The methods are summarized in this section. Detailed methods are published in the trial protocol [32] and the statistical analysis plan [33].
Aim, study design and participants
This was a parallel group, individually randomized trial with a 1:1 allocation ratio. The aim was to assess the effect of the intervention on attitudes towards the non-permanent effective contraceptive methods [34,35,36] available in Palestine (oral contraceptive pills (OCs), intrauterine devices (IUDs), injectables, implants and the patch). Women were eligible to take part if they were between 18 and 24 years of age, did not report using an effective method of contraception, owned a personal mobile phone, lived in the West Bank and could read Arabic. The trial was promoted through PFPPA’s service delivery points and outreach sites, the PFPPA website and the distribution of trial promotional material via flyers and social media sites. We recruited participants through PFPPA’s service delivery points and outreach sites in Jerusalem, Bethlehem, Halhoul and Ramallah.
Intervention and control
The intervention was informed by the integrated behavioral model [37] and was sent by mobile phone text message. It was tailored according to marital status, resulting in two sets of intervention messages: (1) female-married and (2) female-not married. Most of the messages in the two sets overlap, with minor tailoring so that the messages are relevant to marital status (marital status was used as a proxy for sexual activity because in this context it was thought inappropriate to ask about sexual activity if not married). Participants allocated to the intervention group received zero to three messages per day (113 messages for female-not married and 120 messages for female-married) for 120 days.
The intervention messages provided information about contraception, targeted beliefs identified in the development phase that influence contraceptive use (e.g. misconceptions about the side effects and health risks of contraception, belief that non-hormonal methods are better because they are not harmful to health) and aimed to support young women in believing that they can influence their reproductive health. The intervention contained the following behavior change methods, adapted for delivery by mobile phone [38]: belief selection, facilitation, anticipated regret, guided practice, verbal persuasion, tailoring, cultural similarity, arguments, shifting perspective and goal setting.
Participants allocated to the control group received 16 control messages about trial participation over 120 days. Details on the development of the intervention can be found in the intervention development publication [31] and the trial protocol [32].
Allocation and intervention delivery
The online trial database and randomization system were used to generate the allocation sequence, and randomization occurred immediately after the baseline data were submitted by the clinic research staff. The system sent the Palestinian texting platform the allocation, preferred time slot for message delivery, mobile phone number and marital status.
Protecting against bias
Participants would have been aware of the allocation after they started receiving the messages. Allocation was masked from the research staff collecting outcome data unless the participant revealed it to them. Treatment allocation was masked from the researchers who analyzed the data.
Outcomes
Primary outcome
The primary outcome was the proportion of participants reporting that at least one method of effective contraception was acceptable at 4 months post randomization. The acceptability of each method was measured by the following stems: Using the [method] … causes infertility, … causes unwanted side effects, … is easy, … is a good way to prevent pregnancy and I would recommend the [method] to a friend. The IUD and implant included an additional stem: The [method] insertion would not be a problem for me. The response options for each stem were: strongly disagree, disagree, not sure, agree, strongly agree and I do not know what the [method] is. A method was acceptable if participants reported agree or strongly agree for all stems except for … causes infertility and … causes unwanted side effects, for which disagree or strongly disagree indicated acceptability [32].
Secondary outcomes
Secondary outcomes were the use of effective contraception at 4 months and any use during the study, the acceptability of individual methods and service uptake, unintended pregnancy and abortion.
Process outcomes
The process outcomes were knowledge of effective contraception; perceived norms and personal agency in relation to using and communicating with partners about contraception and intention to use effective contraception and intervention dose received.
Data collection
All outcomes were self-reported. At baseline, we collected personal and demographic data and the primary outcome data via self-completed paper questionnaire. We collected data on all outcomes at 4-month follow up. Staff, masked to treatment allocation, collected the follow-up data verbally by telephone.
Sample size
The trial was powered to detect a 15% absolute increase in the acceptability (corresponding to an odds ratio (OR) of 1.86) of effective contraception in the intervention group compared to the control group. A sample of 454 participants would provide 90% power to detect a 15% absolute increase in acceptability, at the 5% significance level, assuming 50% acceptability in the control group. We allowed for 20% loss to follow up and aimed to randomize 570 participants.
Statistical analysis
The trial protocol was accepted for publication on 14 September 2017 [32] and the detailed statistical analysis plan was publicly released before conducting the data analysis, on 7 November 2017 [33]. Analyses were conducted according to randomized group and only participants with complete outcome data were included in the principal analysis. All statistical tests were two-sided and considered significant at the 5% level. The analysis was conducted using Stata 15. Unmasking occurred on 6 February 2018, after the masked data were analyzed as outlined within the analysis plan.
Loss to follow up and missing data
We used the chi-squared test to investigate whether loss to follow up differed by trial arm. We used logistic regression to compare baseline characteristics of participants who completed follow up with participants who did not. We investigated whether predictors of loss to follow up differed by trial arm by testing for an interaction.
Principal analysis
Analysis of the primary outcome
We used logistic regression to compare the proportion of participants in each trial arm who reported that at least one method of contraception was acceptable. We report the crude and adjusted OR with the 95% confidence interval (CI) and p value. We adjusted the primary analysis regression for prespecified baseline covariates [32, 33].
Analysis of the secondary outcomes
The analysis of the secondary outcomes was similar to the analysis of the primary outcome, although in analysis of the acceptability of the individual methods, the acceptability of that method at baseline replaced the acceptability of at least one method at baseline as a covariate.
Analysis of the process outcomes
The process outcomes of perceived norms, personal agency and intention each comprised ordinal scales. Each scale was analyzed individually using ordered logistic regression. For knowledge, we used linear regression to assess the difference in the mean scores between the groups. To quantify the “dose” of the intervention that the intervention participants received, we analyzed the number of messages that participants reported to have read (all, most, some, none) and whether they stopped the messages, along with our monitoring data.
Additional analyses
Sensitivity analyses
We conducted two sensitivity analyses to account for missing data. In the first, we considered that all participants who were lost to follow up did not find at least one method acceptable. In the second, we adjusted for the main baseline predictors of missingness. Both sensitivity analyses were adjusted for the baseline covariates as aforementioned.
Subgroup analysis
We conducted an exploratory subgroup analysis for the primary outcome with prespecified subgroups [33]. Within the subgroups, we assessed the heterogeneity of the estimated treatment effect with a test for interaction [39, 40]. We estimated ORs with 95% CIs for each subgroup.
Contamination
We report the proportion of control group participants who reported that they read another participant’s messages and the proportion of intervention participants that reported that their messages were read by another participant.
Report of physical violence
We report the proportion of participants in each group that reported experiencing physical violence during the study.