There are some studies using derivations of the phenotype model or other categorical frailty models to measure the degree of frailty. Results from the comparisons between them and the original phenotype model have been published, with comparable model performances reported in different populations [
3,
34,
37‐
40]. In addition, evidence indicates that the phenotype model may require model calibration or redevelopment. Some studies have raised concerns about the scoring algorithm for the phenotype model, with emerging reports showing that not all the components contributed equally to the prediction of adverse health outcomes [
7,
41‐
43]. Previous findings have suggested that slow walking appears to be the most important risk factor for adverse outcomes among the five indicators included in the phenotype model [
41,
42]. Moreover, low predictive accuracy of the phenotype model in the prediction of adverse outcomes has been reported [
35,
37,
44]. For instance, one study found the phenotype model could not differentiate the healthy elderly from those with unplanned hospital admission and falls, with an area under the receiver operating characteristic curve (AUC) value of 0.50 and 0.52 respectively [
44]. Similarly, another study reported an AUC value of 0.55 for non-spine fractures and 0.63 for hip fractures in 6701 women using the phenotype model [
37].
Regarding the frailty index approach, much of the literature focuses on the comparison between the frailty index and the phenotype model in predicting risk of adverse outcomes [
33‐
35,
38,
45‐
47]. Of note, it may be methodologically challenging to directly compare the continuous frailty index and the categorical phenotype model. One study based on the GLOW 3-year Hamilton Cohort used three strategies to perform direct comparisons between the frailty index and the phenotype model by (1) investigating the associations between the adverse outcomes and respective per one-fifth (20%) increase of the frailty index and the phenotype model; (2) trichotomizing the frailty index according to the overlap in the density distribution of the frailty index by the robust, pre-frail and frail groups defined by the phenotype model; and (3) trichotomizing the participants based on a predicted probability function of outcomes predicted by the frailty index [
35]. All the three strategies yielded comparable predictive accuracy of the frailty index and the phenotype model in predicting risk of adverse outcomes. Additionally, some studies compare the frailty index with other existing tools for predictions of risk of osteoporotic fractures. For instance, there was one study comparing the frailty index with the fracture risk assessment tool (FRAX) in prediction of risk of major osteoporotic fracture (spine, hip, upper arm or shoulder, or wrist) and hip fracture in 3985 elderly women [
48]. The frailty index was found to be comparable with FRAX in predicting risk of major osteoporotic fractures and hip fracture, indicating that measuring grades of frailty may aid in fracture risk evaluation and fracture prevention for the elderly. Of note, we observed similar results in the women stratified by taking or not taking anti-osteoporotic treatments and/or supplementation, which indicated that the prediction of frailty index and FRAX in major osteoporotic fractures was not significantly influenced by the effect of anti-osteoporotic treatments and/or supplementation [
48]. However, further studies are needed to evaluate whether the assessment of frailty would be a useful addition to FRAX to improve predictive accuracy for risk of fractures in the elderly. Furthermore, despite abundant studies investigating the trajectory nature of the frailty index in the elderly, limited evidence is available for the change of frailty before and after an osteoporotic fracture. In our study, we aimed to assess the change of the frailty index before and after onset of a major osteoporotic fracture during follow-up in the elderly women [
46]. We found that the increase of the frailty index was significantly larger in the women who experienced a major osteoporotic fracture than their controls, indicating their greater deficit accumulation and accelerating frailty after a major osteoporotic fracture [
46]. Investigating the transition nature by the change of frailty index before and after a major osteoporotic fracture may be useful to serve as an indicator for the effect of treatments or interventions [
16]. For example, the change of frailty may be used to identify the minimally important differences (MIDs) in a fracture intervention study, taking into account the frailty transition nature [
46]. Though results of the prediction of frailty status in risk of osteoporotic fractures are consistent in the literature, it still remains largely unknown whether frailty is a cause or a consequence of osteoporosis. For instance, some studies have reported no significant cross-sectional relationship between frailty and osteoporosis [
49,
50], though frailty and osteoporosis share similar biological pathways and common risk factors such as advanced age, low physical activity, weight loss and cognitive decline [
51]. More high-quality evidence is required to further clarify the association between frailty and osteoporosis dependently or independently of the aging process.