Skip to main content
. 2016 Feb 13;57(4):745–756. doi: 10.1093/geront/gnv688

Table 2.

Bivariate and Multivariate Multinomial Regression Predicting Emotional Abuse Severity

Characteristic Bivariate models Final multivariate model
Score 1/2 Score 6/7 Score 12+ Score 1/2 Score 6/7 Score 12+
OR (95% CI) OR (95% CI) OR (95% CI) OR (95% CI) OR (95% CI) OR (95% CI)
Victim physical vulnerability
 Functional capacity 0.98 (0.81–1.18) 0.99 (0.83–1.19) 1.03 (0.84–1.26) 0.90 (0.71–1.14)a 0.85 (0.68–1.06)a 0.89 (0.70–1.14)a
 Poor health 0.75 (0.41–1.38) 1.09 (0.64–1.87) .99 (0.55–1.78)
 Age 0.93 (0.90–0.96)*** 0.90 (0.87–0.93)*** 0.91 (0.88–0.94)*** 0.92 (0.89–0.96)*** 0.89 (0.86–0.92)*** 0.91 (0.87–0.94)***
Victim–perpetrator relationship dynamics
 Victim dependent on perpetrator 1.57 (0.59–4.17) 1.18 (0.46–3.03) 1.47 (0.55–3.89)
 Spouse/partner 1.0 (0.54–1.87) 1.71 (0.98–2.97)† 2.17 (1.21–3.89)* 0.46 (0.18–1.19) 0.72 (0.28–1.86) 0.89 (0.34–2.29)
 Adult child 0.82 (0.43–1.58) 0.96 (0.54–1.73) 0.76 (0.40–1.45)
 Grandchild b b b
 Paid attendant b b b
Home cohabitation
 Lives alone with perpetrator 2.20 (1.02–4.76)* 3.41 (1.66–6.98)** 5.03 (2.36–10.70)*** 2.74 (0.99–7.54)† 2.97 (1.03–8.54)* 5.32 (1.87–15.09)**
 Lives with perpetrator and others 2.57 (0.90–7.32)† 4.43 (1.68–11.68)** 4.16 (1.47–11.76)** 2.05 (0.53–7.91) 2.51 (0.55–11.49) 2.90 (0.68–12.29)
 Lives only with non-perpetrator others 0.84 (0.43–1.64) 0.94 (0.51–1.76) 0.91 (0.44–1.85) 0.65 (0.29–1.47) 0.63 (0.29–1.34) 0.75 (0.33–1.70)
Sociocultural
 Female 1.08 (0.63–1.84) 0.87 (0.54–1.42) 1.18 (0.69–2.01)
 African-American 0.83 (0.42–1.62) 0.80 (0.43–1.49) 1.09 (0.57–2.07) 0.71 (0.33–1.53) 0.69 (0.33–1.45) 1.07 (0.50–2.30)
 Hispanic 2.22 (0.56–8.89) 3.40 (0.95–12.14)† 1.67 (0.39–7.21) 1.99 (0.46–8.61) 3.92 (1.01–15.19)* 1.75 (0.37–8.28)
Race/ethnicity other 0.95 (0.13–6.93) 1.36 (0.24–7.61) 3.5 (0.71–17.36) 0.94 (0.12–7.39) 1.54 (0.25–9.63) 5.03 (0.91–27.92)†
 Less than high-school 0.73 (0.30–1.79) 1.14 (0.51–2.54) 0.67 (0.27–1.68) 0.84 (0.30–2.32) 1.83 (0.70–4.76) 0.80 (0.28–2.32)
 High school 0.39 (0.20 – 0.73)** 0.80 (0.47–1.37) 0.49 (0.27–0.91)* 0.37 (0.18–0.75)** 0.96 (0.51–1.80) 0.45 (0.22–0.90)*
 Household income 1.06 (0.93–1.20) 1.15 (1.02–1.29)* 1.08 (0.95–1.22) 1.0 (0.85–1.17) 1.12 (0.96–1.28) 0.95 (0.81–1.11)
 Suburban 1.25 (0.61–2.59) 0.83 (0.41–1.68) 0.64 (0.29–1.40)
 Rural 0.91 (0.50–1.66) 0.83 (0.48–1.44) 0.66 (0.36–1.21)
Control
 Interviewed by proxy 0.95 (0.32–2.81) 0.85 (0.31–2.35) 0.51 (0.15–1.80) 0.98 (0.24–3.97) 0.93 (0.24–3.61) 0.50 (0.10–2.37)

Note: Multinomial referent category was score = 0. Independent variable referent groups: health status (good), cohabitation status (lives alone), race/ethnicity (Caucasian), education (more than high-school), and geographical context (urban). Adjusted multinomial regression model satisfied the Likelihood Ratio model fit test (p < .001) [χ2 (39, 509) = 129.76]. Independent variables in the final model had tolerance of 0.74 or above and VIF of 1.34 or below, which suggests no multicollinearity. OR = odds ratio. CI = confidence interval.

aFunctional capacity was included in the final model given its central role to age-associated vulnerability.

bNonconverge due to low number of positive cases on independent variable.

*p ≤ .05, **p < .01, ***p < .001, †p < .10 (borderline).