Developing and validating models for predicting nursing home admission using only RAI-HC instrument data.

Area of Science:

• Gerontology and Healthcare Analytics
• Predictive Modeling in Clinical Practice

Background:

• Existing nursing home admission (NHA) risk models often rely on complex, multi-source variable sets, yielding a single risk score.
• There is a need for more accessible and informative NHA risk prediction tools that can guide targeted interventions.

Purpose of the Study:

• To develop a novel NHA risk model utilizing a single data source for simplified application.
• To provide richer output information beyond a single risk value, enabling tailored interventions.
• To identify specific intervention clusters for high-risk individuals.

Main Methods:

• Developed a predictive model using variables exclusively from the Resident Assessment Instrument - Home Care (RAI-HC) system.
• Employed principal component analysis and K-means clustering to categorize high-risk clients for targeted interventions.

Main Results:

• The developed model demonstrated performance comparable to existing complex models in predicting NHA.
• Identified three distinct intervention clusters for high-risk clients: physical functionality deficits, cognitive functionality deficits, and depression/mood disorders.

Conclusions:

• The proposed NHA risk model, using only RAI-HC data, is accurate and practical for clinical integration.
• The identified intervention clusters facilitate the precise targeting of care for individuals at high risk of nursing home admission.
• This approach simplifies the implementation of NHA risk assessment by leveraging a single, integrated data system.