A 10+10+30 radio campaign is associated with increased infant vaccination and decreased morbidity in Jimma Zone, Ethiopia: A prospective, quasi-experimental trial

Mass media interventions have the potential to reach large audiences and influence health behaviours and outcomes. To date, no study has evaluated the effect of a radio-only campaign on infant vaccination coverage, timeliness, and related morbidity in a low-income country. We implemented the “10+10+30” radio campaign involving broadcasting a weekly 10-minute radio drama series on vaccination, followed by a 10-minute discussion by community health workers, and then a 30-minute listener phone-in segment in Jimma Zone, Ethiopia for three months. To assess the impact of 10+10+30, which was aired on a community radio station, we recruited mothers of infants up to 5 weeks old in intervention district clusters that were inside the radio station’s reception range (n = 328 dyads) and control district clusters that were outside of the range (n = 332 dyads). Intention-to-treat and per-protocol analyses, adjusted for pre-intervention differences between the districts, were conducted to examine the co-primary outcome of Penta-3 vaccination coverage and timeliness as well as those of other vaccines and outcomes related to infant morbidity. Both intention-to-treat and per-protocol analyses revealed higher vaccine coverage (p<0.001) and more timely vaccine administration (p<0.001) in the intervention district relative to the control district, with infants in the intervention district being 39% more likely to receive a Penta 3 vaccination (adjusted RR: 1.39, p<0.001). In addition, adjusted regression analyses of maternal retrospective reports over a two-week period revealed 80% less infant diarrhoea (RR: 0.20, p<0.001), 40% less fever (RR: 0.60, p<0.001) and 58% less cough (RR: 0.42, p<0.001) in the intervention district relative to the control district. This study provides compelling initial evidence that a radio drama integrated with discussion and phone-in components may improve infant vaccination coverage and timeliness, and may reduce infant morbidity. Randomized controlled trials are needed to confirm and extend these findings with other samples.


Introduction
According to the study, after controlling for several cofounders, respondents who listened to radio at least once a week were 18% more likely to have completed childhood vaccinations compared with those who did not listen to radio, but the corresponding figure for those who watched television at least once a week compared with those who did not watch television at all was only 1%. A second reason that radio-based health information interventions need to be studied further in LMICs is that radio broadcasting is relatively cheap and less limited by literacy as programming can be produced in local languages [10]. Thus, particularly for rural areas which are more likely to include illiterate and low-income individuals, community radio has the potential to be a more tailored and cost-effective method of transmitting health information.
Despite these potential advantages of radio as a health communication medium, to our knowledge, there have been no quasi-experimental trials to test the effect of multipronged radio intervention on childhood vaccination coverage, timeliness, and child health outcomes in a low-income country.
Moreover, the existing literature on the association between behavioural change communication and the coverage of child immunization has given little attention to the impact of community engagement on behavioural change [5]. Thus, there is a need for studies that highlight how community engagement could influence behavioural change-specifically childhood vaccination.
Community engagement interventions tend to involve the priority population in their design and implementation [11]. Thus, radio programmes that require listeners to phone-in have a stronger community engagement appeal than formats such as radio spots that merely disseminate information.
Community engagement has been found to increase child vaccine uptake in LMICs [12]. In Ethiopia, community engagement interventions on childhood vaccination have included community health workers in their delivery [13][14][15].
A common complaint of behavioural change interventions is the lack of theory that informs their development and implementation [16]. While community health workers have been engaging with parents of infants to aid uptake of childhood vaccinations, and phone-in and drama programmes exist for promoting child health in general, theory-based interventions for promoting childhood vaccination that involve community health workers and radio as a platform for engagement is lacking.
To address this gap, we developed a theory-based "10+10+30" radio intervention for mothers of infants and other community members in Jimma Zone in the Oromia region, Ethiopia that included a 10-minute radio drama on infant vaccination, a 10-minute discussion by community health workers, and a 30-minute phone-in from listeners drama regarding childhood vaccination. The intervention was based on the Health Belief Model (HBM), which emphasizes key constructs that shape health behaviour including perceived benefit, perceived susceptibility, perceived barriers, and cues to action [17,18]. The HBM suggests that mothers or caregivers who believe that their infants are at risk of vaccine-preventable diseases, and who also believe that their infants would benefit from vaccinations are more likely to adopt the behaviour of vaccinating their infants. The HBM has been commonly applied in studies that assess parental beliefs and attitudes to childhood vaccination [17].
Our goal was to address perceived vaccination barriers and promote infant vaccination coverage and timely administration. The intervention was aired on Jimma Community radio from October 23, 2020 through January 10, 2021. We selected Ethiopia because, despite efforts to achieve full immunization coverage of children, the proportion of fully vaccinated Ethiopian children aged 12-23 months increased from 24% in 2011 to only 39% in 2016 [19]. Factors such as health service use and access to maternal and child health information were found to predict full immunization coverage [20], thus making a need for communicating with mothers about immunizations particularly critical in this population.
Moreover, the present study was designed to address gaps in communication identified by the Ethiopia Extended Program for Immunization (EPI), including lack of integration with health extension workers and recommended "appropriate strategies of communication, such as Community Conversation, to improve demand for EPI service by the beneficiaries" [20]. Accordingly, this study evaluated the effectiveness of a participatory radio intervention in improving coverage and timeliness of infant vaccination, and reducing morbidity related to vaccine-preventable diseases.

Ethics statement and trial registration
This study was approved by Ethics Review Board of Jimma University in Ethiopia (IHRPGD/ 668/2019) and the Institutional Review Board of Texas A&M University (IRB2018-0483D). Women gave oral informed consent to participate in the study. Verbal consent, rather than written consent, was sought because the local research team indicated that it is culturally not appropriate to obtain signatures from most rural households, and that enforcing written informed consent would result in mothers' refusal to participate in the study. Additional information regarding the ethical, cultural, and scientific considerations specific to inclusivity in global research is included in the (S2 File).
In the earlier stages, as this was a quasi-experimental trial involving the use of radio for educating community members on childhood vaccination, the requirement to register it as a clinical trial was not apparent. Thus, despite the endline data collection occurring on January 27-31, 2021, registration of the trial was completed on June 4, 2021. Nevertheless, this late trial registration did not affect the study conduct or results. This trial was registered with Clinical-Trial.gov as NCT04913714.

Intervention development
This prospective, quasi-experimental trial was conducted in two districts (Manna or intervention district and Chora Botor or control district) of Jimma Zone in the Oromia region of Ethiopia. Our intervention, known as "10+10+30", involved 10-minute radio drama on infant vaccination, a 10-minute discussion by community health workers, and a 30-minute phone-in from listeners. The 10+10+30 radio programme on infant vaccination was a radio only multifaceted intervention that aimed to improve vaccination coverage and timeliness.
The 10+10+30's theory of change incorporates two main inter-related causal pathways: a) availability of a local, radio drama serial, with each episode lasting 10 minutes; and b) trained health extension workers (HEWs) to aid a 10-minute discussion of each drama episode on air, and to answer questions from listeners for 30 minutes, thereby improving engagement with priority populations to promote uptake of childhood vaccinations. A trained radio journalist acts as a facilitator for the discussion of the drama, takes questions from listeners, and lets HEWs answer them. In Ethiopia, HEWs work primarily in rural areas, and are the equivalent of community health workers in other countries [21]. A low-level health facility called a health post that serves a population of about 3000 to 5000 in a village (kebele) is assigned two HEWs [22].
The use of education-entertainment contributes to behaviour change and increased uptake of preventive services in sub-Saharan Africa [23]. The '10+10+30' intervention is built upon this principle. It anticipates that if culturally appropriate serial radio drama episodes on vaccination are available, HEWs receive necessary media training to both discuss the drama and to answer questions from listeners on air, HEWs will reach more mothers or caregivers of infants. In turn, this will encourage mothers/caregivers to vaccinate their infants in a timely manner and promote overall coverage of childhood vaccinations.
The 12 episodes covered content on infant vaccines, vaccine-preventable diseases, and the roles of parents, community leaders, health care professionals and HEWs in promoting timely infant vaccination.
Prior to airing, three of the episodes were pilot tested among 15 mothers (five for each episode) to assess, among other things, their perspectives on cultural appropriateness and duration. During pilot-testing, mothers generally found the drama episodes interesting and educational, and they noted that the 10-minute duration was adequate.
We officially aired "10+10+30" on Jimma Community radio from October 23, 2020 through January 10, 2021. The episodes aired each week on Fridays and Sundays at 2:00 pm. The days and times were chosen based on feedback from the radio drama design workshop. Fridays and Sundays are typically days of rest, and thus husbands or fathers are more likely to be home. Each episode was aired twice; the first time on Fridays, and then again on Sundays as planned. Although the same drama was repeated on both days, the 10-minute discussion and the 30-minute phone-in were unique to each airing. Each 10-minute studio discussion involved two HEWs and a facilitator. Of the ten trained HEWs, two were selected each week to act as radio panellists to discuss the drama, and to answer questions from listeners. Each episode was preceded by a one-minute song with a slogan of "Now more than ever for children". See S1 File for detailed information on intervention development and implementation.

Study design
To assess the impact of 10+10+30, two districts of Jimma Zone in the Oromia region of Ethiopia were selected as the intervention and control districts. The study districts were strategically selected to ensure that the control (Chora-Botor district) did not have Jimma Community Radio, which was used for the implementation of 10+10+30 radio programme. In other words, Jimma Community Radio's broadcasting did not reach Chora-Botor district due to the long distance between the district where the radio station was located and the control district [See  The episodes were aired on Friday and Sunday at 2:00 pm each week. The days and times were chosen based on the outcome of a formative assessment and the design workshop, which showed that Fridays and Sundays were typical days of rest, and thus husbands-who often owned household radio or mobile phones-were more likely to be at home. The radio drama design workshop also showed that from 2-4 pm, people came together during coffee ceremony, a platform that could be used to listen to the radio [24].
On Jimma Community Radio, sports and entertainment programmes were aired prior to 2 pm. There were no alternative programmes on national radio or other channels which were broadcasting information on child immunization or childcare practices in the two districts. However, there was no information on how often women listen to radio compared to attending community meetings, childcare sessions at health facilities or watching TV in the intervention and control districts.
According to Jimma Zone Health Office, the estimated population of Mana district in 2019 was 202,048 and that of Chora-Botor was 179,077. The number of children under one year old in Mana was estimated to be 6499, with that of Chora-Botor being 5760.
Mothers of infants up to five weeks of age were the main focus of the radio intervention which covered childhood vaccination.
During planning for the intervention, the team agreed to not make access to radio a requirement for recruitment into the study for two reasons. First, the planning phase identified the potential of those listening to the radio programme spreading information to their neighbours. Second, restricting enrolment to those with access to a radio would limit the number of respondents and fail to account for the potential of the project to benefit those who lack access.
A validated questionnaire aimed at collecting information on childhood vaccination and demographic characteristics of households was administered in both districts at baseline (up to 3 days pre-intervention) and 2-3 weeks after the 12 weeks of radio programme were aired (post-intervention). Trained research assistants asked the mothers to complete surveys focused on their child's health, without mentioning the radio campaign. The data collectors were not aware of which participants were in the control or intervention groups to reduce bias.
Both the baseline and post-intervention questionnaires asked respondents whether they had listened to any radio programme in the last 3 months on vaccination that had drama, discussion, and listener phone-in. However, only the post-intervention questionnaire had questions on child health outcomes, including whether in the last 2 weeks their infant had experienced a) diarrhoea, b) fever, c) an illness with a cough, or d) fast, short, rapid breaths or difficulty breathing.
Eligible participants for the survey were mothers (�18years) who had an infant up to 5 weeks old, lived in the control or intervention districts, and consented orally to be interviewed. Trained data collectors visited households in intervention and control clusters (lowest administrative units or neighbourhoods) to identify all eligible mothers/caregivers (a census) at baseline (October 13-20, 2020) and again at post-intervention (January 27-31, 2021). A census of all eligible mother-infant dyads in the intervention district were visited in their households but two of the kebeles or clusters in the control district were not visited upon reaching the required sample size. The intervention was designed to begin before each child reached 6 weeks old, the recommended age for the first Pentavalent vaccine dose. This vaccine, often referred to as Penta, is a combination of diphtheria, tetanus, whooping cough, hepatitis B and Haemophilus influenzae type B vaccines. Penta 1 vaccine is a predictor of subsequent doses in the series (Penta 2 and Penta 3) [25]. Moreover, collecting post-intervention data after the 12-week intervention ensured that the child cohort would be at least 14 weeks old, the recommended age for receiving Penta 3 dose.

Outcomes
The primary outcome was Penta 3 vaccine coverage, which was measured at most 3 weeks after the end of the intervention. Timeliness of the vaccinations was a secondary outcome. Timeliness of vaccination was only assessed among infants with a vaccination date recorded on a health card, and was defined as either early (more than 4 days before the recommended age for vaccination), on-time (from 4 days before the recommended age to 4 weeks after the recommended aged for vaccination), or delayed (more than 4 weeks after the recommended) [26]. The coverage and timeliness of Penta 1, Penta 2, Rotavirus (Rota) 1, Rota 2, Oral Poliovirus (OPV) 1, OPV 2, OPV 3, pneumococcal conjugate vaccine (PCV) 1, PCV 2, and PCV 3 vaccinations were included as secondary outcomes. Any vaccination, either reported on the health card or recalled by the mother, was included as a vaccination. Morbidity indicators of diarrhoea, fever, difficulty breathing, and cough at any time in the prior 2 weeks as reported by the mother were also included as secondary outcomes.

Statistical analysis
The study was powered for the primary outcome: Penta 3 vaccine coverage.
Based on the Penta 3 coverage of 64.0% from both mothers' recall in a survey and use of vaccination cards previously identified in 2017 in Oromia region in Ethiopia [27], we hypothesized that the intervention would lead to a 15% increase in Penta 3 coverage by age 4-5 months. The 15% increase in Penta 3 coverage expected to result from the radio intervention was determined to be significant in Jimma Zone by key health experts during the radio drama design workshop. Our sample size calculations led to a minimum of 337 mother and infant pairs (up to 5 weeks old) per arm (or a total of 674 mother-infant dyads for the two districts), 48 administrative units or clusters per arm (or 96 in total), with at least 7 pairs in each cluster using cluster-randomized sampling.
Our calculations included the following parameters: 1) 80% power, 2) a two-sided alpha of 0.05, 3) an estimated intra-cluster correlation of 0.167, a design effect of 2 relative to simple random sampling, and 10% loss to follow up [28,29]. An intra-cluster correlation of 0.167 is recommended by the WHO to be a conservative choice for post-campaign vaccine coverage survey if there is no strong reason for selecting another value [28].
Based on population figures in the two districts, rather than random sampling within clusters, we conducted an entire census of the clusters to achieve minimum sample size in the intervention and comparison districts. Also, each district had 22 clusters (hence a total of 44 clusters for the two arms). However, because two of the clusters in the control district were excluded after achieving the sample size for the study, there were 20 clusters in the control district and 22 clusters in the intervention district (Fig 2).
Two analyses were completed: an intention-to-treat as the primary analysis and a per-protocol as the secondary analysis. In the intention-to-treat analysis, all mothers in the intervention district were considered exposed to the radio program and all mothers in the control district were considered not exposed to the radio program. In the per-protocol analysis, mothers in the intervention district who reported listening to any radio programme on vaccination that had drama, discussion and listener phone-in during the past three months on the post-intervention survey were considered exposed to the radio programme. Mothers in the control district who did not report listening to any radio programme on vaccination that had drama, discussion and listener phone-in during the past three months on the post-intervention survey were considered not exposed to the radio program. All other mother-infant dyads were eliminated from the per-protocol analysis.
We compared characteristics of the two groups for factors that have been shown to be associated with increased rates of vaccination from the literature in low-and middle-income countries [30], and Ethiopia in particular [31][32][33]. These factors included sex of child (male/ female), birth order (first, second or third, and fourth or later), religion (Islam, Orthodox, other), marital status (yes/no), location (urban/rural), walking time to vaccination centre (<15 minutes, 15-30 minutes, >30 minutes), place of delivery (home/health centre), frequency of antenatal care (0 visits, 1-3 visits, >3 visits), vaccination verification card (yes/no), age of child at baseline (in days), age of mother at baseline (in years), wealth index (number of yes responses to 18 questions about household commodities, livestock, and land), and frequency of radio listenership. These factors were chosen a-priori [34]. A sensitivity analysis was completed where only children with vaccinations with dates that were recorded on the health card were considered to be vaccinated. All others, including those where the health card indicated a vaccination, but no date was recorded and those where the mother recalled the vaccination, were considered to not be vaccinated for the purpose of the sensitivity analysis.
Categorical variables were compared using a χ2 test and continuous variables compared using a t-test. A generalized linear model with a log link and Poison distribution with robust variances was used to compare the risk of the various outcomes between the intervention and control arms. We expect that our outcomes are likely to be correlated within kebeles (lowest administrative units) or neighbourhoods. Therefore, the lowest administrative units were included in the regression models as a random intercept to account for this clustering (correlation) within neighbourhoods. Regression models were both adjusted and unadjusted for the confounders described above. Anyone with missing data or a response of "don't know" was eliminated from the regression that included the variable of interest. The same analysis was repeated for both the intention to treat analysis, the per-protocol analysis, and the sensitivity analysis.
Timeliness of vaccination was only examined in cases where the date of vaccination was recorded on the health card. Two-sided p-values less than 0.05 were considered statistically significant. All analysis was completed using Stata version 15 [35].

Results
Overall, 696 women were assessed for eligibility, and 332 women-infant dyads in the control district and 328 women-infant dyads from the intervention district were enrolled at baseline (Fig 2). Although we estimated 7 infants per cluster, in practice the number of infants per cluster varied, in some cases being nearly three times as large.
At post-intervention, data were completed for 324 (97.6%) and 314 (95.7%) women-infant dyads across 20 lowest administrative units in the control and 22 lowest administrative in the intervention district, respectively. Of these, 302/324 (93.2%) of the control dyads and 108/314 (34.4%) of the intervention dyads were included in the per-protocol analysis based on selfreported exposure to the intervention. The control and intervention groups had some significant differences across demographic and confounding factors (Table 1). Notably, the intervention group had higher educational attainment, was more likely to live in an urban area, was closer to the vaccination site, was more likely to have delivered in a health facility, and had greater numbers of antenatal care visits. Similar trends were observed in the per protocol analysis, although the groups tended to be more balanced across the demographic and confounding factors (S1 Table).

Childhood vaccine coverage
As shown in Table 2, according to the intent-to-treat analysis Penta 3 vaccine coverage was significantly higher in the intervention group (89.8%, 95% CI: 85.9-92.9%) compared with the control group (65.1%, 95% CI: 59.7-70.3%). Similar results were obtained in the per-protocol analysis, with the intervention group showing coverage of 97.2% (95% CI: 92.1-99.4%) versus 62.9% (95% CI: 57.2-68.4%). Indeed, vaccination rates were significantly higher across all of the childhood vaccinations examined in both the intention-to-treat and per-protocol analyses. In a sensitivity analysis, where only children with vaccination dates recorded on a health card were considered, the proportion vaccinated was lower, but similar trends were observed (S2 Table).
As shown in Table 3, in the intention-to-treat regression analysis, infants were 35% more likely to have received Penta 3 vaccination in the intervention group than in the control group (adjusted RR: 1.35, 95% CI: 1.21-1.50). A similar pattern of significant effects was observed Table 1. Baseline characteristics of the participants in the intention to treat analysis.

MEAN (SD) MEAN (SD) T-TEST P-VALUE
Age of the Child (days) 16 In the sensitivity analysis that were restricted to children with vaccination dates recorded on a health card, the likelihood of completing Penta 3 vaccination was 1.96 times higher in the intervention compared to the control (adjusted RR: 1.96, 95% CI: 1.57-2.45) (S3 Table). The likelihood of completing Penta 3 vaccination was higher in the sensitivity analysis as compared to the main analysis, but the risks of other vaccinations were generally lower in the sensitivity analysis. The intervention was associated with significantly higher likelihood of completing all vaccinations examined in both the intention to treat and per-protocol cohorts in the sensitivity analysis (S3 Table).

Childhood vaccine timeliness
As shown in Table 4, the intention-to-treat analyses revealed that the intervention was also associated with enhanced Penta 3 vaccination timeliness. Specifically, among infants with recorded vaccination dates, the proportion of infants receiving their Penta 3 vaccine on time was 60.1% in the intervention group (95% CI: 52.6-67.2%) versus 47.8% (95% CI: 37.7-58.1%) in the control group. Most of the other infant vaccines, with the exception of OPV3 which was not statistically significant, were also had better timeliness in the intervention group as compared to the control group. Finally, Table 4 also shows that similar results were obtained for the per-protocol analyses.

Infant morbidity
As shown in Table 5, infants in the intervention group had fewer reported cases of illness in the previous two weeks across all symptoms examined according to both the intention-to-treat and per-protocol cohorts.
Compared with infants in the control group, intention to treat analysis after adjusting for sex of child, birth order, religion, marital status, urban/rural, time to vaccination centre, place

Discussion
The present study provides promising evidence that a "10+10+30" radio intervention is associated with increased infant vaccine coverage and timeliness, and may help reduce morbidity. b Adjusted for sex of child, birth order, religion, marital status, urban/rural, time to vaccination centre, place of delivery, antenatal care, able to see age/vaccination verification card, age of child in days at baseline, age of mother in years, radio listening, wealth index, and accounting for clustering within kebeles. Unadjusted models still accounted for clustering within kebeles.  The positive effect on vaccine coverage reflects similar results obtained for mass media campaigns in low-and middle-income countries such as Bangladesh, Ecuador, Peru, Philippines and the Democratic Republic of Congo [5]. However, unlike previous studies that used combinations of mass media, and thus making the effect of the radio component alone difficult to establish, the current used only radio, making it possible to quantify potential of radio only intervention in improving childhood vaccination coverage. For example, Hutchinson and colleagues assessed the effect of 26 episodes of a television serial drama with radio spots, newspapers, billboards, community-level activity promotions in rural Bangladesh from Aug 2001 to March 2002 [36]. Results indicated that DPT3 vaccine coverage among those exposed to the interventions was 64% versus 48% among those not exposed. In the current study, intentionto-treat analysis indicated that the coverage of Penta 3 vaccine (which contains two more antigens than a third dose of DPT3 vaccine) was 89.8% among participants in a region exposed to the intervention versus 65.1% among participants in a control region without access to the intervention. Combined with similar positive findings for the per-protocol analysis, the current findings provide initial empirical support for the 10+10+30 radio intervention as a method of significantly increasing infant vaccine coverage.
Timely completion of childhood vaccination is also a key goal in averting vaccine-preventable diseases. Penta 3 vaccine is often a primary focus of childhood vaccination because its completion indicates coverage of the first three series of the childhood vaccinations at ages 6 weeks, 10 weeks and 14 weeks. In the current study, intention-to-treat analysis found that ontime completion of Penta 3 vaccine was 60.1% among participants in the intervention district and 47.8% in the control district, with per-protocol analysis having coverages of 66% and 47%, respectively. To our knowledge, this is the first experimental trial to assess timeliness of Penta 3 vaccines with a radio-only intervention in a low-income country. Our results suggest that listeners exposed to the programming about the benefits of vaccinating their children and the dangers of not doing so were encouraged to have their infants vaccinated in a timely manner. This finding is consistent with results from a recent study of the effect of mass media exposure on childhood immunization coverage in low-and middle-income countries using pooled cross-sectional data from the Demographic Health Surveys of 13 countries in sub-Saharan Africa for the years 2004-2010 [4].
Systematic reviews of interventions for improving infant vaccination had not identified a single study that focused only on radio [3,5], making our study highly innovative in its design and outcomes. In a separate study in Pakistan that did not use radio but had similar design and outcomes like those of the current study, exposure to pictorial cards to increase the knowledge of mothers about vaccination was significantly associated with improved infant vaccination compared with the control [37]. In that study, researchers followed mother-infant dyads from less than 6 weeks of age until 4 months, with 72.1% in the intervention group completing all 3 doses of DPT/Hepatitis B vaccine as compared with only 51.7% in the control group [37]. The current study's focus on the relationship between exposure to radio-only intervention on vaccination and child morbidity outcomes is also novel, but shares attributes of a child health radio campaign called Saturation+ which was implemented in Burkina Faso. Although Saturation+ did not focus on vaccination, it did focus on similar child morbidity outcomes including fever, diarrhoea, and fast or difficult breathing within two weeks prior to data collection [6]. Relative to the present study, Saturation+ was a more resource intense approach that included airing 1-minute spots at least 10 times per day and a 2-hour programme airing 5 nights per week that involved a 10-minute live radio drama acted on location by local actors and remainder of the programme taken up by news, music, and discussion [6,38]. Although mothers in the Saturation+ intervention group reported fewer episodes of fever (58.2% versus 73.5%), diarrhoea (54.1% versus 64.4%), and fast/difficult breathing (48.1% versus 54.1%) in their children relative to control mothers, none of the differences was statistically significant [6]. Thus, it is particularly noteworthy that the present '10+10+30' intervention, which was less resource intensive, may be associated with reduced child morbidity. However, the cost-effectiveness of 10+10+30 has not yet been determined unlike Saturation+ [39].
The '10+10+30' intervention campaign seems to have reached a third of the primary target population given that 108 of the 314 mothers (34.4%) in the intervention district reported having listened to a radio programme with drama, discussion and phone-in segments on childhood vaccination. However, 22 of the 324 mothers (6.8%) in the control district also reported having listened to the '10+10+30' intervention even though the control district did not have such a programme. This may suggest possible confusion with other radio programmes, a tendency to provide socially-desirable answers, or perhaps a lack of understanding of the question. Regardless of the source, the per-protocol analyses excluded those in the control district who reported listening to the programme and included only those in the intervention district who reported that they listened the programme. Hence, the continuing presence of significant effects when this potential source of variability is accounted for provides further support for the benefit among those who actually received the intervention.

Limitations
Although the '10+10+30' radio intervention appears to have promise as a strategy to enhance infant vaccination coverage, timeliness and related morbidity, our findings should be interpreted with caution because of some limitations. First and foremost, despite carefully selecting the control and intervention districts, there were important differences in some key sociodemographic factors that are known to affect childhood vaccination. Although we statistically controlled for these differences in our analyses, statistical control does not guarantee the generalizability of the findings to other communities. Second, due to the low number of infants in both control and intervention districts, a census of the clusters and households were used rather than randomization. Given that two of the kebeles or clusters in the control district were not visited upon reaching the required sample size, selection bias could have affected the results of the study. Third, there were relatively few people with missing data, and thus such observations were eliminated. Using such an approach for handling missing data may have implications for the power of the study [40]. Fourth, the lack of information on how often women in the control and intervention districts listen to radio compared to attending community meetings, childcare sessions at health facilities or watching TV may also lead to confounding errors in the two groups.
Fifth, we relied on parental recall of childhood vaccination status in instances when there was no vaccination card available. While this is a common procedure in some childhood vaccination studies, it nonetheless presents a challenge related to recall bias [30]. Unfortunately, requiring a vaccination card as confirmatory evidence may not be feasible in some populations, and particularly in rural Ethiopia [41]. For example, the Ethiopian Demographic Health Survey of 2016 found that only 25.9% of children aged 12-23 months in the Oromia region had vaccination cards during home visit [42]. Lastly, our reliance of maternal recall of infant illness symptoms during a preceding the two-week period may also be subject to reporting errors; however, efforts to gauge morbidity using other metrics (e.g., medical visits) would introduce other potential errors such as omission of milder symptoms and individual differences in access to, or threshold for seeking, medical care.

Conclusions
In summary, the current study provides initial evidence that the '10+10+30' radio intervention is associated with increased infant vaccination coverage, timeliness and reduced morbidity. Whereas caution is needed in interpreting the present findings, most importantly because of demographic differences between the intervention and control group, this study nonetheless contributes to the literature by providing the first evidence that a radio-only intervention (10 +10+30) may increase infant vaccinations in a low-income country. Additional randomised controlled trials are needed to verify and extend the present results and to examine population-level changes.