This one-year two-arm parallel-design cluster randomized controlled community trial was conducted in the West Arsi Zone, Oromia regional state, Ethiopia, 250 km from Addis Ababa. Clusters (Kebeles), the smallest administrative unit in Ethiopia, were used as the randomization unit. The zone, spanning 12,556 sq. km, includes 16 districts (13 rural and 3 urban) with a population of 2,929,894 in mid-2022. Its climate comprises 45.5% highland, 39.6% midland, and 14.9% lowland, and agriculture is the primary livelihood. Conducted from October 15, 2022, to July 15, 2023, the study enrolled pregnant adolescents before 16 weeks of gestation, excluding those who declined consent.

We used G*Power 3.0.10 to calculate the sample size. The following assumptions were used to calculate the required sample size (28): a level of significance of 0.05, 80% power, an effect size of 0.37, and for the difference between two independent means (two groups), intracluster correlation (ICC) = 0.03 (based on a similar published study) (29). A design effect of approximately 2 was taken, DE = 1 + (m-1) ICC, where m = Average cluster size, assumed number of observations per cluster was 32; m = 32, ICC = 0.03, DE ≈ 2, and a 10% loss to follow-up was taken into account. The total sample size was 488. Therefore, 244 pregnant adolescents in the intervention group and 244 pregnant adolescents in the control group were included. There were 28 clusters. A cluster sampling technique was used in this study.

From 16 districts in the zone, five had nutrition education intervention; therefore, they were excluded from the study. From the 11 eligible districts, four districts, namely, the Dodola Rural District, Adaba District, Gedeb Hasasa District and Siraro District, were selected using simple random sampling (SRS). Then, samples of nonadjacent kebeles (clusters) were selected from the four districts using SRS. Based on the proportional size allocation and considering the equal allocation of clusters in the intervention and control groups, six clusters from both the Dodola Rural district and the Adaba district and eight clusters from both the Gedeb Hasasa district and the Siraro district were taken. The lottery (SRS) method was used to allocate the intervention and control clusters. We employed cluster randomization in this study. The Consolidated Standards of Reporting Trials (CONSORT) guidelines were used for reporting the results.

Pregnant teenagers in the same cluster had a high likelihood of communicating and discussing the intervention messages; therefore, a cluster randomized trial was utilized to prevent message contamination. All pregnant teenagers who met the criteria in one cluster were enrolled in the same arm (either the intervention or control arm) to prevent information contamination. Buffer zones (nonselected clusters) were also placed between the intervention and control clusters (30, 31).

Eligible pregnant adolescents were screened through a house-to-house survey by asking about the first date of their last menstrual period and confirming pregnancy via a pregnancy test. All eligible pregnant adolescents were included in the study. The cluster was randomly assigned, and study participants were screened and enrolled by nurses working in particular districts. Stratified randomization was used, clusters were randomly allocated to study arms within different districts.

Community-based nutritional behavioral change communication intervention (NBCC) using the HBM (health belief model) was the intervention package for this study. This study was adapted from the recommendations of the World Health Organization (WHO), Federal Ministry of Health of Ethiopia, the blended training module on nutritional counseling by EFDRE/MOH, and similar interventional studies (32–36). Moreover, the formative assessment at the beginning of the project guided the development of the intervention tools; husbands of the pregnant adolescents were included in the NBCC, and food preparation demonstrations were included in the intervention. The intervention tools included (1) a training guide for nutrition educators, (2) a manual for nutrition educators, (3) leaflets with the core messages for the pregnant adolescent and families, and (4) counseling checklist cards.

For 1 week, the intervention procedure was pilot tested in a context comparable to the research location, and changes were made in response to the pilot testing results. Counseling was given on the basis of the MDD-W (Minimum Dietary Diversity for Women), the benefits of achieving the MDD-W and the consequences of adequate intake. The consumption of diversified food during pregnancy enables the adequate intake of important micronutrients. Inadequate dietary intake during pregnancy is the major determinant factor in the risk of pregnancy in low-birth-weight infants and is a cause of different unfavorable pregnancy outcomes. The main points of the counseling manual were increasing meal frequency (the number of times a participant take a meal during a day) and portion size as gestational age increased and eating a variety of foods, particularly those high in iron-rich foods, animal products, fruits, and vegetables. The counseling guide’s primary components also included advice on the use of iodized salt and iron/folic acid supplements. Reduced workload, day rest, the use of impregnated bed nets, and the use of medical services were additional messages of the primary contents.

The intervention was framed with the health belief model (HBM), which is a theoretical model that can be used as a guiding framework for health behavior interventions. The HBM theorizes that a person’s perceptions and beliefs about a health behavior can directly influence their probability of adopting it (37). The HBM suggests that individuals perform an internal assessment of the benefits of changing their behavior and decide whether to act. The model identifies different aspects of this assessment: (1) perceived susceptibility, (2) perceived severity, (3) perceived benefits of behavior change, (4) perceived barriers to taking action and (5) self-efficacy (38). During NBCC, it was also emphasized how poor nutritional intake might have negative effects and how susceptible pregnant adolescents are to undernutrition. The NBCC manual also covered the advantages of consuming enough meals that are varied and the obstacles to eating a balanced diet. Each pregnant adolescent in the intervention group attended four counseling sessions throughout her entire pregnancy. Home visits were used to provide personalized trimester-based NBCC. During counseling, the AFDs (counselors) employed a client-centered approach to pinpoint food habits and unique nutritional requirements. Counselors considered the needs of the pregnant adolescent, their household income, and any gaps they had found before letting the adolescent select advice that was readily available, palatable, and cheap in their area. A GALIDRAA step to counseling (Greet, Ask, Listen, Identify, Discuss, Repeat, Agree, and Appoint) was used (39).

Each counseling session for the NBCC lasted 45 to 60 min and was conducted using counseling guide cards that contained the essential information. Before 16 weeks of pregnancy, the first consultation focused on the fundamentals of nutrition, food groups, choosing foods that provide a balanced diet, demonstrating how to prepare meals, frequency of meals, and the use of iodized salt. The second and third counseling sessions, which covered the entirety of the counseling manual, were provided throughout the second trimester of pregnancy. The final counseling, which focuses on weight gain and incorporates all of the module’s important messages, was provided based on the gaps that were discovered during the previous trimesters of pregnancy.

Leaflets with the core messages in Afan Oromo and Amharic (local languages) and appropriate pictures were prepared and delivered to each pregnant adolescent in the intervention arm. For pregnant adolescents who could not read, anyone at home or in the neighborhood who can read was requested to help by reading to the adolescent and other family members.

Fourteen AFDs were used as counselors and were closely supervised by 4 BSc nurses; counselors were selected by health extension workers based on their previous experience in public health services. The counselor and supervisor had 1 week of intensive training that included role-playing and fieldwork utilizing the training manual. Additionally, three additional days of training were provided for the supervisors and counselors after the intervention had been in place for approximately 2 months to ensure that the providers continued to follow the standardized practices. Dietary practice was the primary outcome, and nutritional knowledge was the secondary outcome of this trial.

Cooking demonstration: Cooking demonstrations were offered, focusing on teaching pregnant adolescents how to prepare foods that are balanced, maintain nutrient retention, prevent nutrient loss and ensure food safety. Cooking demonstrations were performed two times on a group basis by experienced nurses at health posts. The diets in the communities participating in the study were predominantly plant-based and thus contained a high content of phytate and polyphenols that inhibit the absorption of iron, zinc and calcium. Thus, the pregnant adolescent demonstrated food processing that enables increased absorption of nutrients (40).

Based on the best practice recommendations developed by the National Institutes of Health Behavioral Change Consortium, criteria were constructed to evaluate the intervention’s integrity (9). Checklists evaluating the intervention design, counselor training, counseling process, reception of the intervention, and application of the skills learned from the intervention were included in the study (41). We chose nonadjacent clusters to avoid information contamination. To equalize the variations, the intervention and control groups each received an equal number of clusters from each district.

Prior to the start of the experiment, the intervention method was pretested. Additionally, to standardize the procedure, each pregnant adolescent in the intervention group had the same number and frequency of counseling sessions, and the durations of contact within an intervention group were comparable. Using a training manual, role-playing, and simulated counseling sessions, counselor training was provided. Pre-and posttraining examinations as well as a practical evaluation were used to gage the knowledge and abilities of the counselors. Randomly chosen counseling sessions were chosen for process evaluation, and one process evaluator examined all of the selected sessions. Employing a “yes/no” rating scale, the process observer evaluated the educator on factors including employing a counseling guide, providing the entire topic, length and frequency of counseling, preparation, accuracy, and the capacity to react to inquiries in a suitable manner (82). The pregnant adolescents’ knowledge of diet during pregnancy was evaluated through interviews through checklists about their understanding of the main points of the intervention. The checklist was applied to a demonstration of food preparation and consumption to evaluate the implementation of the intervention.

Due to the nature of the intervention, participant allocation concealment was not practical. The groups were labeled with a unique number until analysis was finished, which also served to blind the data input clerk. The lead investigator and counseling supervisors oversaw the counseling process.

The primary outcome of this study was dietary practice measured within 24 h. dietary recall method, while the secondary outcome was nutritional knowledge. We used a pretested structured questionnaire to collect the data. The questionnaire included questions on sociodemographic factors, health service usage, dietary categories, meal frequency, and Household Insecurity Access Scale (HFIAS). Two master holders in public health (MPH) holders and six clinical nurses worked as supervisors and data collectors, respectively. Three female laboratory technicians performed the pregnancy tests. At the participants’ homes, the data collectors did face-to-face interviews to fill the questionnaire. In order to preserve the teenager’s privacy as much as possible, other family members were not permitted entry to the site of the interviews.

Dietary intake was calculated using 24-h recalls in accordance with the Food and Nutrition Technical Assistance (FANTA) III recommendation from the Food and Agricultural Organization (FAO) and the United States Agency for International Development (USAID). The goal was to determine whether pregnant teenagers’ diets were diverse (42). According to the recommendations, there are 10 food groups, including grains, plantains, white roots and tuber pulses (peas, beans, and lentils); dairy; meat; nuts and seeds; eggs; dark green leafy vegetables; poultry and fish; other vegetables and other fruits and vegetables rich in vitamin A. A pregnant adolescent was deemed to consume a sufficiently diverse diet if she had consumed at least five of the foods on the lists in the 24 h prior to the data collection period (42, 43). Participants were asked to recall every meal they had consumed during the preceding 24 h, both inside and outside of the home. In addition, the participants were prompted to recall any between-meal snacks they may have had. Food items were scored “1″ for consumption over the reference period and “0″ for nonconsumption.

The HFIAS guidelines (44) were used to assess food security. Twenty-seven questions on the Household Food Insecurity Access Scale were used to evaluate the households’ level of food security. Prior to this, the questions were approved for use in underdeveloped nations (45). Food-secure households that experienced fewer than the first two food insecurity indicators, households that experienced between 2 and 10, 11 to 17, and more than 17 food insecurity indicators were considered mildly, moderately, and severely food insecure, respectively.

Principal component analysis (PCA) was used to generate a wealth index by considering fixed assets. Thirty-five variables were used, and some of the variables included the availability of a water source, type of living house, presence of a latrine, bank account, agricultural ownership, livestock, and household assets (46–50). The responses of all no dummy variables were dichotomized and coded as “1” for the household possessing the asset and “0” for the rest, based on the WHO/UNICEF definition. Household income was classified into quintiles (46).

The assumptions of PCA were checked, and the correlation matrix for the variables containing two or more correlations was ≥ 0.30; the Bartlett test of sphericity was used (p value < 0.05); variables with sampling adequacy measures less than 0.50 were removed (looking anti-image); and the overall measures of sampling adequacy were Kaiser–Meyer–Olkin (KMO) ≥ 0.5 and (47) communality > 0.5, not having the complex structure of correlation ≥ 0.40. Components that collectively explained more than 60% of the variance in the set of variables and had eigenvalues ≥ 1 (48) were used to identify variables to be included in further analyses. The economic status of the study subjects was categorized as poorest, poor, medium, rich, or richest (46).

Eight questions were used to measure the autonomy of pregnant adolescents. When a choice was made by the pregnant adolescent alone or jointly with her husband, code 1 was assigned for each question; otherwise, code 0 was assigned. The ability of pregnant adolescents to make decisions was categorized using the mean (46). Maternal knowledge of food intake during pregnancy was evaluated using 12 questions. When a response was correct, a code of 1 was issued; otherwise, a code of 0 was given. The data were collected on HBM constructs with 25 questions. Code 1 was provided for each question when participants responded yes, and code 0 was given if they responded no; the means were compared between the intervention and control groups (46).

MDD-W: MDD-W or adequate dietary diversity was achieved when the pregnant adolescent consumed food from five or more food groups on the day or within 24 h before data collection (42).

Inadequate dietary diversity: Pregnant adolescents consumed food from fewer than five food groups on the day/within 24 h/before data collection (42).

Nutrition knowledge: Pregnant adolescents who scored ≥75% on the knowledge questions were considered to have good knowledge; otherwise, they were considered to have a low knowledge score (46).

A test–retest method was used for reliability analysis of the pregnant adolescent nutritional knowledge variable, and questions were reviewed to assess whether the results were the same for all respondents. A Cronbach coefficient of 0.83 was obtained after the pretest. To ensure validity, nutrition experts from Dilla University, the Department of Nutrition, were involved in checking the questionnaire to ensure that the questions elicited the required response, and a pretest was also conducted.

The Kobo toolbox was used to enter data, which were subsequently exported to SPSS version 25 software for analysis. A chi-square test was used to evaluate the differences in the sociodemographic parameters between the two groups. The McNemar test was used to analyze differences between pre-and postintervention dietary practices and nutritional awareness.

The difference in outcome measures (Good dietary practice and Good nutritional knowledge) between the intervention and control groups was examined using the generalized estimating equation (GEE) with a binary logit function. First, we performed correlations on all the structures, and the correlation matrix structure with the least amount of Quasi-Information was the exchangeable correlation structure; therefore, this structure was used. The clustering variables in the GEE models were participants. The impacts of various confounding factors were taken into account when fitting the model. The analysis included sociodemographic factor information, obstetric trait information, food security information, time, treatment, and time and treatment interactions. Time and treatment interactions were used to evaluate the intervention’s impact. A multivariable GEE binary logistic model was carried out to account for within-cluster correlation of measurements. Odds ratios and the associated 95% confidence intervals were calculated, and statistical significance was defined as a p value of 0.05 according to the multivariable GEE binary logistic model. A per-protocol analysis was performed in this study. The per-protocol analysis included all study participants who adhered to the predetermined guidelines. Therefore, in this study, pregnant adolescents who attended four education sessions and provided end-line data were included. Difference-in-differences (DID) was used; the DID analysis controls for both unobservable and observable time-invariant characteristics with time-invariant effects. In the correlation matrix analysis dietary practice was measured by number of food groups consumed and Nutrition knowledge was measured by percentage of correct answers from 12 items on nutrition knowledge questions.

Patients or the general public were not actively involved in the design of this study; however, the study’s objectives were disclosed to and approved by the study participants, Jimma University, and the zonal and district health offices. Additionally, we intended to disseminate the findings to the West Arsi zonal and relevant district health offices.

Ethical approval was obtained from the Jimma University IRB/ethics committees with reference numbers JUIH/IRB/194/22 and the Oromia Regional Health Office. Each study participant received a thorough description of the study’s title, goal, protocol, and duration, as well as the potential risks and benefits, prior to providing informed consent. Each teenager provided verbal, written, and sign informed consent prior to any interview or measurement. Informed consent was obtained from the LAR (legally authorized representative) for study participants aged 18 years and younger. Participants were made aware of the publication of their anonymous comments. Informed consent was obtained from participants prior to the commencement of interviews. The researcher remained truthful to the academic and ethical requirements. Finally, the researcher kept the data in a locked file cabinet in a safe place after the completion of the study. Informed consent was obtained from both the adolescent and their husband or parents. Finally, any ethical issues that arose during this research were resolved through discussion between the researcher and JU’s IRB.

Among the 459 pregnant adolescents who were enrolled in this study, 426 [intervention group (IG) =207 and control group (CG) = 219] fully adhered to the protocol and were included in the analysis. At baseline, there was no significant difference in socio-demographic characteristics between the intervention and control groups (p > 0.05; Table 1).

At end-line, Pregnant adolescents in the intervention group had significantly higher scores in perceived susceptibility (p < 0.001), severity (p = 0.004), and benefits (p < 0.001) than the control group. There were no significant baseline differences between groups (p > 0.05). Some constructs in the control group also showed significant changes over time, which may be due to external influences such as routine healthcare messages or community discussions (Table 2).

Before the implementation of the trial, there was no statistically significant difference in nutritional knowledge or dietary practice between the two groups (Table 3). After the trial, pregnant adolescents in the intervention group showed significant improvements in dietary practices and the frequency of meals compared with their dietary practices before the intervention. At the end of this trial, the proportion of pregnant adolescents who had appropriate dietary practices increased by 20.3% (from 21.7 to 42%) in the intervention group. However, the number of pregnant adolescents who had appropriate dietary practices decreased by 5.6% (from 21.5 to 15.9%) in the control group. There was a statistically significant difference in the mean minimum dietary diversity score (MDD-W) and mean meal frequency between the intervention and comparison groups (p < 0.001). There were statistically significant differences in the mean nutritional knowledge score, MDD-W score and mean meal frequency (Table 4).

On the multivariable generalized estimating equations model, after adjusting for probable confounders, the odds of appropriate dietary practices increased more in the intervention group than in the comparison group [AOR = 4.2, 95% CI: 2.6–5.3]. Pregnant adolescents in the intervention group had an AOR which was 5.5 times that of the comparison group regarding nutritional knowledge (AOR = 5.5, 95% CI = 3.78–8.1; Table 5). Differences between baseline and baseline dietary practices and differences in DID between the intervention and control groups are shown in Table 6. Furthermore, the correlations of the health belief model constructs with knowledge and dietary practices are shown in Table 7. There was no statistically significant difference in the mean score of nutritional attitudes, food security and autonomy of pregnant adolescents (Table 8).

The multivariable generalized estimating equation (GEE) logistic regression model revealed that having a college education or above (AOR = 4.37, 95% CI = 2.90, 5.62), being a merchant (AOR = 6.37, 95% CI = 5.20, 7.18), having a richest wealth quintile (AOR = 3.34, 95% CI = 2.27, 4.47) and good nutritional knowledge (AOR =5.47, 95% CI = 4.2, 6.29) were significantly more likely to increase dietary diversity (Table 9).