Study population and study design
A community-based cross-sectional study was conducted in Gondar city. Gondar has located 727 Km from the capital, Addis Ababa, and 182 km from Bahir Dar, the capital of Amhara National Regional State. According to the 2007 national central statistical agency, Gondar city has a population size of 351,675 and the city is subdivided into 24 city kebeles (local administrative) holding approximately 53,725 households and 112, 800 adult population. There is one governmental tertiary eye care center that provides different specialty eye care services and three private eye care specialty clinics in the city.
The study was conducted from April 25 to May 15, 2017. Adults’ age ≥18-years-old living in Gondar city, were the source population. Individuals who were unable to respond to the questionnaire due to either serious illness or mental problem and those who resided less than 6 months in the study area were excluded.
Sample size determination and sampling procedure
The sample size was determined by using open Epi computer software for a single population. By considering the proportion of good attitude towards traditional eye medicine use (P= 50%), a design effect of 1.5 and 10 % of non-response rate, the final sample size was 633.
To select study participants, a multi-stage random sampling technique was employed in 24 homogenous kebeles (local administrative). Six kebeles were selected from the city by a simple random sampling technique. In selected kebeles, there were 12, 952 households. Population proportion to size sampling was used to determine the sample size in each kebeles. Households in kebeles were selected by a systematic random sampling method using a sampling fraction of 21. Finally, the lottery method was used to select a participant in a household, where more than one adult (≥18 years old) was found (Figure 1).
Data collection tool and procedures
Data collection was done by 6 trained diploma secretaries through a face to face interview, using a pre-tested structured Amharic (local language) Questionnaire. The data collection instrument (questionnaire) was developed in English for this study purpose and then translated to Amharic (local language) and later translated back to English by language experts to ensure the accuracy and reliability of data. To assure data quality, a pre-test (5% of the sample size) was done outside the study area on a similar population. The reliability of the questionnaire was tested using Cronbach’s alpha and the result was 0.80. During a home to home visit to collect data, each data collector started by spinning a pen at a crossroads to select a specific direction. Besides, data clean up; checking for completeness, outliers, and missing values were done each day by the supervisor.
Operational definitions
Good attitude: A participant was said to have a good attitude towards traditional eye medicine use when he/she had scored mean and above the mean of attitude questions.
Poor attitude: A participant was said to have a poor attitude towards traditional eye medicine use when he/she had scored less than the mean of attitude questions.
Availability of traditional healer: A traditional healer was considered as available when a participant answered “yes” for the question “do you know a functional traditional healer?”
Health insurance: The participant was said to have health insurance when the participant has had an annual payment agreement with the government and the government covers the annual health cost of the participant.
Data processing and statistical analysis
Data from the entire questionnaire was coded, entered into Epi info 7, and exported to SPSS version 20 software for analysis. Frequency and cross-tabulations were used for the descriptive analysis of data. Associations between dependent and independent variables were analyzed using binary logistic regression with enter method variable selection. Variables with a P-value of ≤ 0.2 in the bivariable analysis were entered into multiple logistic regressions and a variable with a P-value of < 0.05 with a 95 % confidence interval was considered as a statistically significant association. Model fitness was checked using Hosmer and Lemeshow goodness of fit and the result was 0.985. Besides this, predictor variables were checked for multicollinearity and the variance inflation factor values were less than 10 through cross-checking across each variable.