Jackson, Meagan2024-08-292024-08-292024-08-292024-08-21https://hdl.handle.net/10012/20914Background: As Canada's population continues to age, it is essential to screen for nutrition risk in frail older adults to support older adults to benefit from services and care that can improve their nutrition and prevent outcomes of nutrition risk, such as frailty. Frailty is defined as "an age-related syndrome of physiological decline, characterized by marked vulnerability to adverse health outcomes" (Walston et al., 2022). Over time, insufficient protein and energy intake may increase the risk of malnutrition (Hoogendijk et al., 2019) and may lead to weight loss in an older population, which is an indicator of frailty (Cederholm et al., 2019; Fried et al., 2001: Wei et al., 2018). The Seniors in the Community Risk Evaluation for Eating and Nutrition (SCREEN)-14 is a 17-item questionnaire created to measure nutrition risk in older adults aged 50+ (Keller et al., 2005). As SCREEN was created before frailty became a relevant concept in clinical care, refinements to SCREEN-14 may be needed to specifically target frail older adults. Further, to promote use of nutrition screening tools in practice, they need to be as brief as possible. SCREEN-8 was developed using a mixed method of classical test theory and expert opinion to meet this need. However, it is not known if this is the best short version of the tool, specifically considering frail older adults. Objectives: The objectives of this thesis were to: (1) use expert opinion to determine the minimum set of SCREEN-14 questions needed to identify nutrition risk in potentially frail older adults (i.e., content validity), and if any new questions are required (Chapter 4) (2) use Item Response Theory (IRT) to also identify a minimum data set that can differentiate those most likely to be at risk, and specifically within the oldest age groups (proxy for frailty) (Chapter 5), and (3) evaluate the differences and overlap in these tools as compared to SCREEN-8. (Chapter 5). Two discrete studies will be completed to address these objectives. Methods and results: The first study (Chapter 4) uses a Delphi technique which gathers expert opinion and builds consensus on an issue by way of anonymous surveys. Forty-six experts in the field who have used SCREEN in the past and had contacted the originator were invited to participate in a two-round Delphi process. Eighteen experts participated throughout the two rounds. They were asked to rate (first round survey) and confirm (second round survey) the importance of current SCREEN-14 items for the inclusion in a minimum set that specifically could identify risk in those who are frail. Additional questions were asked on expansion of items (round one) and consensus on inclusion of new items (round two). The experts concluded that eight current SCREEN questions should be included in the final set with no new questions added. The second study (Chapter 5) addresses objectives two and three in a secondary data analysis of the Nutri-eSCREEN dataset. Nutri-eSCREEN contained SCREEN-14 in an online platform as well as self-reported sex, age, and geographic location. After cleaning and applying the exclusion criteria there were n=20,093 participants in the dataset used in the analyses. To address objective two, psychometric models, specifically a Graded Response Model (GRM) was conducted to identify a draft minimum SCREEN (i.e., what items discriminate risk) using the whole sample as well as only participants 75+years (n=2,749) to serve as a proxy for frailty. From both samples eight different questions (six of them overlapping) were identified using the item parameter estimates and consideration on coverage of the concept of nutrition risk. The IRT 75+ version was decided to be a better tool and used in subsequent analyses. Using demographic variables, a bivariate analysis was also completed to determine the statistical differences among SCREEN-14 questions by sex and age groups to demonstrate construct validity of the abbreviated IRT version. Variations of SCREEN (SCREEN-8 and SCREEN resulting from the Delphi) were created for comparison to this IRT version resulting from the GRM analysis to address objective three. SCREEN-8 identified the highest proportion at risk (42.1%) and the SCREEN-IRT the least (27.7%). The overlap of all three versions identifying participants to be at risk was 77.4%. Finally, model fit statistics among the three brief versions were calculated with SCREEN-8 and SCREEN-IRT found to be better tools than SCREEN-Frail. Conclusion: As the population is rapidly aging, easy to use and brief tools that target nutrition risk identification to those most in need of further evaluation are required. These studies helped to identify two separate potential versions of SCREEN through different methods. The different versions both contained eight questions as does SCREEN-8. The conclusion is that although these two new versions were created via consensus and IRT methods they do not appear to be superior to SCREEN-8 with respect to, brevity, model fit, and identification of prevalence of risk. Future work should further test and validate these versions for comparison to SCREEN-8.enolder adultsagingMEDICINE::Physiology and pharmacology::Physiology::Nutritionscreening toolsfrailtyItem Response TheorGraded Response ModelDelphiSCREENvalidity testingMaster Thesis