Application of behavioral economics principles to reduce injectable contraceptive discontinuation in rural Ethiopia: A stratified-pair, cluster-randomized field trial

Abbreviations

BE                     Behavioral Economics

CPR                  Contraceptive Prevalence Rate

CBDDM           Community-Based Data for Decision Making

DHS                 Demographic and Health Survey

FMOH             Federal Ministry of Health of the Government of Ethiopia

FP                    Family Planning

HEW               Health Extension Worker

HEP                 Health Extension Program

IUCD               Intra Uterine Cervical Device

JSI                   John Snow, Incorporated

L10K               Last Ten Kilometers

LAFP               Long Acting Family Planning

MNCH             Maternal, Newborn and Child Health

PHCU              Primary Health Care Unit

RMNCH          Reproductive, Maternal, Newborn and Child Health

SNNP              Southern Nations, Nationalities and Peoples’

WDA              Women Development Army

Introduction

The national family planning program in Ethiopia has made considerable strides in providing access to services across the country. The contraceptive prevalence rate (CPR) in Ethiopia jumped from 6% in 2000 to 36% in 2016. This increase was mainly the result of a sharp increase in the use of injectable contraceptive methods, from 3% in 2000 to 23% in 2016. Nonetheless, 35% of contraceptive users discontinued use within 12 months of uptake, while the discontinuation rate among injectable contraceptive users was 38% (Central Statistical Agency and ICF, 2016).

The Ethiopian Federal Ministry of Health (FMOH) aims to increase the CPR to 55% by 2020 (Federal Ministry of Health, 2015). High rates of discontinuation pose a threat to the achievement of this target, as well as to the intentions of many Ethiopian women to space births and to limit the number of children they have. High rates of contraceptive discontinuation impose a cost on the woman, her family, and her country’s health system. Discontinuation can lead to unintended pregnancies, causing social, economic, and emotional distress, and also contribute to maternal morbidity and mortality (Ahmed et al., 2012; Ali et al., 2012; Cleland et al., 2012).

Women may choose to discontinue a contraceptive method because their reproductive intentions have changed, and they no longer intend to use family planning. However, if women are discontinuing, but still have intentions to use contraception, the system has not met their needs. Discontinuation rates may be high for a variety of reasons. Some of these are problems at the system level—for example, high rates of stockouts or distance from the source of the family planning method could be structural barriers impeding continued use. However, various nonstructural factors may also contribute. Often, the dominant reason for stopping a modern, reversible method is dissatisfaction (World Health Organization, 2012). Dissatisfaction with the method could stem from gaps in counseling that lead to suboptimal initial method choices, health concerns or side effects, or choice of an inconvenient method, among other reasons (Bradley et al., 2009). In contexts where structural barriers have for the most part been eliminated, nonstructural, or behavioral, factors may be driving discontinuation. Ethiopia is likely to be such a context. The introduction of the country’s flagship Health Extension Program (HEP) has brought family planning services to the doorsteps of the rural population (Admasu et al., 2016).

Behavioral economics (BE) offers a promising toolkit for designing interventions to change health-related behaviors and decision-making, from both the provider and the client perspective, by improving understanding of why people behave as they do and what motivates their decision-making and action (Buttenheim & Asch, 2013; Datta & Mullainathan, 2014). Contraceptive discontinuation has been identified as a critical reproductive health challenge that is persistent across contexts and may be responsive to behavioral economics interventions. Contraceptive discontinuation is especially relevant for a behavioral economics lens because there is an explicit gap between stated preferences and behavior (Ashton et al., 2015).

Compounding the cognitive factors surrounding contraceptive use, recent behavioral science research suggests that individuals living in poverty experience a chronic state of scarcity, which imposes additional cognitive demands that make forward-looking decisions and actions difficult (Mani et al., 2013; Shah et al., 2012). Individuals experiencing scarcity tend to “tunnel”; focusing on what is most urgent or most salient, displacing all else. This may have harmful effects on future-looking decisions or actions like those involved in contraceptive use. In low-income countries, health workers as well may experience diverse forms of scarcity, living in poverty themselves and working in contexts characterized by large catchment areas, heavy workloads, personnel shortages, and limited time or task management resources.

Behavioral economics offers many promising solutions to address these types of behavioral challenges. Reminders have proven effective in promoting medication adherence and medical appointment attendance across several contexts (Kannisto et al., 2014; Lindenmeyer et al., 2006). Job aids are another widely used behavioral solution that has been effective in improving family planning counseling (Kim et al., 2007; Leon et al., 2005). Insights from BE and a structured behavioral design methodology have recently been used to improve reproductive health programming (Buttenheim et al., 2019; McConnell et al., 2018; Spring et al., 2016), suggesting that the approach could be applied to mitigate the contraceptive discontinuation problem in Ethiopia. Further application is needed to understand how insights from behavioral economics can be applied to solve persistent public health challenges particularly in low-resource settings where scarcity increases the likelihood of intention-action gaps. This paper contributes to this understanding by sharing the experience and results of the application of behavioral economics to a family planning program in rural Ethiopia.

Abandonment of injectable methods is an urgent matter, given that they are the primary method of choice among married women of reproductive age in Ethiopia, and that the government sector provides 82% of contraceptive methods (Central Statistical Agency and ICF, 2016). The Last Ten Kilometers (L10K) Project of John Snow Research & Training Institute, Inc. (JSI) collaborated with ideas42 to work with the HEP to design and test behavioral approaches to mitigate the problem of discontinuation of injectable contraceptives.

The overall goal of this study is to increase the rate of contraceptive continuation among women of reproductive age who are new users of injectable contraceptives in rural Ethiopia. The primary objectives in support of this goal are 1) to use behavioral insights to develop a package of interventions that will improve the rate of contraceptive continuation of women using injectable contraceptives in Ethiopia and 2) to trial this package of interventions and assess the preliminary impact of this package and its limitations.

Methods

The intervention

To identify behavioral barriers impeding continued use of injectable contraceptives, design a targeted solution, and test the impact of this approach, the study team applied a four-stage behavioral diagnosis and design methodology. This process involved four distinct phases depicted in Figure 1: problem definition, behavioral diagnosis, intervention design, and testing (Datta & Mullainathan, 2014).

Figure 1. Stages of Behavioral Design methodology (Datta & Mullainathan, 2014).

Problem definition (Step 1). The general problem of contraceptive discontinuation is broad, and it encapsulates diverse methods, subpopulations, and manifestations of discontinuation. Thus, the goal of the problem definition process was to (a) identify the different problem definition candidates that represent these multiple dimensions and components of discontinuation in rural Ethiopia, and (b) evaluate these candidate definitions systematically along with a set of criteria so that the strongest and most actionable problem or problems are identified. The criteria were:

• 1. The key drivers for discontinuation of contraceptives included behavioral issues among others (e.g., cultural issues, problems with the supply chain).

• 2. The existence of an intention-action gap, i.e., women have the desire to continue using contraceptives, but fail to follow through on this intention and discontinue.

• 3. The problem could be moved by a behavioral intervention; and,

• 4. Structures were in place for implementing cluster-randomized clinical trials to test the effectiveness of the behavioral intervention.

Problem definition included the analysis of 2011 Ethiopian Demographic and Health Survey, informal conversations with L10K senior and mid-level managers, process mapping, preliminary behavioral mapping, and a field visit in October 2014. The field visit included conversations with frontline HEP workers (i.e., HEWs and the WDA) and contraceptive users, along with observation of family planning provision.

The process gathered anecdotal evidence that both passive (unconscious decisions) and active (conscious decisions) contraceptive discontinuation could be occurring. All long-acting (i.e., IUCD and implants) and short-acting (i.e., injectables, pills, and condoms) reversible methods may be subject to active discontinuation, while only short-acting reversible contraceptive methods may be subject to passive discontinuation. Several candidates for defining the problem were identified and evaluated based on the four criteria described above. The final problem definition from the process was: “Women of reproductive age who are using injectable contraceptives discontinue use within 12 months of uptake.”

Behavioral diagnosis (Step 2). The objective of the behavioral diagnosis was to identify the primary behavioral drivers or the “behavioral bottlenecks” leading to discontinuation of injectable contraceptives within the study population. The process included detailed behavioral mapping to generate hypothesized bottlenecks; contextual reconnaissance to gather evidence; and synthesis of evidence and evaluation of hypothesized bottlenecks. Evidence gathered during the contextual reconnaissance stage included data from client and provider interviews, observations of family planning provision, analysis of data from Demographic and Health Surveys, conversations with project staff, and analysis of secondary literature. This evidence was used to confirm, reject, or refine the hypothesized bottlenecks.

The set of hypotheses that described the primary bottlenecks driving observed behavior (i.e.: discontinuation of injectable contraceptives) and its intervention design implications are presented in three categories: bottlenecks contributing to passive discontinuation, active discontinuation, and provider behavior (Table 1).

Table 1. Overview of behavioral bottlenecks and their design implications.

Behavioral Bottlenecks	Design Implications
Bottlenecks Contributing to Passive Discontinuation
1. Prospective Memory Failure to Follow Through on Next Injection: Due to prospective memory failure, women do not come to the health post at the appropriate time.	Salient, timely reminders for women.
2. Tunneling Leads to Myopic Planning or Failures to Plan for Next Injection: Women tunnel on the specific date of a follow-up appointment, yet they do not form any contingency plans for receiving their injection if they cannot make that exact date.	Prompt plan-making and contingency planning for return appointments.
3. Hassle Factors: Women may be deterred by hassles associated with successful continuation of injectable contraceptives.	Timely reminders of fertility intentions.
Bottlenecks Contributing to Active Discontinuation
4. Perception of Limited Choice Set for Continued Use: Women experience and wish to eliminate side effects from injectable contraceptives, yet they do not feel that they have viable alternatives, nor do they consider switching methods.	Expand the choice set of methods women consider to be viable options; increase and maximize HEW and WDA touch-points with women.
Bottlenecks Contributing to Provider Behavior
5. Bandwidth Tax: HEWs have limited time and attention to devote to client tracking and case management. Thus, the date of clients’ follow-up appointment is not salient to HEWs at the appropriate time to provide prospective appointment reminders and retrospective follow-up for missed appointments.	Simplify case management systems to reduce the cognitive load of HEWs.

Intervention design (Step 3). Building directly on the behavioral diagnosis and related design implications, a package of scalable interventions was selected and refined to ensure that the interventions were feasible, culturally acceptable, aligned with HEP policies, contextually appropriate, and human-centered. To ensure human-centered design, the tools were co-created with the HEWs and clients. The intervention design package included three components—a HEW planning tool located at the health post; a client counseling job aid used by HEWs during service provision; and appointment cards distributed by HEWs during service provision and used by clients (Figure 2).

Figure 2. Clockwise from left: appointment card, client counseling job aid, and Health Extension Worker (HEW) planning calendar.

HEW planning calendar: The planning calendar is a desktop-based calendar, located at the health post and utilized by the HEW. The planning calendar serves as an aid for the HEW when scheduling future appointments and outreach, and it guides her toward better time management and planning.

Client counseling job aid: The counseling job aid is a simple guide to frame the HEW service provision interaction with clients to (1) to prompt HEWs to provide high-quality services at every interaction and to create an atmosphere in which the client feels comfortable and supported, (2) prompt the key components of family planning counseling, and (3) reduce the extent to which a provider needs to remember exact protocols for setting return appointments and managing records. Three job aids—one for counseling on available family planning methods for informed choice; one for counseling new family planning clients; and one for counseling returning clients—are displayed on three sides of a triangular tower and kept on the HEW’s counseling table.

Appointment cards: HEWs distribute appointment cards to all clients requiring return appointments. They remind clients about future appointments and provide critical information on what to do if they cannot make their appointments or experience health concerns in the interim.

Test (Step 4). The final step of the behavioral design methodology was to test the package of interventions to assess implementation feasibility and impact on the outcomes of interest.

Results

In general, the catchment populations of the health posts in the intervention areas were near twice those of the control areas. The average catchment population of the health posts in the intervention area was 7,868; for the control area it was 4,064.

Assessment of the implementation status of BE tools at the health posts indicates that 12 of the 19 health posts in the intervention area had appointment cards available on the day of the visit, while only two health posts were stocked out of the appointment cards for more than a year (Table 3). Client counseling towers were available in all the intervention area health posts, and they were on top of each HEW’s counseling table. However, the HEWs from one of the intervention area health posts indicated that they did not use the counseling job aid. Use of HEW the planning calendar was poor. Only four of the intervention area health posts were using the planning calendar appropriately, while the rest were not using it during the four weeks preceding the survey, or they did not have the calendar for the most current year.

Health post records indicated that 2,490 women in the intervention area and 1,684 women in the control area visited the study area health posts between February 2016 and November 2016 to obtain injectable contraceptives (Table 4). Among them, 335 women (13%) from the intervention area and 408 women (24%) from the control area were eligible for the study and were interviewed after obtaining consent; 688 women (28%) in the intervention area and 550 women (33%) in the control area were not eligible; 1,047 women (42%) in the intervention area and 559 women (33%) in the control area could not be located due to erroneous records; while 420 women (17%) in the intervention area and 167 women (10%) in the control area could not be reached for interview. If the eligibility rate for participating in the study among those who could not be located for an interview was similar to those who were located, then the loss to follow-up of eligible study participants was 22% in the intervention area and 18% in the control area. There were no non-responses among those who were eligible.

Table 5 compares the background characteristics of the study participants between the intervention and control areas. Compared to those of the control area, the intervention area participants were less likely to have higher education (p < 0.05); and the age of the last child was less likely to be more than three years (p < 0.001). Otherwise, participant characteristics between treatment and control areas did not differ with statistical significance.

Table 6 shows the distribution of the study participants according to the health system functional status of the health posts to which they belonged. Although it appeared that compared to participants from the control areas, participants in intervention areas were more likely to be in the catchment area of health posts with better-functioning health systems (i.e., more Level 4 HEWs, higher density of active WDA team leaders, fewer contraceptive stockouts, more frequent supportive supervisory visits, more frequently updated family folders, more updated CBDDM registers, and less likelihood of closure for more than a month during the past two years), the differences were not statistically significant. However, the HEWs in the intervention area were serving a larger population. Of the participants in the intervention area, 81% were in the catchment area of health posts that had more than 2,500 people for every HEW deployed, compared to 37% in the control area (p < 0.01).

The frequency distribution of person-months of follow-up according to the independent variables can be found in Table A1 of the data repository (see extended data (Karim, 2019)).

Table 7 gives the adjusted point estimates of the outcomes of interest by study arm and the intervention effects with 95% confidence intervals. The adjusted estimates considered the differences between the study arms observed in Table 5 and Table 6. Figure 3, Figure 4 and Figure 5 display the analysis in Table 7 using figures. The mean number of side effects from injectable contraceptive methods recalled was 0.24 higher (p < 0.05) among the intervention area participants than among those in the control areas. The proportion who recalled being told of a local holiday or event as a reminder for the next appointment for injectable contraceptives was 16.7 % points higher (95% Confidence Interval [CI]: 7.8 %, 25.6%) among the intervention area participants than among those from the control areas (p < 0.001). Although the intervention’s effects on the proportion told about side effects, proportion told what to do if they had side effects, the proportion told about other methods, the proportion given appointment card, and the proportion who missed any appointment were in the expected directions, they were not statistically significant (p > 0.1).

Figure 3. Mean number of side effects, family planning (FP) methods, and action to take for side-effects recalled by the participants.

Figure 4. Percentage of participants who receiving method information from the provider, given appointment card and missed an appointment.

Figure 5. Injectable contraceptive discontinuation rate.

The primary outcome of interest, i.e., injectable contraceptive discontinuation rate within 12 months of uptake, was 10.8 % points lower (95% CI: 2.4 %, 19.1%) among intervention area participants than among control area participants (p < 0.05) (Table 7 and Figure 5).

Table A2 in the data repository (see extended data (Karim, 2019)) gives the list of independent variables that were used in the final models to obtain the adjusted intervention effect estimates.

Discussion

This is a unique study that applies a behavioral design methodology to reduce injectable contraceptive discontinuation in rural Ethiopia. The results of this study indicate that behavioral design methods are a promising approach for creating effective public health solutions. The four-stage methodology was feasibly implemented in a participatory manner alongside uninterrupted program and service delivery. We show that two of the three elements of the behavioral intervention were implemented with fidelity and that, as a package, were effective in decreasing injectable contraceptive discontinuation in the study environment, thus supporting the hypotheses generated during the behavioral diagnosis phase.

Several factors related to the internal and external validity warrant discussion. First, erroneous identification records of injectable users in the study area health posts were substantial. It is common knowledge among HEP workers, though not well documented, that some contraceptive users choose to use injectables discreetly and provide false identities to obtain services. The data collectors confirmed this assumption with the HEWs in the study area. Poor record-keeping at the health posts may also be partly responsible for the errors. Second, differential loss to follow up by study arm could have biased the study result. However, if the injectable contraceptive discontinuation rate among women who were lost to follow-up was similar to that of women who were not lost to follow-up within the respective study arms, then the estimate of the intervention effect was unbiased. Although we cannot directly test differential loss, we show that age, religion, and region of residence are all broadly similar across intervention groups. If these characteristics are the primary drivers of whether clients provide false identities, we would not expect the desire to use injectables discreetly to differ by treatment arm either. Third, the threat to the validity of the randomization is that there was no baseline on a wide range of characteristics and that some of the treatment effects are highly sensitive to the specific controls used. To validate these treatment effects (especially for the primary outcome) we rely on the backward selection process using variables in Table 5 and Table 6, and the model specification derived from it, as most appropriate and that unobserved differences do not drive the measured treatment. Fourth, the design of participant recruitment and eligibility may have introduced sample selection bias. Specifically, participants were eligible for the study if they were new injectable users (defined as someone new to family planning, switching from one method to another, or who had taken a hiatus from family planning for six months or more). Because the intervention reduced discontinuation and encouraged dissatisfied users to switch methods, it is possible that the study participants who were included during the latter part of the enrollment period were different between the study arms. The intervention arm may have included more women who were switching to injectable contraceptives from a prior method. Table 5 and Table 6 show that for both individual- and provider-level characteristics, there is broad balance between treatment and control participant. However, we are unable to completely eliminate the possibility of this bias. Fifth, having more than 12 months of the observation period, there was an opportunity for health system factors (for example, injectable supply and availability of HEWs) to mask the effects.

Injectable discontinuation rates in both arms of the study were much higher than initially expected. The difference in the representativeness and measurement method of the discontinuation rate between the expected (nationally representative data with a five-year recall period) and the observed (which was in eight districts with one-year recall period) may partly explain it. It could be that the respective Regional Health Bureaus purposely selected low performing districts for L10K to support the family planning program. Nonetheless, variations in the injectable discontinuation rate between geography is observed in Ethiopia. An independent study in 2014 in Agarfa district, Oromia, report significant higher rates of injectable discontinuation (67%) than the national average (Bekele et al., 2015).

Results on the primary outcome of this study show that the overall intervention package was successful in reducing injectable contraceptive discontinuation. Examining the secondary outcomes of interest sheds light on the distinct components of the intervention package and the accuracy of the behavioral diagnosis insights. Study results indicate that the appointment cards and the client counseling job aid were utilized in the intervention areas, whereas the planning calendar was generally not utilized by HEWs. Before the study, the existing family planning client tracking system for HEWs was the tickler filing system to track defaulters (Chewicha & Azim, 2013). The behavioral diagnosis revealed that it was not a useful tool to help HEWs provide prospective appointment reminders for the next dose of injectable contraceptives nor was it an effective tool to track missed doses. The planning calendar was designed to help the HEWs provide prospective follow-up reminders to the FP clients. However, it is unlikely that the HEW planning calendar had any effect on the primary outcome, mainly because nine of 19 intervention area health posts did not have the most recent planning calendar, and among the ten health posts that had copies of the most recent planning calendar, six were not using them. One of the possible reasons for not using the planning calendar could be that the HEWs were already overwhelmed by the existing recording requirements of the HEP, and the planning calendar was an extra recording burden for them. The planning calendar was a parallel system to the HEWs’ existing record-keeping efforts.

Study participants in both arms reported receiving appointment cards from the HEWs during the last time they visited the health post to obtain injectable contraceptives. This was not surprising because, during the problem definition and behavioral diagnosis exercises, it was noticed that some HEWs provided injectable contraceptive clients a paper note as a reminder for the next appointment. Thus, formal appointment cards institutionalized an already used ad-hoc strategy used by the HEWs as one of the intervention tools. Additionally, while providing the appointment card and counseling clients, the HEWs in the intervention area were advised to relate the appointment date for the next injectable dose to a local holiday or an event and to write it down in the appointment card. Thus, as expected, it was found that the intervention area study participants were more likely to recall being told about a local holiday as a reminder for the appointment date. Given it was an already-used practice by HEWs in this context, associating the return appointment date with a local event or holiday likely increased appointment salience and recall, even without the availability of printed appointment cards. The use of the formalized appointment card validates the participatory and user-centered behavioral design approach in which solution concepts are sourced from the end-users themselves and designs build off of existing practices, habits, and patterns of behavior.

Recall of the side effects of injectable contraceptives was also higher among the intervention area participants, which is consistent with HEWs using the client counseling job aid component of the behavioral package. Although L10K provided refresher training to HEWs in both study arms on the family planning client care job aid provided by the national program, this job aid was long and cumbersome for counseling clients, and rarely used by HEWs. The behavioral design process made the original job aid more concise by increasing the salience of key messages and reducing the amount of information included. The new job aid was mounted on a desktop tower (rather than presented in a book or table tent that required page-turning) to be readily accessible and salient during client counseling. These two design adaptations may have made the job aid more usable and effective for HEWs.

It is interesting to note that using the appointment card and the client counseling job aid were within the routine workflow of the HEWs, but the planning calendar required an alteration of the HEWs’ workflow. Thus, HEWs complied with the use of appointment cards and client counseling job aid, but not with the use of the planning calendar. This is a critical insight for solution design—both within global health programs and across behavioral design more generally. Designs that introduce new tools or processes within an existing workflow, require an additional step or action, or require the shifting of habits, should be carefully considered and heavily pre-tested prior to implementation.

The study design cannot distinguish whether the appointment card or the counseling job aid was more effective. Ideally, the tools could have been tested one by one, with a shorter duration of observation, to identify the effectiveness of each. The one-year observation period required to measure the primary outcome for the study prevented the use of rapid tests one by one. This could be a blessing because it is possible that the appointment card and counseling job aid had a synergistic effect. Nonetheless, a shorter testing duration would have been beneficial to keep policy-makers interested, and to convert research findings into policy more rapidly.

The behavioral diagnosis pointed to both provider-side and client-side behavioral bottlenecks contributing to discontinuation of injectable contraceptives. Given the available data from this study, it is difficult to disentangle exactly how the intervention package interacted with each specific bottleneck. However, we can offer some observations. On the provider-side, the diagnosis revealed that HEWs experienced a condition of scarcity driven by heavy workloads, time-intensive tasks, and a high number of services to provide. This context of limited time and attention may have led HEWs to tunnel on more pressing issues, putting aside lower priority client tracking and case management. Low utilization of the planning calendar (which would have required an additional step or process by the HEWs) reinforces this insight. Though the planning calendar was designed to ease HEWs bandwidth constraints, it does not seem to have been effective in this regard. The client counseling job aid and the appointment card, while also targeting client-side bottlenecks, also served to address the HEWs taxed bandwidth. These tools acted as salient reminders and physical cues prompting providers to employ a specific protocol and provide key pieces of information during appointments.

On the client-side, the diagnosis highlighted behavioral bottlenecks contributing to both passive and active discontinuation. The appointment card targeted passive discontinuation by serving as a physical reminder for a future appointment and providing a salient memory cue (the closest holiday to the appointment) to help clients remember the appointment date. It may have also targeted active discontinuation by including personalized information from the HEW (name and phone number), thereby offering a solution if the client was actively considering discontinuing. The client counseling job aid targeted active discontinuation by ensuring that the HEW provide key messages around side effects, continued use of the method, and what to do if the client was dissatisfied with the method or missed a follow-up appointment. Further exploration would be necessary to understand in greater detail exactly how the intervention package targeted the five individual bottlenecks identified in the diagnosis.

Across contexts, dissatisfaction is the dominant reason for stopping a modern, reversible method (World Health Organization, 2012). Dissatisfaction may stem from side effects, among other reasons (Bradley et al., 2009). The client counseling job aid prompted HEWs to discuss with clients both side effects and what to do if side effects were experienced. The results show that clients in the intervention area were able to recall a higher number of side effects than the control area. However, across the intervention and control areas, there was no significant difference in client recall of the actions that could be taken for side effects. Given the reduction in discontinuation in the intervention area as a result of the intervention package, this represents an interesting data pattern. Perhaps the awareness of potential side effects alone was a meaningful factor in reducing discontinuation. Behavioral research shows that not only do we react more strongly to negative events than positive ones (Baumeister et al., 2001; Taylor, 1991), our responses are heightened in response to surprising or unexpected events as compared to expected events (Shepperd & Mcnulty, 2002). The suggestion that the mere awareness of potential side effects may lessen the intensity of a negative experience is an area for future practical exploration.

Conclusion

The overall findings of this study suggest two primary conclusions: first, that the behavioral design methodology is a useful approach for solving persistent behavioral challenges in family planning programs, and second, that two components of the intervention package (the appointment card and the client counseling job aid) are promising solutions for enhancing family planning programs in this low-resource, rural setting in Ethiopia. The behavioral design methodology was successfully implemented by a multidisciplinary team to identify novel behavioral insights around injectable discontinuation, design a context-appropriate solution, and test the effectiveness of the intervention package. Two of the three behavioral tools in the intervention package were consistently utilized by HEWs in the intervention area. A reduction in injectable discontinuation in the intervention area of over 10 percentage points can be attributed to the design package. In addition, increases in secondary outcomes such as the number of side effects recalled and use of a local holiday or event as an appointment reminder were consistent with the behavioral intervention being utilized effectively to reduce discontinuation.

The appointment card and the client counseling job aid emerged as the two tools that HEWs most consistently used in service provision and appear to seamlessly integrate into their routine workflow. This finding is critical for behavioral design generally. Where possible, new solutions must seamlessly integrate into existing patterns of behavior, workflows, or processes. If new solutions being introduced require new habits, processes, or steps, they may need additional scaffolding and support to ensure successful utilization.

We show that the behavioral design methodology is a practical approach for identifying relevant insights and designing targeted solutions. The study’s results indicate that insights from behavioral economics can improve family planning programming and outcomes. The resulting intervention package reduced injectable discontinuation, likely reducing unintended pregnancies and improving participant health and autonomy.

Consent

During participant recruitment, after eligible participants received services at a health facility, HEWs informed them about the study and the data collection activities. They were given the opportunity to ask any questions about the study and its participants, and then were asked if they consented to participate. HEWs were supposed to have first asked for verbal consent and then supply a written consent form to record participant contact information for the follow-up interview. All HEWs reported having administered and secured verbal consent to all eligible study participants. However, very few paper consent forms were retrieved from health posts. This implementation failure presents an unknowing violation of client privacy. Learning from this experience will inform how consent is secured in future activities.

Therefore, for the FP client follow-up survey, verbal consent was sought and documented by the interviewers. If the respondent was under 18 years old, then consent was sought from her husband, parents or guardian. As it was expected that most of the respondents could not read or write, written consent was not sought. Only when consent was given by the respondent, the electronic data collection tool allowed to continue with the survey.

Data collection from health posts was considered a structured routine supportive supervisory visit. Data was collected mainly through observations and review of health post records and extraction of administrative data. However, verbal consent from the HEW or HEWs (when more than one HEW was present during health post visit) were obtained before continuing with data collection from the health post.

The consent forms for the FP client follow-up and health posts data collection were part of the data collection tool and can be accessed with the data repository.

Ethical approval was obtained from the ethical review boards of the Regional Health Bureaus of Amhara, Oromia, SNNP and Tigray regions and from the ethical review board of JSI Research & Training Institute, Inc. The ethical approvals can be accessed with the data repository.

This study was retrospectively registered with the International Standard Randomized Controlled Trials Number registry (ISRCTN17390653) on the 10 April 2019.

Data availability