Risk factor profiles
The risk factors that were found to be associated with falls among people living with dementia in the community (Figure 1, Table S3) were used to create risk factor profiles (Table 1). Each profile is a grouping of people who have similar risk factor combinations and can be used to understand if certain groups of people are at a higher risk of experiencing first hospitalised falls or severe outcomes following a fall hospitalisation.
| Risk factor profile | Profile description | Number of people in profile |
|---|---|---|
| 1 | Low proportion of risk factors other than medicated hypertension (high blood pressure) | 32,910 |
| 2 | Low-moderate proportion of risk factors | 10,497 |
| 3 | Polypharmacy (dispensed 5 or more different medications) and cardiovascular disease | 20,384 |
| 4 | Diabetes and polypharmacy | 8,555 |
| 5 | Multimorbidity (people with multiple health conditions) | 8,509 |
These risk factor profiles were created using cluster analysis, for further details see the following section, Identification of risk factor profiles.
Cluster analysis aims to uncover structure in data, such as underlying subpopulations, by assessing similarities across a set of characteristics. These similarities are measurable and can be used to group the data in such a way that individuals in a group are most similar to each other and least similar to those in other groups. In this study, the characteristics used for clustering are the risk factors that were found to significantly increase the likelihood of falls among people living with dementia (when the relationship of each risk factor with falls was explored individually – bivariate logistic regression, and again when controlling for presence of other risk factors – multivariable logistic regression). Each cluster (risk factor profile) is formed by grouping together people with similar risk factor combinations.
The optimum number of profiles is based on assessment of the stability of risk profile groupings over multiple iterations (that is, the groups and their characteristics do not change when the process is run multiple times) and their distinction from one another (that is, each profile is clearly different). For the present analysis, 5 risk profiles each for community and aged-care residents was found to be optimal in terms of stability and distinctiveness. For further details on the cluster analysis performed on this data, see the technical notes.
Figure 2 shows the 5 most common risk factors within each risk factor profile identified among people living with dementia in the community, the proportion of people within each profile with each risk factor, and the age and sex breakdown of each profile. While hypertension was common across profiles, there were distinguishing risk factors within each grouping which the profile names are based on. The full list of risk factors by profile can be found in Table S4, and the total number of people in each profile by fall or no fall, and age and sex breakdown, can be found in Table S5.
Figure 2: Characteristics of risk factor profiles for people living with dementia in the community: proportion of people by top 5 risk factors, sex and age
See the following extended description for details about the data displayed in this image.
The top 5 risk factors and the proportion of people with each of these risk factors are shown for each profile (as outlined in Table 1). Hypertension and/or antihypertensive medication appears in the top 5 of all profiles. The majority of people across all profiles are aged 30–84 years. There are more women than men in profile 1, about equal men and women in profiles 2 and 3, and more men than women in profiles 4 and 5.