Samenvatting

Background: The proportion of frail older adults is increasing and is expected to further increase in the coming years, both globally and in the Dutch population. This poses a great challenge to public health. To determine the prevalence of frailty in a population, a frailty index (FI) is recommended. A FI is an accumulation model encompassing health deficits in multiple domains. Previous research has shown that a FI can be created out of existing health surveys, since it is a flexible instrument, fairly insensitive to the use of specific items. However, this is based on scale development using Classical Test Theory, while few studies have investigated the psychometric properties of their FI using Item Response Theory (IRT). The aim of this study was to create a FI using the Dutch Health Monitor 2016, and to investigate its psychometric properties using Item Response Theory (IRT). Methods: Forty-two deficits were selected in three health domains, i.e., physical, psychological, and social. Psychometric properties were investigated by using an IRT model for polytomous response categories: the Graded Response Model (GRM). Items were evaluated by Cronbach’s Alpha, Factor Analysis, Point Polyserial Correlations, and GRM. Results: The analyses showed that all items demonstrated a positive association with the scale. However, five items did not fit well to the FI scale. From the physical domain these were body mass index and three items about adherence to physical activity guidelines: moderate activity per week; bone and muscle strengthening activities; balance exercises. From the psychological domain this was an item about a sense of control over one’s own future. Conclusions: By using IRT, we showed that while 37 items were adequate and fitted the scale well, five items in our FI were redundant, indicating that it does matter which items are selected for a FI. IRT is a strong method for item selection and thus for creating a more concise Frailty Index. Key messages: Creating a solid and more concise Frailty Index with IRT is promising for epidemiological research and public health. For creating a Frailty Index, item selection needs careful consideration. [ABSTRACT FROM AUTHOR]