Table 3.

Logistic regression estimates of the treatment on treated effect of SMS reminders on adherence to treatment advice

Primary outcome	Percentage (N)		SMS	SMS
	Received SMS (n = 174)	SMS not sent (n = 206)	SMS risk ratio (95% CI)	Risk difference (95% CI)
1 Followed treatment advice	81.6% (142)	65.5% (135)	2.33 (1.62–3.36)	16.1% (8.89%–23.2%)
2 Followed treatment advice*			1.84 (1.19–2.85)	11.3% (3.76%–18.9%)
3 Followed treatment advice**			2.34 (1.46–3.74)	15.5% (7.86%–23.2%)
4 Followed treatment advice***			1.68 (1.08–2.61)	9.68% (1.84%–17.5%)

Exclude 39 observations who were sent a text message but who reported not seeing the text message in the follow-up survey, N = 380.

*Controls for city survey was conducted (Ibadan/Ogbomosho) and treatment* city survey was conducted (Model 2).

**Controls for city survey was conducted (Ibadan/Ogbomosho), treatment* city survey was conducted, whether assignment to the treatment/control group was made off-protocol, and treatment* whether survey assignment to the treatment/control group was made off-protocol (Model 3).

***Controls for city survey was conducted (Ibadan/Ogbomosho), treatment* city survey was conducted, whether assignment to the treatment/control group was made off-protocol, and treatment* whether survey assignment to the treatment/control group was made off-protocol, an indicator variable for surveys conducted at PPMVs, an indicator variable for participant who had a bank account, and an indicator variable for participant could not read or write (Model 4).

Standard errors are clustered by retail site.