, as the objective was to map evidence breadth rather than assess quality. Selection and data extraction were managed by two independent reviewers using the Rayyan® tool and a standardized Google Sheet® grid. Data synthesis comprised a descriptive analysis of study characteristics, technical methods, and performance metrics, presented in a PRISMA flow diagram and a narrative synthesis supported by summary tables and charts.

Results:

Among 2,735 identified studies, 49 met the inclusion criteria. Most studies (60%) were retrospective cohort studies conducted in North America (64%), in inpatient settings (80%), and in oncology (55%). Structured data were used in 82% of studies, primarily for mortality risk prediction (65%) or symptom identification (24%). Machine learning was the most common identification method (63.3%), with AUROC values typically ranging between 0.8 and 0.9. Patient-reported outcomes and unstructured data were less frequently used.

Conclusions:

HDWs are valuable tools for identifying unmet palliative care needs. However, there is a significant research gap regarding primary care populations, non-oncological chronic diseases, and the clinical integration of unstructured data. Future prospective studies in outpatient settings are needed to demonstrate clinical utility.