The impact of self-interviews on response patterns for sensitive topics: a randomized trial of electronic delivery methods for a sexual behaviour questionnaire in rural South Africa

Table 2 Item response rates for general sexual behaviours

	ITT analysis							AT analysis
	PI arms		SI arms		Z	p	φ	PI arms		SI arms		Z	p	φ
Total number of respondents	174		166					203		137
Ever had sexual intercourse
Yes	165	95%	159	94%	0.47	0.49	0.04	194	97%	130	95%	0.00	0.99	0.00
Declined to answer	0	0%	1	<1%	1.05	0.31	0.06	0	<1%	1	<1%	1.49	0.22	0.07
Age at first sex
Median (IQR)	18	[17–20]	18	[16–20]	1.87	0.17	0.07	18	[17–20]	18	[16–20]	2.26	0.13	0.08
Declined to answer	30	17%	25	15%	0.35	0.56	0.03	35	18%	20	15%	0.39	0.53	0.03
Lifetime number of sexual partners
Median (IQR)	2	[1–5]	2	[1–4]	0.07	0.79	0.01	2	[1–4]	3	[1–5]	0.44	0.51	0.04
Declined to answer	31	18%	28	17%	0.08	0.78	0.01	33	16%	26	19%	0.47	0.49	0.04
Number of partners in past 12 months
Median (IQR)	1	[0–1]	1	[1–1]	4.13	0.04	0.11	1	[0–1]	1	[1]	9.10	0.003	0.16
Declined to answer	16	9%	25	15%	3.40	0.07	0.10	18	9%	23	17%	6.17	0.01	0.13
Number of sex acts with MRP in past 4 weeks ^a
Median (IQR)	2	[1–3]	2	[1–4]	0.11	0.74	0.02	2	[1–4]	1.5	[1–4]	1.41	0.23	0.06
Declined to answer	18	10%	31	19%	9.88	<0.001	0.17	21	10%	28	20%	13.8	<0.001	0.20
Frequency of condom use in past 4 weeks ^a
Never	14	17%	20	30%				16	55%	18	69%
Sometimes	8	10%	5	7%				9	29%	4	20%
Most of the time	14	17%	8	12%				14	16%	8	16%
Always	43	52%	31	46%	3.80	0.28		53	50%	21	28%	6.28	0.10
Declined to answer	3	2%	3	2%	0.06	0.80	0.01	3	1%	3	2%	0.51	0.47	0.04
Any concurrent relationships today
Yes	5	3%	12	6%	3.39	0.07	0.10	8	5%	9	6%	1.19	0.28	0.06
Declined to answer	11	6%	11	7%	0.01	0.92	0.01	13	6%	9	7%	0.01	0.92	0.01
Any concurrent relationships in past 12 months
Yes	8	8%	12	11%	1.35	0.25	0.06	11	13%	9	11%	0.22	0.64	0.03
Declined to answer	1	<1%	2	1%	0.43	0.51	0.04	1	<1%	2	1%	0.87	0.35	0.05

ITT Intention-to-treat, AT As-treated, PI personal interview arms (PAPI, CAPI), SI self-interview arms (CASI, ACASI), IQR inter-quartile range, MRP most recent partner. Z test statistics compare PI to SI arms. In each pair the upper value is a \( {\chi}_1^2 \) comparison of the proportion of affirmative responses amongst valid responses for binary outcomes and a non-parametric Kruskall-Wallis test with k − 1 degrees of freedom for continuous and ordinal variables. The lower value is a comparison of item non-response rates using a \( {\chi}_1^2 \) test. φ is the effect size associated with the relationship between interview arm and the outcome of interest. ^a These questions had not been asked in recent annual surveillance questionnaires

ISSN: 1471-2288