TABLE 6. Multinomial Logit Coefficient Estimates and Estimated Program (Interaction) Effect for LARC/PM Use and Other Method Use, Mayer Hashi Evaluation, 2013 (N = 37,902)

	LARCs/PMs			Other Methods
	Coefficient	SE	P Value	Coefficient	SE	P Value
Program (ref: nonprogram)	–0.084	0.088	.34	–0.243	0.049	<.001
Year 2013 (ref: 2010)	0.616	0.157	<.001	0.145	0.070	.04
Program X Year	–0.359	0.182	.05	–0.223	0.093	.02
Age, years (ref: 25–29)
<20	–2.396	0.280	<.001	–0.812	0.072	<.001
20–24	–1.039	0.166	<.001	–0.283	0.056	<.001
30–34	0.691	0.122	<.001	0.247	0.072	.001
35–39	1.071	0.127	<.001	0.460	0.071	<.001
40–44	0.487	0.155	.002	–0.054	0.078	.49
45–49	0.169	0.157	.28	–1.269	0.081	<.001
Education (ref: no education)
Primary incomplete	0.242	0.110	.03	0.189	0.057	.001
Primary complete	–0.164	0.127	.20	0.273	0.053	<.001
Secondary incomplete	–0.276	0.140	.05	0.231	0.061	<.001
Secondary complete or higher	–0.589	0.197	.003	0.279	0.085	.001
Wealth quintile (ref: lowest)
Second	–0.207	0.115	.07	0.064	0.056	.25
Middle	–0.285	0.124	.02	–0.139	0.061	.02
Fourth	–0.263	0.135	.05	–0.221	0.072	.002
Highest	–0.380	0.171	.03	–0.378	0.083	<.001
Religion (ref: Muslim)
Non-Muslim	0.545	0.137	<.001	0.278	0.063	<.001
Sector (ref: rural)
Urban	0.278	0.115	.02	0.286	0.054	<.001
Constant	–1.988	0.129	<.001	0.380	0.073	<.001
Program effect^a
Interaction effect	–0.017	0.011	.14	–0.038	0.021	.07

Abbreviations: LARC, long-acting reversible contraceptive; PM, permanent method; SE, standard error.
↵a Estimated from the predicted probabilities of LARC/PM use or other method use obtained from the model for each program area by survey year combination, in line with the difference-in-difference approach to estimate program impact, as described in the main body of the article.