Predicting Care Needs of Older Persons in the Healthcare System through AI-enabled Analysis of Patient-Monitoring Data (PAI)

Problem: Increased care demand of older people and limited healthcare staff resourcesEvery country is experiencing growth in both the size and proportion of older persons in the population. Older age is associated with increased health-related care needs and increased healthcare costs.

Much of the cost is related to acute episodes, which require immediate attention and thus costly emergency care, and increased level of aftercare. The care of older persons is managed at different levels of care. Healthcare staff resources are more and more limited in light of increasing demands due to the demographic shift.

Consequently, the European healthcare of the future needs to be able to reduce the total care needs of older persons, by either early adequate interventions at the current care level, and/or a timely transfer to the appropriate care level.Aim: Predict and enhance care needs for older people by early detection of signs of decline in health and functionThe PAI project predicts care needs of older persons by AI-based data driven analysis of current and historical patient-monitoring data.

A benefit of the early detection of decline in health and function is that preventive measures can be taken so that costly acute care with increased  aftercare can be reduced or avoided.

Avoiding hospitalisation can e.g be achieved by preventing fall incidents in nursing homes through assisting older persons in risky situations, or by avoiding heart attacks through timely adjustment of medication in primary care.

In addition, early insight in changing care needs means older patients can be transferred to another care level at the right moment to receive the care they need and enduro that health care staff resources are used effectively.

The latter contributes to the continuity of care and the coordination between inpatient and outpatient care.Implementation and use of care need predictions is challenging as data availability, data quality, type of possible deteriorations and type of possible interventions may differ strongly between care levels and different diagnosis.

Hence, a central aspect of the project is implementation of the solutions through redefining work processes in healthcare facilities, analyzing which different cultural or context-based differences may affect implementation results and developing guidelines on how implementation can be achieved in different countries and care levels.MethodsThe project follows a design science approach and a multiple case study design where context adapted analytics for predicting older patient care needs is developed in close collaboration with care recipients and health care practitioners, and implemented in six real-life settings (nursing homes, primary care units and hospitals in Spain, Sweden and Norway).

Data collection and data analysis follows a mixed-methods strategy that combines quantitative and qualitative elements, including the thematic interpretative analysis of meeting notes, interviews and observations, as well as the statistical longitudinal trend analysis of Key Performance Indicators in the respective healthcare systems.Impact on research community and societyPredictions of care needs benefits older persons (to enable healthcare providers to implement targeted interventions and preventive measures to prevent injuries and avoid worsening disease states), healthcare staff (effective use of their scarce resources will create a sustainable and more satisfying work environment) and the health system (efficient attention will ensure the system sustainability).

From an academic perspective, promising results have been shown in demonstrators, but there is a need to investigate how these methods can actually be applied for selected patient groups in real-life settings where data availability and conditions may not be optimal.

The outcomes of the project in terms of methods and tools are scalable and can be applicable and valuable in various other contexts.