. 2016 Feb 17;72(1):140–150. doi: 10.1093/geronb/gbw007

Table 3.

Decomposition of the Ethnicity Coefficient Using the Karlson, Holm, and Breen (KHB) Method

	Onset model (n = 960), %		Progression model (n = 1,199), %
	Arabs	FSU immigrants	Arabs			FSU immigrants
			Improving	Worsening	Death	Improving	Worsening	Death
Total % of the ethnicity coefficient explained by all mediators	25.4*	−1.8	32.6	15.4	50.8	−2.8	181.0*	−58.7^†
% of the ethnicity coefficient uniquely explained by each mediator
Education	18.0*	−13.6*	12.8	22.5*	116.1^†	−12.4	135.6*	−36.1^†
Household income (€)	1.9	1.4	−15.9	1.6	−16.9	−7.1	−3.6	−2.6
Household net worth (€)	8.6	11.5	0.2	−5.3	−27.4	0.8	36.1	−8.6
# of chronic diseases	−2.8	−3.8	25.4^†	0.8	43.3^†	5.8	−1.2	3.1
Smoking status
Never smoker^a
Current smoker	0.0	0.17	5.6	0.8	−14.8	9.2	−7.9	−7.8
Former smoker	−1.1	−0.8	6.7	−0.4	−18.5	3.2	1.0	−2.7
Physical activity
Inactive^a
Somewhat active	−0.9	2.1	−1.7	−8.6^†	−46.3	0.6	−18.7	5.1
BMI
Nonobese^a
Obese	1.4	−1.6	0.1	−0.5	−3.1	0.2	6.4	−2.0
Orientation to time score	0.8	3.3	0.2	4.6	19.2	−0.2	29.9	−6.3
Memory function score	−0.5	−0.3	−0.7	−0.1	−0.5	−2.9	3.5	−0.6

Notes. Positive values indicate attenuation of the association between ethnicity and functional limitations. Negative values indicate suppression of the association between ethnicity and functional limitations. BMI = body mass index; FSU = former Soviet Union.

^aReference group.

*p < .05. ^† p < .10 (two-tailed test).