Study design and setting
We conducted community based cross-sectional study design from March 21 to April 20, 2018. Data was collected by using both quantitative and qualitative methods intended to achieve stated objectives. This study conducted in Hossana town, Hadiya zone, southern Ethiopia. It is 230 km far from Addis Ababa, and 160 km far from regional town, Hawassa, Ethiopia. According to 2007, Hosanna town census projection, estimated total population was 105,371. Out of these, 51,632(49.47%) were male population and 53739 (50.53%) were female population. Out of 53739 (50.53%) total female population, 3,667(6.8%) female were reproductive age group and of 3,646(6.78%) were pregnant women. There were a total of 21504 households reside in the town. There were also one hospital, three public health centers and more than 10 private higher and medium clinics were provision of health care services for the communities reside in the urban.
Those eligible households were selected by using systematic random sampling techniques from Hosanna Town, Hadiya Zone, southern Ethiopia. The eligibility criteria were those female head of households or spouses with age greater than 18 years old and reside for more than one year in the study setting were included in the study. First select health extension worker (HEW) and then households under the health extension worker catchment areas were selected. Those head of households or spouses unable to listen, unable to talk, critically ill during data collection period were excluded from the study.
Sample size determinations and sampling procedures
Sample size was calculated using single population proportion formula for stated objective. By taking the previous proportion of 39% of utilization urban health extension program service, which is conducted in West Shoa Zone, Oromia Regional State, Ethiopia, 2014 , and using 95% confidence interval with alpha value at 5% by considering 10% of non-response rate then, sample size was estimated as follow:
N = Z2 *P (1–P)/d2
= (1.96)2 (0.39) (1- 0.39) = 366
= 366 (with 10% for non-response a total of 403 sample size was calculated. Within each selected household, the female head (the mother or the wife) of the household was interviewed. In cases when the mother or the wife was not available, the husband (or the male head of the household) was interviewed.
Sample size determination for qualitative study participants: A total of four focus group discussions with an average of 10 study participants and 10 individual interviews were conducted. Groups were homogeneous according to the main inclusion criterion. However, there were heterogeneity within each focus group in terms of age, residence occupation and education status. Each discussion lasted average of 30 to 90 min until completion of discussion. Data saturation was sufficiently met after 4 focus group discussions and 10 individual interviews. Purposeful sampling technique was used to select the study participants which were significant to provide information representative samples to achieve stated objective.
A systematic random sampling technique was used to select eligible households at every ten (10th) interval which was reside in selected kebelle in the town. All eight kebeles were included in the study. To achieve representativeness of this study, households were selected by using proportional allocation of size in each kebeles (the smallest administrative unit in Ethiopia). A list of frame of all the households in each kebeles was obtained from the kebeles administration office. If more than one eligible respondent were exist in the household’s data collectors select one respondent by using lottery method. .
Data collection tools and procedures
Study instruments were adapted and modified according to the context of the study area. Thus, they could include both the outcome variables and independent variables in the structured questionnaire. Quantitative data were collected by using a structured questionnaire. The questionnaire was adapted after a review of documents, guidelines, and manuals related to UHEPs, and various previous literatures conducted in urban areas. The questionnaires had four parts. The first part was including the socio-demographic and economic characteristics of study participants; the second part was involved questions related to the community knowledge of health extension program services utilization, third part was service related contact with UHE professional’s factors and the fourth part was about nature of health development army.
Qualitative data collection process includes: four focus group discussions and 10 in-depth interviews. Development of focus group discussion includes: one group formed from health extension program members eight in number, one group from kebele steering committee ten in number (8 males and 2 females), one group from kebele administrative members nine in number and other group formed from Health Development Army ten in number were participated in each sessions. Purposive sampling techniques were used to select study participants selected kebeles in Hosanna town. Participants in each group assuming that they had rich information regarding health extension program packages and provision of primary health care service at the community level.
Data quality assurance
The questionnaire adapted in English language and translated into working language (Amharic), then back-translated to English to check for consistency and further analysis. A pre-test was conducted in Fonko town using 5% of the study sample size and necessary adjustments were done based on pre-test findings.
Two day training was given data collectors and supervisors about the objective of the study and the process of how to collect data. Face to face interview was conducted between four trained and grade12 completed data collectors and study participants. Two Supervisors (Diploma nurses) were checking completeness and consistency of day to day collected data. In addition to this, all questionnaires were crosschecked completeness and consistency daily by the principal investigator.
For the qualitative data collection process interview guidelines prepared and used to guide the focus group discussion and one supervisor moderated the FGD. In addition to this, a tape recorder was used to record the discussions. The data was transcribed and then translated in to English for further interpretation. Similar responses was grouped and summarized based on thematic area or key variables. Results of the qualitative study were presented in narrative form triangulated with quantitative results.
Data processing and analysis
Data entered in to Epidata version 3.1 and for further analysis exported in to STATA 14. Frequency with percentage was reported to explain the amount of participants in categorical variables. Cross tabulation was performed to identify how much cell consists in each category of explanatory variable with outcome variable.
The existence of measure of association between each explanatory variable and the outcome variable identified using binary logistic regression model. In the binary logistic regression model those explanatory variables whose p-value < 0.25 were the candidate predictors for the multivariate binary logistic regression model . Strength of measure of association between independent variable with outcome variables reported using the Odds Ratio (OR) with 95% confidence interval. In the multivariate binary logistic regression model strength of measure of association was reported by using Adjusted Odds Ratio (AOR) with 95% CIs, by controlling the effect of other predictors. In multivariate binary logistic regression model, p-value < 0.05 was used identify significantly associated factors with outcome variable.
The research team was checked existence of confusing variable which means confusing association between explanatory with outcome variable . In addition to this, redundancy of independent variable explains the outcome variable was checked; due to its existence, concerned body made intervention on both of repeated variable produce wastage of resources, so, we removed one and made the appropriate intervention. Then this was evaluated by using Mean of Variation Inflation Factor (VIF) value <10, this indicate absence of rigorous collinearity among predictors .
During the analysis phase, model selection criteria performed by using log likely hood ratio test due to the reason of each model was nested with each other models. Furthermore, log likely hood ratio test was used to explain the maximum parameter estimates of parsimonious model. Moreover, reasonably fit model was selected by calculating the Hosmer and Lemeshow goodness-of-fit test which compares observed cell with expected cell tally . The result of this showed that, non-significance value confirmed the data reasonable well.