Involving hard-to-reach ethnic minorities in low-budget health research: lessons from a health survey among Moluccans in the Netherlands

Table 4 Association between recruitment strategy and response rate: results from multivariate and multilevel logistic regression

	OR (95 % CI)
	Model 1	Model 2	Model 3	Model 4
No strategy (ref.)	1.00	1.00	1.00	1.00
Announcement letter	1.03 (0.47–2.25)	0.56 (0.26–1.17)	0.69 (0.36–1.29)	0.83 (0.42–1.62)
Information meeting	1.27 (0.69–2.35)	1.41 (0.86–2.31)	1.31 (0.86–1.98)	1.16 (0.74–1.80)
Local Media	1.08 (0.49–2.39)	1.89 (0.91–3.94)*	1.52 (0.81–2.83)	1.46 (0.73–2.93)
Social Media	0.71 (0.36–1.40)	0.52 (0.29–0.93)	0.64 (0.38–1.05)	0.58 (0.33–1.01)*
Door-to-door collection	1.69 (0.87–3.25)	1.64 (0.97–2.79)*	1.57 (1.01–2.43)	–
Involvement of community organizations	0.75 (0.37-1.54)	0.58 (0.31-1.06)*	0.53 (0.32–0.88)	0.60 (0.34-1.05)*
Key informant (cont.)^a	–	1.32 (1.11–1.58)	1.25 (1.07–1.45)	1.25 (1.06–1.48)
Active key informant^b	–	–	1.68 (1.18–2.38)	1.72 (1.16–2.53)

Model 1: single recruitment strategies
Model 2: model 1 + key informants
Model 3: model 2 + active key informants
Model 4: model 3—door-to-door collection
A p value ≤ 0.05 is presented in italics
* A p value ≤ 0.1
^aThe continuous variable of key informants represents the increase in OR corresponding to one additional key informant contributing to the recruitment process
^bKey informants were considered to be ‘active’ if they made major effort during the recruitment process (see text for the details)

ISSN: 1756-0500