Study setting and recruitment
Ghana has a population of about 30 million, of which 54.8% is urban [23]. Life expectancy for women and men was 68.4 years and 62.6 years, respectively in 2017 [23]. Study participants were recruited from two regions; one predominantly urban and representing the southern province (Greater Accra region (GAR)), and the other predominantly rural and representing the northern province (Upper West region (UWR)). The GAR houses the nation’s political capital, Accra, and records a population of 4,010,054 [30]. The UWR occupies the north-western corner of the country, with a population of 702,110 [30]. We employed a cross-sectional study design which included 12 communities in the capital towns of the GAR and UWR, i.e., Accra (8 sites) and Wa (4 sites), respectively. A total of 400 surveys were completed, 200 per region, with each interview lasting about 20-30 minutes. A questionnaire survey was administered in selected locations during September-October 2018 to gather data on individual knowledge, attitudes, and practices concerning antibiotics and antibiotic resistance. Eligibility was restricted to adults aged 18 years and older who had used antibiotics in the past and were able to speak, read, or understand English or any of the predominantly spoken languages in the regions, i.e., Ga, Akan, Dagaare, or Waale [30]. In-person surveys were administered at university campuses, shopping malls, market centers, and hospital surroundings. Individuals were purposely selected at different locations, days, and times of the week to obtain views from across diverse demographic groups. We also endeavored to obtain a balance in gender representation by setting at least 45% female participation. Participants were screened and recruited by trained research assistants (RA). Upon completion, participants received a soft drink and informational leaflets on antibiotic resistance.
Survey instrument and measures
We adapted the Antibiotic Resistance: Multi-Country Public Awareness Survey, developed and validated by the World Health Organization (WHO) [4]. The original survey instrument was enhanced by including questions such as the specific antibiotics participants had used during the last time of infection, their preferred method of receiving treatment, as well as preference for payment of treatment and medicines. The survey was first administered in twelve low-and high-income countries in 2015 and showed distinct results among various population groups [4]. To increase content validity, the survey was pilot-tested with fifteen volunteers to better understand participants’ comprehension. A few questions were subsequently revised prior to final dissemination. Pilot participants were excluded from the analysis. The 29-question survey targeted individual knowledge, attitudes, and practices, and composed of both nominal and ordinal closed-ended questions on participant demographics (8 items), access and use of antibiotics (10 items), as well as knowledge and perceptions about antibiotics and antibiotic resistance (11 items).
Dependent variables
Three outcomes were examined- Knowledge, Attitudes, and Practices. Knowledge was examined based on the number of correct responses to ten questions regarding different infections that could be treated with antibiotics, three questions relating to knowledge about prudent use of antibiotics, and eight questions about antibiotic resistance. For example, Question: “It’s ok to use antibiotics that were given to a friend or family member, as long as they were used to treat the same illness”; Question: “It’s ok to buy the same antibiotics, or request these from a doctor if you’re sick and they helped you get better when you had the same symptoms.” Response: True/False/I don’t know. The outcome ranged from 0 to 21, with a higher score indicating better knowledge. Attitudes was defined on a Likert scale and determined whether a person agreed strongly, agreed slightly, neither agreed nor disagreed, disagreed slightly, or disagreed strongly with a set of six statements regarding one’s perceptions about actions to reduce antibiotic resistance. An average score was determined, ranging from 1 to 5. The final dependent variable, Practices, was based on a total score for three items that assessed appropriate use of antibiotics. For example, Question: “On that occasion did you get the antibiotics (prescription) from a doctor?”; Question: “On that occasion did you get advice from a doctor, nurse, or pharmacist on how to take them?” Response: Yes/No/Cannot remember. We made a similar decision for Practices as for Knowledge, where any response other than the correct was considered incorrect. Antibiotic use (Practices) was defined appropriate only if one correctly answered all three questions, and inappropriate if otherwise.
Independent variables
We selected four main independent variables to include in our analysis; place of residence/region (Greater Accra, Upper West), gender (female, male), education (primary or lower, junior high school, senior high/vocational school, tertiary (university degree or currently enrolled)), and age (18-24; 25-34; 35-44; 45-54; 55-64; 65+). This was based on similar KAP studies which found significance for these variables [11, 14, 31].
Data analysis
We first examined several of the individual items for the Knowledge, Attitudes, and Practices outcome variables separately to understand the percent breakdown in our overall sample. Then we looked at the differences in responses to individual items by region, gender, age, and education using Chi-square and unadjusted regression analyses. We conducted three regression models to examine differences in Knowledge, Attitudes, and Practices scores by the demographic characteristics of interest (region, gender, age, and education), starting with unadjusted models, followed by multivariate analysis using ordinary least squared regression models for the Knowledge and Attitudes outcomes, and logistic regression for the Practices outcome to derive estimates net of other covariates. We retained the same independent variables in all models. Data were analyzed using Stata/SE 15.1.