The Effect of Behavior Change Communication via the Health Development Army on Pregnant Women’s Nutrition and Health Practices in Ambo District, Ethiopia: A Cluster Randomized Controlled Community Trial

Research Hypothesis

H0: Behavior change communication through the health development army has no effect on nutrition and health practices of pregnant women. H1: Behavior change communication through the health development army has a positive effect on nutrition and health practices of pregnant women.

Study Design, Study Period and Setting

A cluster randomized controlled community trial (CRCCT) with baseline and endline measurement was conducted from June 2018 to October 2018 among pregnant women in the Ambo district of West Shoa Zone, Ethiopia. Ambo district is located in the western part of Ethiopia, 114 km from Addis Ababa, the capital of Ethiopia. Based on 2017 district health office data, it has 37,454 and 6976 reproductive age group and pregnant women, respectively [24].

Sample Size Determination and Study Population

The sample size was calculated using G power 3.1.9.2 program with a power of 80% for Fisher’s exact test and precision of 5%. According to Fekadu B, et al. [25] the proportion of optimal dietary practice among pregnant women (p1) 0.34 was used, effect size (h) of 0.3 and with the allocation ratio case to control (N2/N1) of 1, Proportion (p2) 0.65 was calculated, and with the allocation ratio of intervention to control group (N2/N1) of 1. The calculated sample size was multiplied by design effect of 1.5 due to cluster sampling. Considering a 10% loss to follow up, the final sample size was 770 pregnant women (385 of controls and 385 of interventions group). Due to problems reported in Figure 1, the actual data were collected from 372 women in the intervention group and 372 women in the control group. The study population was pregnant women aged 18–49 years, before 16 weeks of gestation, and permanent residents (who lived in the study area for more than six months) of the Ambo district. Pregnant women with chronic diseases such as hypertension, diabetes, tuberculosis, and HIV/AIDS, those with incomplete baseline data, and those enrolled in supplementary feeding programs were excluded from the study, as these interventions would have an impact on nutritional status and thus bias the results of the study.

Recruitment of Participants

A multi-stage sampling technique was used to select study participants. The total number of kebeles (sub districts) in the district was stratified into rural and urban areas. Simple random sampling (SRS) with the lottery method was used to select 12 kebeles (2 urban and 10 rural) from the existing 39 kebeles (6 urban and 33 rural). Each kebele has its own sub kebeles or clusters (locally known as “Gots”), which have been pre-determined by government bodies and are grouped into 149 distinct clusters. The number of clusters in each kebele was then recorded. Forty-eight clusters were selected at random from the entire number of clusters available in the district, including non-adjacent clusters and those that were somewhat far apart from one another. Lists were established to include these clusters in the study. As a result, each arm of this study had 24 clusters. We left at least one cluster between each cluster that was used in our research as a buffer zone. We used a proportionate stratified sampling technique to obtain clusters from each kebele. Using the block randomization technique, we randomly assign the chosen clusters into intervention and control groups, as described in more detail below. Eligible pregnant women were screened utilizing a family folder prepared by the kebele’s health extension workers (HEWs), as well as by inquiring about the first date of their last menstrual cycle and confirming pregnancy using a pregnancy test. Then, were employed simple random sampling technique to select study participants from each cluster for each arm. Equal numbers of participants were selected from each cluster. Health extension workers and data collectors visited a group of selected women at each arm at their residence. Women who agreed to take part in the study were asked to sign an informed consent form to confirm their voluntary involvement. Each subject was interviewed after receiving consent.

Random Selection of Clusters

The clusters (locally known as “Gots”) were the units of randomization in our study. We stratified the subdistricts (or kebeles, i.e., urban vs. rural kebeles) before randomization and produced a separate list of kebeles in alphabetical order. The stratification technique was used to distribute any known and unknown confounders at the kebele level equally among the study arms. A unique code was provided to each kebele. As was already mentioned, thirty-nine sub-districts make up the Ambo district. Using simple random sampling (SRS) with the lottery method, we selected 12 kebeles. Each kebele has sub kebeles (or clusters, often known as “Gots”). We used the proportionate stratified sampling method to obtain clusters from each of the 12 kebeles proportionately (48/149) to their population size. A unique cluster code was given to each cluster. The 48 clusters were then divided into two blocks of size 4 in accordance with their alphabetical order by the author. The author randomly selected the randomization sequence of clusters for each block using the sealed lots of the six possible permutations within the block. Each block’s clusters were randomly assigned to the intervention and control arms in the order indicated by the block’s chosen permutation for the stratum. In order to preserve a 1:1 random allocation ratio, we created 24 clusters for the intervention arm and another 24 clusters for the control arm. Actual and potential confounding factors were distributed equally across the study arms due to the stratification and randomization approaches, which ultimately ensured the comparability of the arms. Pregnant women who met the inclusion criteria were enrolled in the same arm in one cluster (either in the intervention or control arms). There were about 12–17 pregnant women in each cluster. Between the intervention and control clusters, buffer zones (non-selected clusters) were also left to reduce information contamination. Health Extension Workers recruited, enrolled, and assigned individuals to interventions after the author randomized the clusters.

Thirteen people from the intervention and thirteen from the control groups were removed from the study for reasons such as discontinuing the intervention and being lost to follow-up. The study was completed with 744 people (Figure 1).

Intervention

To continue with the intervention strategy, the researchers enlisted the help of health extension workers to recruit Health Development Armies (HDA). HDA were chosen from clusters that had been designated as intervention groups. Following recruitment, HDA were trained for one week using a protocol developed by the principal investigator for pregnant women based on the Essential Nutrition Action (ENA) framework, a framework for promoting maternal nutrition developed by the Manoff Group for developing countries, and making a balanced plate for pregnant women in Bangladesh [3, 4, 26]. The ENA framework was recommended by the World Health Organization (WHO) for developing countries and promotes a “nutrition through the life cycle” approach, addressing seven packages, but the authors’ focus was on one package, which was on maternal nutrition during pregnancy. A framework for promoting maternal nutrition developed by the Manoff group for developing countries reflect the benchmarks related to diet and micronutrients, disease prevention and treatment, and lifestyle factors against which maternal nutrition programs can be developed and monitored. Making a balanced plate for pregnant women in Bangladesh was a study protocol which follows practical demonstration (that is) exhibiting the meal preparation in front of the pregnant women. For this, participant women will be encouraged to share food from their own kitchen. Sharing food between participants is possible and culturally acceptable, so Authors adapted this Bangladesh study protocol approach to its study area and participants; it was further explained in the intervention protocol (Supplemental file S1). So, our intervention includes. Both theoretical and practical demonstrations were included in the training, HDA followed the criteria (based on the protocol) to assure the intervention’s adoption. A training handbook, role-playing, participatory meal preparation, and mock BCC sessions were used to deliver HDA training to a group.

To reduce variability among HDA, the researchers established a key assessment question (checklist) (incorporating both theory and skill). The health development army’s knowledge and skills were assessed before and after training, as well as through practical evaluations. After the training, the authors conducted standardization tests to ensure that everyone was on a similar level. Furthermore, the researchers and supervisors actively monitored HDAs during the pretest and during the intervention to see if they were delivering the intended message according to the intervention protocol. Those who failed to communicate the health and nutrition BCC message were identified and prompt corrective action was taken.

The intervention for pregnant women took place in the community’s intervention clusters. However, the exact location of the BCC training was determined by agreement between the pregnant women and the HDA (and so that they have traveled to a common site, but not out of the intervention clusters).The intervention was conducted once every two weeks for 1:00 up to 1:30 hour length on nonworking days; similarly, the researchers monitored the intervention activities every two weeks, and the intervention lasted three months. The typical cluster size was 12-17 pregnant women; therefore one HDA was in charge of each cluster.

The main components of the intervention were adapted from the recommendations of ENA, a framework for promoting maternal nutrition developed by the Manoff group for developing countries, and making a balanced plate for pregnant women in Bangladesh as described above. The core content of the intervention includes the following nutrition and health practice messages: on eating additional foods during pregnancy (increase meal frequency and portion size with gestational age); eating a variety (diversified) of foods from vegetable, fruit, and animal sources; utilization of iodized salt; appropriate timing and storage of iodized salt; decrease consumption of iron-inhibiting foods, such as tea and/or coffee, with meals; decrease alcohol consumption; not avoiding important foods during pregnancy (associated with taboos); importance of taking iron/folate during pregnancy for at least three months; sleep under an insecticide-treated bed net; seek treatment from health institution if developed illness; keep a clean environment, wash hands with soap during critical moments; keep food and food containers clean; drink only treated water (if out of pipe water); reducing heavy workload; taking day rest; improving self-decision making in food and her own health, importance of obtaining support from the family/community during pregnancy and utilization of health care services (Antenatal care (ANC) follow up, plan to deliver at health institution, and plan to have Post Natal Care (PNC) follow up) and birth preparedness and complication readiness. Similarly, at each BCC session, the consequences of not using the above message were discussed. During each BCC session, participants’ knowledge and attitudes about good nutrition and health were also assessed. Then, depending on the detected gaps, a BCC letter was sent out.

A message about food intake centered on locally available, acceptable, and affordable foods. In addition to delivering BCC messages, HDAs were provided posters and brochures with relevant images to display to pregnant women. Pregnant women were given leaflets with key messages written in Afan Oromo and Amharic (local) languages with illustrations. If a woman couldn’t read, it was suggested that she have the leaflet read to her by someone at home or in the community who could. Because the selected HDA were educated (at least read and write), they received written teaching aid; they had a schedule and topics to address in each contact (Supplemental file S1). Each pregnant woman in the intervention group had attended six BCC sessions on nutrition and health.

Meal preparation, exhibiting samples of iron/folate tablets, type and time of adding iodized salt while cooking, how to use an insecticide-treated bed net, and personal and food cleanliness were all covered in a practical demonstration. Participants were urged to share food from their own kitchens (home) to demonstrate meal preparation. Based on this visual display, pregnant women actively identified the food groups and preparation methods they should change.

Pregnant women in the control groups did not receive the intervention but were instead exposed to the typical care provided by the healthcare system during an ANC visit and any intervention at the community level by health extension workers. Women in the control group received nutrition education during their ANC visits, though it may not have been as thorough as other care like diagnose and treat diseases or complications. This service was available to pregnant women in both the control and intervention groups. They were monitored for the same duration of the intervention and given the same evaluations as the intervention group.

From April 1 to June 1, 2018, baseline data on first- trimester pregnant women were collected. All pregnant women who were classified as intervention groups began receiving the intervention on July 1, 2018. The intervention was given over three months, from July 01, 2018 to September 30, 2018.

Blinding

Allocation concealment is not possible due to the nature of the intervention. The participants, as well as the HDA, are aware of the intervention. However, the participants, HDAs, and data collectors were blinded to the study hypothesis. The field supervisors were blinded to the outcome of interest and, finally, we blinded the data collectors for the intervention. Furthermore, the data entry clerk was blinded by coding the groups.

Data Collection Tools and Procedures

Data were collected using a pre-tested and semi- structured questionnaire written in English. To maintain consistency, the questionnaire was translated into two languages (Afan Oromo and Amharic) and then back to English by language experts to keep its consistency. The questionnaire was pretested in Ginchi town, which is nearby to Ambo district, on 39(5%) of the total sample size to identify any ambiguity, length, completeness, consistency, and acceptability of the questionnaire, and some skip patterns were corrected before the real data collection.

Eight diploma nurses were recruited to collect data. Training was given to the data collectors on the objective and relevance of the study; confidentiality of information; respondent’s rights; informed consent; and techniques of interview. The filled questionnaires were checked for consistency and completeness daily by four supervisors who had BSc degrees in Nursing and principal investigators on the spot. Parts one and two of the questionnaire cover socio- demographic and maternal characteristics; part three covers maternal characteristics; part four covers nutritional status; and part five covers birth outcomes such as live or dead, place of delivery, and birth weight. At the baseline, data on sociodemographic, economic, and maternal characteristics were collected. Before and after the trial, data on nutrition and health knowledge, attitude, and practice were obtained, as well as mid-upper arm circumferences.

The household wealth index was assessed using wealth constructs covering household assets, utilities, and agricultural land ownership adopted from the Ethiopian Demographic and Health Survey [23]. The latent factors describing the wealth data were generated using principle components analysis (PCA) and then grouped into wealth terials.

Women’s decision-making power was assessed using eight questions. When a choice was made by the woman alone or jointly with her husband, code one was assigned to each question; otherwise, code zero was assigned to each question. The mean was used to categorize a woman’s ability to make decisions [23].

Food security status was assessed using 27 previously validated questions [27]. A household that experienced less than the first 2, 2–10, 11–17, and > 17 food insecurity indicators was considered food secure, mildly, moderately, and severely food insecure households, respectively.

Questions related to Knowledge, Attitude and practices was adapted from essential nutrition action frame work and the formative research done by the Manoff group to promote maternal nutrition in developing countries [3, 4, 15].

Knowledge of pregnant women about nutrition and health practices was assessed by using 14 questions that were used to assess knowledge of pregnant women regarding optimal nutrition and health. A knowledge score was calculated for each participant based on the number of questions that were correctly answered in the knowledge assessment questions. Each correct response was scored 1, and each incorrect response scored 0. The factor scores were summed and ranked into terciles (three parts). A pregnant woman was considered to have poor knowledge about nutrition and health if she scored below the highest tertile (i.e., in the first and/or second tertile) and good knowledge about nutrition and health if she scored in the highest tertile (third tertile) [28].

Attitude towards nutrition and health practice was assessed by 20 Likert scale questions using PCA. The factor scores were summed and ranked into terciles (three parts). Then the highest tercile was labeled as a favorable attitude, if not unfavorable attitude [28].

The rapid test kit (RTK) was used to determine if the salt used was iodized or not. The interviewer asks each pregnant woman to provide a teaspoon of salt used for food preparation and fill a small cup spread flat, then add two drops of test solution to the surface of the salt by piercing the white ampule, compare the color of the salt to a standard color chart, and determine the iodine concentration. Finally, salt with > 15 parts per million (ppm) was classified as adequately iodized salt, while salt with 15 ppm was classified as inadequately iodized [29, 30].

The outcome variable nutrition and health practice was measured using questions of practice about the dietary quantity, dietary quality, micronutrient intake, disease prevention and treatment and supportive life style and care. The respondents had asked to choose Yes=1 or No=0 answers by indicating whether each practice of nutrition and health were practiced or not. One point was allocated to a correct response for each questions and zero for incorrect response. The factor scores were summed and ranked into terciles (three parts). Pregnant woman was considered to have suboptimal practice about nutrition and health if she scored below the highest tertile (i.e. in the first and/or second tertile) and optimal practice about nutrition and health if she scored in the highest tertile (third tertile) [28].

Data Processing and Analysis

Thirteen pregnant women discontinued the intervention from the intervention group and thirteen lost to follow-up from the control group were excluded from the analysis. Before entering data, data were manually checked for completeness and consistency.Then it was entered into EPI data version 3.1 and exported to SPSS for Windows version 21 for cleaning and analysis. The effect of intervention was measured at the endpoint of follow up.

First, descriptive statistics like mean and Standard Deviation was done for continuous variable and frequency and percentage for categorical data. A chi-square test was performed to compare the baseline characteristics of the intervention and control groups.

The Hosmer-Lemeshow goodness-of-fit statistic is used to assess model goodness of fit. Multicollinearity was checked using Variance Inflation Factors (VIF) at < 8 but the VIF of all variables was less than 2, which means no multicollinearity. Moreover, the interaction of the variables with the intervention at a p-value of < 0.05 was assessed and there was no interaction. Simple logistic regression analysis was performed between nutritional and health practices and associated factors one at a time. Their odds ratios (OR) with 95% confidence intervals (CI) and p-values were obtained. Factors that were significantly associated with nutritional and health practice at a p-value <0.25 in simple logistic regression analysis were entered into multivariable logistic regression analysis. p- Values <0.05 were used to declare statistical significance. Statistical analyses were conducted on an intention-to-treat (ITT) basis, which is appropriate for a cluster randomized design study data analysis.

Data Quality Control

For one week, the health development armies were trained together. Data collectors and supervisors received three days of training as well. Four supervisors with BSc degrees in nursing and investigators supervised the HDA every two weeks. Furthermore, pregnant women in each cluster had the same number and frequency of BCC sessions, and the lengths of contact within each intervention group were similar, resulting in a standardized process. Participants were reminded of the necessity of attending all sessions and acting at home according to the protocol presented to prevent dropout and promote adherence to the intervention and follow-up program. The HDA monitored and recorded adherence to the BCC sessions in a personal training diary (attendance sheet). Pre-testing of the questionnaire was done among 5% (39 pregnant women). Every two weeks, HDAs and their supervisors met to discuss and provide possible remedies to the problems that arose during BCC sessions. The data were collected using a pre-tested questionnaire. Cronbatch’s alpha value of the questionnaire was checked and it was > 0.7 indicating that it is suitable for use in the research domain. The questionnaire was also translated into the most widely spoken languages in the study area (Afan Oromo and Amharic) to aid respondents’ comprehension. The data gathering process was thoroughly monitored by supervisors and the principal investigator. Daily, completed questionnaires were reviewed for completeness, and any missing or incorrect information was updated. Field supervisors randomly evaluated 5% of the data and alerted them to a possible measurement issue [26].

Ethical Consideration

All methods in the study were performed in accordance with relevant institutional, national, and international guidelines. This study is a community trial that was registered with the Pan African Clinical Trial Registry. The date of registration was March 5, 2018. The registry’s unique identification number is (PACTR201805003366358). The results were reported using the CONSORT (Consolidated Standards of Reporting Trials) guidelines (Supplemental file S2). The Jimma University Ethical Review Committee reviewed and approved the study protocol (ref no: RPGC/40724/2016). Permission to conduct the study in the respective kebeles was granted by the Ambo district health office (ref no: ADHO/134/2018). Prior to participation in the study, the nature of the study was clearly disclosed to the study participants in order to gain their written informed consent (fingerprint for those who could not read and write) was secured from the study participants, and all information obtained was kept anonymous. Soft copy data is password protected, while hard copy data is secured with a key and lock to guarantee confidentiality. Personally identifiable information will not be used in any form in the presentation of the findings.

Quantity of Food

There was no significant difference at base line measurement between the control and intervention groups in the quantity of food-related issues during this pregnancy (p>0.05). However, there was a significant difference in the consumption of additional food during pregnancy (44.1% versus 68.5%, p<0.001) and the absence of food sharing (8.3% versus 27.2%, p<0.0001) at the end of the study as compared to the intervention and control groups, respectively (Table 3).

Quality of Food

There was no significant difference at base line measurement between the control and intervention groups related to the quality of food (p>0.05). However, at the endpoint measurement, a significant increase was seen in the case of consumption of a variety of foods (15.1% VS 29.0%, p<0.001); utilization of iodized salt (18.3% VS 50.8%, p<0.001), appropriate utilization (52.9% VS 86.2%, p<0.001) and appropriate storage of iodized salt (58.8% VS 81.4%, p<0.001). Similarly, there was a significant decrease in avoidance of important foods (because of taboos) during this pregnancy (10.0% VS 26.6%, p<0.001) at endpoint measurements between the control and intervention groups, respectively (Table 3).

Iron/Folate Supplements

There was no significant difference at base line measurement between the control and intervention groups related to micronutrient intake (p>0.05). At the endpoint measurements, a significant difference was observed in iron/folate tablet supplementation at the endpoint measurements (59.9% VS 81.5%, p<0.001) between the control and intervention groups, respectively. However, in this study, there was no significant difference between the control and intervention groups for the duration of iron/ folate supplementation (30.9% VS 34.0%, p=0.622) (Table 3).

Disease Prevention and Treatment Related Practices

There was no significant difference at baseline measurement between the control and intervention groups related to disease prevention and treatment practices (p>0.05). A significant difference was observed between the control and intervention groups in the number of pregnant women who started ANC visits (62.44% versus 80.1%, p<0.001), number of pregnant women who followed ANC at least 4 times (15.1% versus 31.9%, p<0.001) at the end point measurements. However, no significant difference was seen in the number of pregnant women who went to health institutions for help after developing illness (especially fever and parasitic related diseases) (40.8% VS 39.1%, p = 0.719) and those pregnant women who drank only treated water during this pregnancy (26.6 % VS 32.8%, P = 0.065) (Table 3).

Supportive Life Style and Care Related Practices

At baseline measurement, there was no significant difference observed between the control and intervention groups related to supportive lifestyle and care practices (p>0.05). Similarly, at the endpoint of the study, pregnant women made no significant differences in food and health decisions (70.2% versus 68.8%, p = 0.691) or support from family or community (89.2% versus 91.9%, p = 0.209). However, a significant difference was seen between the control and intervention groups in getting enough rest during this pregnancy (73.4% versus 80.9%, p = 0.014) (Table 2). Generally, this study revealed that there was a significant difference between the control and intervention groups in the overall optimal nutrition and health practices at the end point of the study (39.0% VS 62.6%, p<0.001) (Table 3).

Risk Factors for Suboptimal Nutrition and Health Practices among Pregnant Women in the Ambo District at the End of the Study

The control group had a larger absolute risk of suboptimal nutrition and health practice (0.613 vs. 0.374, p<0.001) than the intervention group. Pregnant women in the intervention group had a relative risk and an absolute risk difference of 0.613 and 23.9% for suboptimal nutrition and health practice, respectively.

Simple log-binomial model analysis showed that there was an association between nutritional and heath practice of pregnant women and study group, residence, age of the respondent, respondent occupation, husband occupation, respondent educational status, husband educational status, household size, food security status, number of pregnancy, number of ANC, estimated time to reach health institution, knowledge about nutrition and health, and attitude towards nutrition and health. Whereas, religion, age at marriage, marital status, wealth status, number of previous birth, and gap duration between pregnancy had no association with nutrition and health practice of pregnant women. However, study group, respondent educational status, husband educational status, estimated time to reach health institution, knowledge about nutrition and health, and attitude towards nutrition and health were found significantly associated with nutrition and health practice in multivariable log-binomial model analysis (p<0.05) (Table 3).

The study group (being in the intervention group) was found to have a significant association with the nutrition and health practices of pregnant women. Pregnant women who received intervention were 30.9% less likely to be at risk of being suboptimal in nutrition and health practice compared to pregnant women who were in the control group (ARR = 0.691, 95%CI: 0.594-0.804). Those pregnant women who had secondary and above education were 66.5% less likely to be at risk of being suboptimal in nutrition and health practice compared with those with no formal education (ARR =0.335, 95%CI: 0.139-0.808). Similarly, husband education is significantly associated with optimal nutrition and health practices. Those pregnant women whose husbands had secondary and higher education were 55.0% less likely to be at risk of being suboptimal in nutrition and health practice compared with those with no formal education (ARR =0.450, 95%CI: 0.280-0.722). Pregnant women who travel more than 60 minutes to reach a health institution were 1.258 times more likely to be at risk of being suboptimal in nutrition and health practice compared to pregnant women who travel less than 30 minutes (ARR = 1.258, 95% CI: 0.829-1.499). Pregnant women who had good knowledge of nutrition and health were 31.9% less likely to be at risk of being suboptimal in nutrition and health practice compared to pregnant women who were in their counterparts (ARR=0.681, 95%CI: 0.675- 0.921). The study also revealed that those pregnant women who had a favorable attitude towards nutrition and health were 27.3% less likely to be at risk of being suboptimal in nutrition and health practice compared to pregnant women who were in their counterparts (ARR = 0.727, 95%CI: 0.541- 0.978) (Table 4).

Table 4: Result of the log-binomial model to identify risk factors for suboptimal nutrition and health practices among Pregnant women at endpoint in Ambo district, Ethiopia, 2018, (n=744). Notes: *Significant at p-value <0.05. Parameter estimates were adjusted for the tabulated variables. N: Number, Abbreviations: CRR, crude relative risk; ARR, adjusted relative risk, CI: Confidence Interval.

Strength

The strengths of this study are the use of the randomized controlled design that included a fairly large sample size. Because the intervention was delivered through health development armies who come from the same community, it improved the chance of sustainability.

Limitation

This study admits the following limitations: due to the nature of the intervention, allocation concealment was not achievable; nonetheless, the participants, HDAs and data collectors were blinded to the study hypothesis.

Some nutrition and health-related practices were self- reported and unable to be cross-checked during the data collection period.