Table 3 Linear regression models showing the independent variables that best explain the variation in HRQoL (EQ5D index) and burden (Negative impact scale COPE index)

MODEL A: Dependent variable

B

β

p

R²

Adj R²

Independent variables

Lower

Upper

 Relatives working/studying (0 = no;1 = yes)

.139

.074

.204

.275

<.001*

 Care interaction perceived by relatives (1 = never to 6 = always)

.027

−.004

.057

.117

.086

 Quality of support perceived by older people (4 = never to 16 = always)

.011

−.002

.023

.112

.087

 Burden perceived by relatives (7 = never to 28 = always)

−.007

−.016

.001

−.111

.093

 Explained variance

.112

.096

MODEL B: Dependent variable

B

β

p

R²

Adj R²

Independent variables

Lower

Upper

 Relatives’ health–related quality of life (−0.594 = worse than death to 1 = full health)

−1.732

−3.272

−.193

−.112

.028*

 Positive value in caring perceived by relatives (4 = never to 16 = always)

−.468

−.646

−.290

−.307

<.001*

 Quality of support perceived by relatives (4 = never to 16 = always)

−.211

−.352

−.070

−.171

.003*

 Changed behaviour of older person perceived by relatives (1 = never to 5 = very often)

1.002

.660

1.344

.308

<.001**

 Explained variance

.386

.376