Transitions to residential aged care after hospital for people living with dementia

Linear regression models

Linear regression models, fit with a gamma distribution, were used to examine predictors of length of first overnight hospitalisation ending in 2017 (first hospitalisation) for people with and without dementia, focusing on the role of potentially preventable complications of interest to the dementia cohort.

Regressions were performed separately depending on people’s use of residential aged care before and after their first hospitalisation:

• continued to live in the community (community-dwellers),
• moved from the community to live in residential aged care (transitioned to residential aged care),
• continued to live in residential aged care (aged care residents).

The outcome of our regression model was a continuous variable: length of hospital stay (days)

The predictors of the regression model were all categorical variables:

• dementia status (people living with or without dementia),
• age (65–74, 75–84, 85+)
• sex (male or female)
• delirium (reported during hospitalisation or not)
• in-hospital fall (reported during hospitalisation or not)
• pneumonia (reported during hospitalisation or not)
• pressure injury (reported during hospitalisation or not)
• urinary tract infection (UTI, reported during hospitalisation or not)

For people who transitioned to residential aged care, the model also included:

• Eligible and awaiting entry to residential aged care (reported during hospitalisation or not)
• Use of respite residential aged care after discharge (used within 7-days of discharge or not)

For aged care residents, the model also included:

• Whether the person moved to a different aged care facility after discharge (or returned to their previous facility)
• Note: a sensitivity analysis was conducted and removing aged care residents who were accessing respite care in a residential aged care facility prior to their hospitalisation (rather than as a permanent resident) had no effect on the results.

Interactions between dementia status and each predictor were also tested. Where significant these tests show that the impact of having this predictor on length of stay is different for people living with dementia compared with people without dementia.

Estimating additional bed-days in hospital attributable to each predictor

The estimated number of additional days spent in hospital (bed-days) for people with potentially preventable factors was calculated separately for people living with dementia and people without dementia. Using predictors included in the linear regression models, the estimated number of additional bed-days was calculated by:

1. Using the adjusted effect associated with predictors of interest to estimate the adjusted additional length of stay for those exposed to a risk factor comparing to those unexposed to the risk factor.
2. Multiplying the estimated additional length of stay by the number of people exposed to the risk factor of interest to estimate the total additional bed-days attributable to this risk factor.