Table 3: Logistic regression analysis: violent caregivers versus all others ( ).

Variable S.E.Sig. EXP( )95% C.I. for EXP( )

Patient vulnerability
 Number of dementia symptoms2.4701.050.0194.8173.50912.518
 Level of functional impairment1.231.581.0342.0491.0934.912
 Number of drugs taken.463.538.3901.589.5534.563
 Verbal aggression2.1291.450.142.119.0072.841
 Violence 3.1851.228.0104.1682.1768.399
Caregiver risk
 Hassle experienced by caregiving.173.234.4601.189.7511.880
 Level of impairment sub-scale−14.52354.843.9964.962.6559.877
 Cognitive status.953.800.2332.594.54112.435
 Level of social support.023.418.9561.023.4512.322
 Psychiatric symptoms.249.486.6081.283.4953.326
 Self esteem−.472.514.046.662.591.748
 Coping style−17.49133.673.9992.534.5445.436
Interaction of risk and vulnerability
 Pt dementia * cg depression1.728.408.0075.4833.2177.075
 Pt depression * cg hassle * Cg psychiatric symptoms .162.425.0526.1764.5119.706

(a) Step wise logistic regression model: variables were considered significant risk factors if they were found significant on step one and if they remained significant when tested against all other significant variables in the model onstep two. All variables found significant on step 1 remained significant at step 2.
(b) Model accounts for 48.3% of the variance. Nagelkerke square = .483.
(c) Homer-Lemeshow goodness of fit test ( = 9.628, df = 8, ).