Effects of sending a series of mobile SMS follow-ups to beneficiaries who have received referrals for family planning services from an IFPP family planning promoter. The activity seeks to understand whether behaviorally informed SMS messages can be a cost-effective means of increasing the referral redemption
rate among beneficiaries.

RCT- sending a series of mobile SMS follow-ups to beneficiaries who have received referrals for family planning services from an IFPP family planning promoter VS business as usual

The primary outcome of interest is the referral redemption rate. The raw data from DHIS2 contains individual records of issued referrals and which of these have been redeemed with nurses to be able to calculate this overall redemption rate.

We will estimate the intent-to-treat (ITT) effect between the control group (no SMS messages) and the intervention group (SMS messages). We will estimate an ordinary least squares (OLS) regression of the binary outcome (redeemed referral) on intervention assignment for beneficiary i in strata s. The strata corresponds to a supervisor-date cell: i.e., all women who had been provided with a referral by promoters working under a certain supervisor, who were visited on a certain date

The trial will conclude following the larger of: seven months of enrollment or the enrollment of 5,000 women.

Information from individuals who do not provide consent to be part of the study or are not issued a referral are still recorded within the CwS application per standard data collection practices conducted by PSI-Mozambique. However, data access and analysis will pertain only to the sample that provided consent. All enrolled participants who provided consent will be included in the intent-to-treat analysis; any participant for whom we do not have a consent form on file will be excluded from analysis. In addition, there may be some limited cases of duplicate phone numbers in which the same number is registered twice; anecdotal evidence suggests that this primarily corresponds to cases in which two women share the same phone, and both are visited by promoters. These women will be included in the sample, but each pair jointly has a probability of assignment to treatment of 75% (and 25% of these women will be “doubly” exposed and receive messages twice.) In analysis, we will also present results excluding beneficiaries reporting duplicate phone numbers.

Since PSI maintains a comprehensive administrative dataset and our analysis is restricted to the consenting participants, we do not expect true missing values. Any missing values for referral redemptions will be interpreted as absence of redemption (i.e.,“not redeemed”). We will not impute any covariates for those not reporting in the DHIS2 data.
However, data on message delivery will be provided by the SMS vendor that is managing the intervention and must be compiled and maintained by PSI. We are aware there is a significant risk of missing data here. If data is missing, we will code any recipients for whom delivery was not recorded as 1 for message delivery, thus ensuring a conservative estimate of the intervention’s implementation.