Tunatar da ni. The Immunization Reminder and Information SMS System (IRISS)

Background and rationale:

Vaccines are one of the most effective and cost-effective interventions available to public health today. But the benefits of vaccines are not getting to all children, particularly in Nigeria. The 2016 National Immunization Coverage Survey (NICS) reports that only 33% of children aged 12-23 months received the third dose of the pentavalent vaccine nationally. In addition, as much as 40% of children had zero dose of vaccines from the routine immunization (RI) program by the age of two years, having been left out of the "service grid". Even among children that have access, retention and utilization is poor; as much as 31% of children who received their first pentavalent vaccine dose, drop out. They do not return to complete the third dose in the series.

In the 2016 NICS, barriers related to lack of knowledge were the most frequently cited barriers by parents of under-vaccinated children. In Nigeria, RI is delivered through fixed sessions at health clinics and outreach sessions in the communities. To complete the RI schedule, caregivers are required to proactively take their infants to the service sites a total of 5 times in the child's first year of life. As much as 42% of parents cited knowledge-deficits such as, not appreciating the importance of vaccines, not knowing the schedule or not recognizing the need to seek out vaccination services as reasons for not vaccinating their children. The complacency to seek RI services may be linked to some parent's erroneous belief that the door-to-door vaccination of their children during polio eradication campaigns (with oral polio vaccine) provides their children with all the vaccines they need. Low confidence, inconvenient clinic schedules and location may also underline the poor demand for RI services.

The health belief model and theory of planned behavior provides an organizing framework to articulate how knowledge, perceptions, subjective norms, self-efficacy and cues to action influence behavioral intention and action. Based on this theory, the team identified that Short Messaging System (SMS) messages and reminders could serve the three-fold purpose of providing information, shaping subjective norms and cueing caregivers to action.

Study investigators hypothesize that providing community leaders with positive and actionable messages on immunization services will improve their understanding of the value of vaccines and provide them facts to drive discussions about vaccination in the community. These discussions about vaccination will result in a common understanding of its importance, increase support for it and build greater positive norms about it in the community. Furthermore, if vaccination schedules for local clinics are sent to the community leaders through SMS, and they, in turn, ask their town announcers to announce the session times in the community, this may serve as cues to action for parents of vaccine age children. In the same vein, providing targeted reminders to parents about their child's vaccination due date and the schedule of their health center will motivate their timely action to seek vaccination services for their children.

The evidence that SMS reminders improve vaccine uptake and coverage is strong, however, most of these studies come from small research settings with tightly controlled and well-resourced program infrastructure. The feasibility of implementing an SMS messaging and reminder intervention like this, at scale, and in a rural and low coverage setting like Kebbi, will be tested by integrating the intervention into RI services, using existing government immunization providers and managers.

This is a mixed methods study comprising a qualitative component (formative study) to understand the context, test the messages and refine intervention strategies, and a quantitative component (evaluative studies) to evaluate the impact of the intervention on immunization uptake, with a cluster randomized controlled trial (CRCT), covering all LGAs, and a complementary before-and-after study, in one sentinel LGA. A costing of the program will also be done to evaluate affordability by the government. Only the CRCT is described here.

The objective of the CRCT is to: assess the impact of the SMS interventions on the age-appropriate rates of routine vaccination among children 0-11 months of age.

Methods. The intervention.

The setting: Kebbi state, is one of the 37 states in Nigeria, with an estimated population of 4.4 million and 180,000 newborns every year. Located in the Northwest of the country, Kebbi has 21 LGAs further divided into 225 wards. Each ward 1 to 3 health facilities.

The system: The project will develop the SMS platform, The Immunization Reminder and Information SMS System (IRISS), to register contacts, manage the contact database and coordinate, schedule and send the messages, as well as track messages sent. The SMS messages will be sent in three ways:

The message targets: As described above in the summary, the project will send out a general SMS broadcast to active phone users in the intervention LGAs, targeted messages to community leaders and other community members who opt-in for the messages, and individualized reminders of a child's vaccination dates and clinic schedule to parents of infants who opt-in for the messages. The SMS messages will be deployed over a period of 9 months between June 2019 to March 2020.

All individuals, including Leads, community gatekeepers, community volunteers, health workers, and caregivers living in the intervention LGAs will be eligible for enrollment into the IRISS e-registry without limitations if they meet the following criteria: for the public broadcast - have an active phone; for the targeted messages to traditional (Mai-unguwas) and religious leaders (Imams) at the community level - have a phone or access to one; for caregivers - have a vaccine-age child (less than 12 months of age), have a phone or access to one, and currently reside in the intervention area.

Based on the number of community leaders in the 14 intervention LGAs, the phone ownership rate, and an assumption of the level of uptake, the projection is that about 7,000 traditional leaders, 1,000 Leads, and 72,000 parents will be enrolled in IRISS making it a total of 80,000 enrollees. The projected enrollment of 72,000 parents of newborns is based on an assumption that 20% of the parents of the 360,000 babies born annually in the intervention LGAs will uptake the service.

The messages. Four types of messages will be sent: normative, educational, informative and motivational. Vaccination schedule reminders will also be sent.

Advertising IRISS: To sensitize community leaders about IRISS and build capacity of RI providers to assist caregivers with opting in, when needed, the study will leverage a community engagement strategy (CES) being rolled out by the government. This strategy aims to involve traditional leaders in the registration of newborns, reconciliation of the data on children vaccinated, comparing community and health facility records, and the follow up of children who missed their vaccine doses. The state immunization officers will conduct cascade of sensitization meetings from the Emirs at the top of the traditional leadership ladder to the Mai-unguwas at the lowest rung. This CES will be implemented in both intervention and control LGAs by government officials, the difference is that IRISS sensitization module will be presented only in the intervention areas. It is likely that advert for enrollment into IRISS for health information may spill over to control LGAs, particularly if information is shared by the public on social media in ways beyond the project's control. If this happens and individuals from control LGAs register into IRISS expecting a message, they will receive a general message about the importance of hand-washing. This is to avoid people losing trust in the program, if they feel SMS messages were advertised but not sent.

Evaluation - CRCT Clusters were defined as the LGAs, the next lower administrative unit below the state. In line with the goal of implementing at scale, all 21 LGAs in Kebbi state were eligible to participate. The 21 clusters (LGAs) were randomly assigned in a 2:1 ratio to maximize the number of LGAs receiving the intervention while maintaining sufficient power to detect meaningful difference between the two parallel arms. The intervention arm (14 LGAs) are to receive both the scheduled and responsive SMS messages on immunization. The control arm (7 LGAs) is designed to receive no SMS intervention, unless some individuals erroneously registered in IRISS, then only those individuals received a one-time message on hand-washing.

Randomization will be accomplished with un-stratified covariate-constrained randomization using a macro accessible from: https://github.com/ejgreene/ccr-sas to generate multiple allocations of each LGA into one of three groups. The covariates are: phone ownership/access (continuous variable) - % of population with mobile phones; target population (continuous variable) - number of children 0-11 months; immunization coverage (continuous variable) - % of children 0-11 months appropriately vaccinated for age; emirate (4 binary variables for each emirate where the variable equals 1 if the LGA is in that emirate and 0 otherwise).

Each LGA will be allocated to one of three groups, and the process is repeated 100,000 times. To achieve the 2:1 ratio, two groups will be combined later to form the intervention arm, while the third becomes the control arm. For each of the 100,000 allocations, the within-group mean for each covariate will be calculated and compared to the overall state covariate mean. The allocations where the groups' covariates means fall with 30% of the state covariate mean, will be kept, otherwise, they will be discarded. The criteria used for the emirate variables is that the groups could not differ in size by more than 1 LGA.

Once the qualifying allocations are generated, the next step is to conduct a transparent and public draw. At a meeting of stakeholders from the state primary health care development agency, a volunteer will be asked to randomly select one allocation out of those that qualified. For example, if 200 allocations met the criteria, a random number from 1 to 200 will be generated from www.numbergenerator.org. The allocation corresponding to the generated number is then taken. Next, the groups in the selected allocation are assigned to either intervention or control by having three volunteers draw three balls labelled - intervention 1, intervention 2 and control from an opaque black bag. The groups assigned intervention 1 and 2 will be collapsed into one intervention arm.

Sample size: In a CRCT, sample size depends on both the number of clusters and of respondents within each cluster. In this study, the number of clusters is 21 LGAs with 14 intervention and 7 control LGAs. Similarly, the number of respondents per LGA in the Lot Quality Assurance Surveys (LQAS) that study team plans to leverage to assess impact, is fixed at 60. The LQAS are conducted by the government quarterly. Given a fixed sample size, the power of the study to detect a meaningful difference in the primary outcome at the cluster level was calculated as follows:

A total of 60 households will be sampled from each LGA. One infant per household will be interviewed through their caregivers. Total no. of respondents is 21*60 =1,260. Based on the 2:1 allocation ratio of the 21 clusters, a baseline rate of 12% of infants appropriately vaccinated for age in the state, assumed increase to 30% in intervention arm after 9 months of intervention, a design effect of 3, type 1 error of 5%, a one-sided test; the study has 99% power to detect a difference of 18% percent between the intervention and control LGAs.