Participants characteristics.

Background
For the oldest-old residents around their 90s living in facilities, quality end-of-life care is crucial. While an association between reduced food intake and death is known, specific patterns of intake changes before death are not well-documented.
Aims
This study aims to classify food intake changes among residents in Japan’s special nursing homes during the 6 months before death, enabling precision care for each group using routinely recorded data.
Methods
Sixty-nine deceased older adults from five special nursing homes were studied over 3.5 years (January 2016 to June 2020). Criteria included: at least six months’ residency before death, ability to eat orally during the study period, and death within the facility. We created a time-series dataset for 69 participants, documenting their average weekly food intake (on a scale of 0-10). Subsequently, we used cluster analysis to identify clusters of change in the average weekly food intake from the 6 months before death.
Results
Eligible residents’ mean age was 89.7 ±  6.7 years, and 79.7% were female. Cluster analysis classified 4 clusters of decline in food intake changes during the last 6 months before death: immediate decrease (n = 14); decrease from 1 month before death (n = 24); decrease from 3 months before death (n = 7); and gradual decrease for 6 months before death (n = 24).
Conclusion
This study identified four groups of food intake prior to death. Recognizing food intake clusters in practical settings can help manage and provide appropriate end-of-life care in facilities with few medical providers but many care providers.