Study design and area
A community-based cross-sectional survey was conducted among women in West Gojjam Zone, Amhara National Regional State, Ethiopia, over a period of three months (March to May, 2012) to assess the utilization of HEP and factors associated with it. West Gojjam Zone was one of the 11 zones of Amhara Region. The HEP was started in 2003, and during the study period, there were 6,530 HEWs (6,401 in rural and 129 in urban areas) of the region, and 782 HEWs (772 in rural and 10 in urban areas) of West Gojjam Zone.
Study population and sampling
Women who were alive, aged 15–49 years, and had at least one child less than 5 years old were included in the study. A multistage sampling procedure was used to select the study units (mothers in the selected households). At the first stage, six districts (Yilmana Densa, Mecha, Semien Achefer, Bure, Jabi Tehnan and Dega Damot) were randomly selected out of the 13 rural districts. At the second stage, 44 kebeles were selected randomly from the six districts. In the third stage, 30 households were selected randomly from each kebele to get 1320 mothers required for the study. The list of women who were alive, aged 15–49 years, and had at least one child less than 5 years of age in each kebele was prepared in collaboration with the HEWs. Whenever there was more than one mother in a household, only one mother who was responsible for the household was selected randomly by lottery method.
The sample size was determined by using a single population proportion formula. The computation was based on the 95% confidence interval (Zα/2 = 1.96), 5% marginal error (d), and 50% utilization of HEP (p) by the community, and a 10% non-response rate.
The sample, 423, was multiplied by the design effect of 3 (number of stages), and the final sample size was 1269, which was again, raised to 1320 in order to take 30 households from each kebele.
Data collection
In order to collect information, interview was conducted based on a structured questionnaire consisting mainly of close-ended questions. The questionnaire was developed based on previous health service utilization studies carried out in developing countries [11, 18, 21]. The questionnaire was based on three factors, such as predisposing factors, enabling factors, and needs factors.
Twelve data collectors and six supervisors were recruited and trained to administer the questionnaire. The data collectors were diploma holder nurses and health officers (with a bachelorette degree) served as field supervisors. The interview was conducted at study participants’ houses, and the data collectors had to go house to house in each selected kebele until a sample of 30 mothers was obtained for each kebele.
The structured questionnaire was pre-tested in kebeles of the same administrative zone which were not included in the actual study. The pre-test was done on 66 mothers (5% of the study sample size), and the questionnaire was assessed for its completeness, clarity, and length. A daily data quality-check was done during the study period, and data were double-entered to minimize error during data processing.
Operational definitions
Utilization of HEP was defined as health post visits by mothers in the community in the last 12 months for health services, such as immunization, family planning, antenatal care (ANC), delivery, postnatal care (PNC), and diagnostic treatment.
Predisposing factors were any demographic (sex, age, marital status), social structure (education, occupation, family size, ethnicity, religion), HEP related (hearing and understanding HEP), and belief related factors (perception about health services and health workers) explaining an individual’s decision to use health services. Enabling factors were situations that influence utilisation behaviour (income, home visits by HEWs, and frequency of home visits, services provided in the HP). Need factors were individuals’ perceived need to use heath care services based on their perceived illness and/or clinical evaluated illness (type and stage of illness).
Model households were defined as households that attended at least 75% of the training given by HEWs and implemented at least 75% of the HEP packages. Frequency of home visits by HEWs was the number of visits HEWs made in a week or month, and we categorized it into more frequent visits if there was at least one visit per month, and no visits or less frequent visits if there was no visit at all or if there was one visit in more than one month. The number of years after household graduation means years since the household has become a model family by attending at least 75% of the HEP training and implementing at least 75% of the HEP packages.
The conduct of HEWs and the quality of services were measured according to study participants’ perception of HEWs’ conduct and services provided in the HP, respectively, using a 5-points Likert scale ranging from 1 (very bad) to 5 (very good). Comprehension/understanding of the HEP was also measured according to study participants’ perception of their HEP knowledge using “Yes” or “No” responses and study participants who mentioned the exact number of HEP packages, i.e.16 packages, were labelled as ‘accurately mentioned the number of HEP packages’.
Data analysis
Data were coded and entered into Epi-Info 3.5.1 and transferred to SPSS 16 for analysis. Age was coded as “0” for ages below the mean, and “1” for ages above the mean. Mothers’ occupation was coded as “0” for housewife and “1” for farmer and other jobs. Other jobs meant jobs other than housewife and farmer, and included merchant and daily labourer. Education was coded as “0” for illiterate and “1” for at least read and write; income was coded into four categories (0 = Birr 562 and below, 1 = Birr 563–760, 2 = Birr 761–960, and 3 = Birr 961 and above).
The multilevel binary logistic regression was used to assess the predictors of HP visits by the community. The multilevel models allowed us to consider the individual level (household) and the group level (kebele) in the same analysis, rather than having to choose one or the other. Due to the multistage cluster sampling procedure, individual women were nested within kebeles; hence, the likelihood of women seeking HP visits was likely to correlate to the kebele members.
We examined the effect of the individual level variables, and the kebeles using a two-level binary logistic regression modelling. During analysis, the characteristics of women and households were taken as individual level (level-1), and kebeles were treated as level-2. For the dependent variable, HP visit, two models were estimated: the intercept-only model, an empty model that contained no covariates, and the full model that included individual variables and the kebeles. The intercept-only model allowed us to evaluate the extent of cluster variation on the utilization of HEP. Based on this model, the intra-class correlation coefficient (Rho) was calculated to evaluate whether the variation in the scores was primarily within or between clusters.
Both the crude odds ratio (COR) and the adjusted odds ratio (AOR) were used for reporting the results of the binary logistic regression. COR was an unadjusted crude ratio that reported the odds ratio without taking confounders into considerations, while AOR reported the odds ratio taking into account and controlling confounders. Therefore, AOR gave a more accurate picture of the association than COR did. Estimates of population parameters were presented with 95% Confidence Interval (CI). In every application of inferential statistics, P value of 0.05 was taken as significance level.
Ethical issues
The University of Gondar Ethics Review Committee approved the research proposal. A written, informed consent was obtained from each study participant. Confidentiality was assured by not taking personal identifiers. The respondents were also informed about their freedom to withdraw at any time while they were being interviewed.