APPENDIX A.

Associations between current contraceptive use and composition of family planning network, stratified by type of contraception (N=162 women)

	Currently using modern contraception*
	Injectables		Pill^		Implant
Independent variables	aRR	95% CI	aRR	95% CI	aRR	95% CI
Has a FP network	3.16	(1.39 – 7.17)	1.72e+07	(5.64e+06 – 5.22e+07)	5.56	(2.51 – 12.33)
Composition of FP network~
No FP network (reference group)	--		--		--
Social network only	2.89	(1.32 – 6.32)	1.94e+07	(6.90e+06 – 5.46e+07)	6.23	(2.78 – 13.94)
Provider network only	3.11	(1.22 – 7.97)	1.51e+07	(4.36e+06 – 5.23e+07)	6.34	(3.34 – 12.02)
Social and provider network	4.21	(1.94 – 9.12)	1.42e+07	(1.51e+06 – 1.34e+08)	3.53	(1.37 – 9.10)
Knows someone using that same method	2.28	(1.62 – 3.20)	17.48	(1.77 – 172.75)	6.30	(3.36 – 11.80)

^*Each estimate corresponds to the results of a single multivariable Poisson regression model. Contraceptive use is the dependent variable and the row variable is the primary explanatory variable of interest. All models adjust for respondent’s age, parity, marital status, primary school completion, and household earnings and sub-district fixed effects.

We excluded women who were currently pregnant and used cluster-correlated robust standard errors at the sub-district level.

^{^}The estimates for pill use are large because, among pill users, all women had a FP network and almost all women knew someone else using the pill.

^~Composition of FP network is a categorical variable. The estimates for each category within that variable are from one multivariable model.

Notes: aRR = adjusted risk ratio; FP = family planning