Effects of a clinic-based reproductive empowerment intervention on proximal outcomes of contraceptive use, self-efficacy, attitudes, and awareness and use of survivor services: a cluster-controlled trial in Nairobi, Kenya

Figures & data

Figure 1. ARCHES cluster-controlled trial consort flow chart (Nairobi, Kenya; 2019)

Notes: *Participants were contacted at the 6-month follow-up point irrespective of their participation in the 3-month follow-up, unless they had specifically declined further study participation. **Numbers at exit reflect those included in the analytical sample via the modified ITT approach

Table 1. Female family planning client characteristics at baseline by ARCHES modified intent-to-treat (mITT) treatment groups, unweighted (unbalanced) and IPTW (balanced) (Nairobi, Kenya; 2019)

Characteristic	Unweighted (n = 659)				mITT IPTW^b (n = 661.05)
		Treatment group, n (%)				Treatment group, n (%)
	Total Sample (n = 659)	ARCHES Intervention (n = 328)	Control (n = 331)	p^a	Total Sample (weighted n = 661.05)	ARCHES Intervention (n = 328)	Control (weighted n = 333.05)	p^a
Age, mean (std dev)	27.29 (7.27)	26.70 (6.63)	27.86 (7.81)	0.04	26.67 (6.69)	26.70 (6.63)	26.64 (6.77)	0.91
Married	429 (65.10)	230 (70.12)	199 (60.12)	0.01	457.67 (69.23)	230 (70.12)	227.67 (68.36)	0.62
Education level				<0.001				0.92
Primary or less	145 (22.00)	95 (28.96)	50 (15.11)		188.11 (28.46)	95 (28.96)	93.11 (27.96)
Secondary	238 (36.12)	160 (48.78)	78 (23.56)		327.60 (49.56)	160 (48.78)	167.60 (50.32)
Tertiary or higher	276 (41.88)	73 (22.26)	203 (61.33)		145.33 (21.99)	73 (22.26)	72.33 (21.72)
Parity				<0.001				0.70
Nulliparous	147 (22.31)	46 (14.02)	101 (30.51)		85.91 (13.00)	46 (14.02)	39.91 (11.98)
Uniparous	201 (30.50)	110 (33.54)	91 (27.49)		228.09 (34.50)	110 (33.54)	118.09 (35.46)
Multiparous	311 (47.19)	172 (52.44)	139 (41.99)		347.04 (52.50)	172 (52.44)	175.04 (52.56)
Food insecurity past 30 days	120 (18.21)	80 (24.39)	40 (12.08)	<0.001	175.22 (26.51)	80 (24.39)	95.22 (28.59)	0.22
Paid work past year	455 (69.04)	230 (70.12)	225 (67.98)	0.55	474.69 (71.81)	230 (70.12)	244.69 (73.47)	0.34
Language of survey administration				<0.001				0.59
English	192 (29.14)	66 (20.12)	126 (38.07)		138.73 (20.99)	66 (20.12)	72.73 (21.84)
Swahili	467 (70.86)	262 (79.88)	205 (61.93)		522.32 (79.01)	262 (79.88)	260.32 (78.16)
Current modern contraceptive use	490 (74.36)	263 (80.18)	227 (68.58)	<0.01	528.28 (79.92)	263 (80.18)	265.28 (79.65)	0.86

^a P-values are based on chi-square analyses for categorical variables and on Satterthwaite T-tests assuming unequal variances for continuous variables.

^b Adjusted for inverse probability of treatment weights (IPTW – weighted on treatment against baseline characteristics) using ATT (average effect among the treated) weighting where the control group is weighted to match the characteristics of the intervention group. Baseline characteristics used for weighting included age (continuous), marital status, education level, parity, food insecurity, language of survey administration, employment for paid work in the past year, and current modern contraceptive use.

Figure 2. IPTW descriptive outcomes [proportions (binary) and means (continuous)] graphed over time within mITT treatment groups (Nairobi, Kenya; 2019)

^aReporting using inverse probability of treatment weights (IPTW - weighted on treatment against baseline characteristics) using ATT (average effect among the treated. Baseline characteristics used for weighting include age (continuous), marital status, education level, parity, food insecurity, language of survey administration, employment for paid work in the past year, and current modern contraceptive use. ^bBinary outcome, reported in IPTW proportions ^cScale resized to improve readability ^dContinuous outcome, reported in IPTW means Symbols are utilized to denote where the beta (i.e. average change in slope) between intervention and control at three- and/or six-month follow-up is statistically significant (see Table 3): ^p<.1, * p<.05

Table 2. IPTW descriptive outcomes at each time point within mITT treatment groups (Nairobi, Kenya; 2019)

	Control (weighted n = 333.05)^a					ARCHES Intervention (n = 328)
	Baseline	3-month follow-up	p-value^b	6-month follow-up	p-value^c	Baseline	3-month follow-up	p-value^b	6-month follow-up	p-value^c
Current modern contraceptive use, n (%)	265.28 (79.65)	242.63 (84.13)	0.15	233.40 (86.39)	0.03	263 (80.18)	251 (89.01)	<.01	251 (87.76)	0.01
Covert use of contraceptives, n (%)	39.12 (11.83)	40.40 (14.09)	0.40	44.23 (16.37)	0.11	41 (12.81)	68 (24.20)	<.001	65 (22.73)	0.001
Contraceptive self-efficacy in the face of opposition, mean (std)	3.73 (1.93)	4.29 (1.94)	<.001	4.23 (1.88)	<.01	3.79 (2.03)	4.38 (1.85)	<.001	4.35 (1.88)	<.001
Awareness of local violence survivor services, n (%)	176.75 (53.07)	188.01 (65.93)	<.01	204.72 (75.77)	<.001	214 (65.24)	248 (87.94)	<.001	267 (93.36)	<.001
Use of local violence survivor services, n (%)^d	3.02 (1.31)	1.16 (0.56)	0.44	0	N/A	1 (0.41)	2 (0.95)	0.48	3 (1.41)	0.25
Acceptability of RC, mean (std)	2.64 (1.96)	1.98 (1.82)	<.001	1.94 (1.76)	<.001	2.87 (2.04)	2.00 (1.84)	<.001	1.74 (1.72)	<.001
Acceptability of IPV, mean (std)	1.29 (1.70)	0.81 (1.34)	<.01	0.73 (1.13)	<.01	1.27 (1.68)	0.72 (1.25)	<.001	0.66 (1.11)	<.001

^a Adjusted for inverse probability of treatment weights (IPTW – weighted on treatment against baseline characteristics) using ATT (average effect among the treated) weighting where the control group is weighted to match the characteristics of the intervention group. Baseline characteristics used for weighting include age (continuous), marital status, education level, parity, food insecurity, language of survey administration, employment for paid work in the past year, and current modern contraceptive use.

^b P-values for the within treatment group difference from baseline to 3-month follow-up are based on chi-square analyses for categorical variables and on Satterthwaite T-tests assuming unequal variances for continuous variables.

^c P-values for the within treatment group difference from baseline to 6-month follow-up are based on chi-square analyses for categorical variables and on Satterthwaite T-tests assuming unequal variances for continuous variables.

^d P-values based on Fisher's exact test due to small n.

Table 3. Difference-in-differences average treatment effect among the treated (time × treatment) comparing treatment groups (ref = control) from multi-variable mixed-effect regression (Nairobi, Kenya; 2019)

Outcome	IPTW difference-in-differences (time × treatment) parameter ^a (95% confidence interval)
	Baseline to 3-month follow-up	p-value	Baseline to 6-month follow-up	p-value
mITT approach
Logistic regression [Beta – time × treatment (95% confidence interval)]^b
Current modern contraceptive use	0.43 (−0.19, 1.03)	0.17	0.13 (−1.05, 1.32)	0.83
Covert use of contraceptives	0.77 (−0.09, 1.63)	0.08	0.38 (−0.92, 1.69)	0.57
Awareness of local violence survivor services	0.84 (0.13, 1.55)	0.02	0.92 (0, 1.84)	0.05
Linear regression [Beta – time × treatment (95% confidence interval)]^b
Contraceptive self-efficacy in the face of opposition (Cronbach alpha = 0.75)	0.11 (−0.19, 0.42)	0.47	0.07 (−0.24, 0.37)	0.68
Acceptability of RC (Cronbach alpha = 0.70)	−0.12 (−0.42, 0.18)	0.44	−0.34 (−0.65, −0.04)	0.03
Acceptability of IPV (Cronbach alpha = 0.78)	−0.04 (−0.26, 0.18)	0.72	0.02 (−0.20, 0.24)	0.86
As-treated approach
Logistic regression [Beta – time × treatment (95% confidence interval)]^b
Current modern contraceptive use	0.28 (−0.45, 1.01)	0.45	0.12 (−0.94, 1.17)	0.83
Covert use of contraceptives	0.61 (−0.15, 1.37)	0.12	0.35 (−0.54, 1.24)	0.44
Awareness of local violence survivor services	0.78 (0.05, 1.51)	0.04	0.96 (0.28, 1.63)	<.01
Linear regression [Beta – time × treatment (95% confidence interval)]^b
Contraceptive self-efficacy in the face of opposition (Cronbach alpha = 0.75)	0.01 (−0.30, 0.31)	0.99	−0.04 (−0.35, 0.27)	0.80
Acceptability of RC (Cronbach alpha = 0.70)	−0.36 (−0.67, −0.05)	0.02	−0.24 (−0.54, 0.07)	0.13
Acceptability of IPV (Cronbach alpha = 0.78)	0.03 (−0.18, 0.25)	0.75	0.10 (−0.12, 0.32)	0.37

^a Adjusted for inverse probability of treatment weights (IPTW – weighted on treatment against baseline characteristics) using ATT (average effect among the treated) weighting where the control group is weighted to match the characteristics of the intervention group. Baseline characteristics used for weighting include age (continuous), marital status, education level, parity, food insecurity, language of survey administration, employment for paid work in the past year, and current modern contraceptive use.

^b DiD estimator reported is the beta for the interaction term of time × treatment from linear or logistic multi-variable mixed effect regression. Null value on these parameters is equal to zero with positive values (>0) indicating positive time trend and negative values (<0) indicating negative time trend. Odds ratios comparing intervention and control at each time point are shown in Supplement 1.

Data availability

Statistical code and the de-identified analysis dataset is available from the Dryad repository, DOI https://doi.org/10.6076/D1Q30G.