SCREEN III validation for community living older adults

Year:

2017

Applicant:

Keller, Heather

Institution:

University of Waterloo

Email:

hkeller@uwaterloo.ca

Project ID:

170602

Approved Project Status:

Complete

Project Summary

Older adults living in the community are nutritionally vulnerable. Nutritional risk predicts poor quality of life and health outcomes. Identification of at-risk individuals in primary health care is needed to ensure that treatment can be initiated early. Brief tools that can be self-administered are needed for busy doctors’ offices. An 8-item version of a valid and reliable nutrition screening tool was used in the Canadian Longitudinal Study of Aging. A 3-item version is believed to work just as well, but needs to be tested further. Both versions will be compared to other variables such as health care utilization and falls to determine their equivalence. Further, nutrition risk will be explored to determine those characteristics associated with risk and if nutrition risk is an important predictor of health outcomes. This analysis will provide crucial evidence to use the 3-item version of this nutrition screening tool in primary health care settings.

Project Findings

With this project a 3-item SCREEN was developed from the current 8-item version available in the CLSA. The 3-item version was contrasted with the 8-item version and sensitivity and specificity determined for various cut-points to develop a final score on the 3-item version, at which point participants were considered to have nutrition risk. Construct validity was assessed by comparing the 3- and 8-item versions to health variables as well as a composite outcome (any of low life satisfaction, poor self- perceived general health, hospitalization, ER use). Misclassified participants were determined based on the 3- and 8-item tool cut-points for identification of nutrition risk. We found that the 3-item tool identified 45% of sampled participants to be at risk vs. 32% for the 8-item version. Construct validity was comparable between the two versions; however 16.7% of participants were misclassified with the 3-item tool when compared to those identified by the 8-item version (more false positives than false negatives).