Menu
Menu
Menu
Menu
Objectives: Our aims were to examine AMR-specific and AMR-sensitive factors associated with antibiotic consumption in Nepal between 2006 and 2016, to explore health care-seeking patterns and the source of antibiotics. Methods: Cross-sectional data from children under five in households in Nepal were extracted from the 2006, 2011 and 2016 Demographic Health Surveys (DHS). Bivariable and multivariable analyses were carried out to assess the association of disease prevalence and antibiotic use with age, sex, ecological location, urban/rural location, wealth index, household size, maternal smoking, use of clean fuel, sanitation, nutritional status, access to health care and vaccinations. Results: Prevalence of fever, acute respiratory infection (ARI) and diarrhoea decreased between 2006 and 2016, whilst the proportion of children under five receiving antibiotics increased. Measles vaccination, basic vaccinations, nutritional status, sanitation and access to health care were associated with antibiotic use. Those in the highest wealth index use less antibiotics and antibiotic consumption in rural areas surpassed urban regions over time. Health seeking from the private sector has overtaken government facilities since 2006 with antibiotics mainly originating from pharmacies and private hospitals. Adherence to WHO-recommended antibiotics has fallen over time. Conclusions: With rising wealth, there has been a decline in disease prevalence but an increase in antibiotic use and more access to unregulated sources. Understanding factors associated with antibiotic use will help to inform interventions to reduce inappropriate antibiotic use whilst ensuring access to those who need them.
Cross‐sectional data on living children under five in households in Nepal were extracted from the 2006, 2011 and 2016 DHS surveys through datasets and survey reports. The DHS survey collects data using a stratified 2‐stage cluster sampling method. The time frame for data collection was from February to August in 2006, February to June in 2011 and June to January in 2016. ARI was defined as fast breathing and/or difficulty breathing due to a problem in the chest with or without cough in the 2 weeks preceding the survey. This is concordant with the definition given by the WHO Integrated Management of Childhood Illnesses (IMCI) for pneumonia. Fever (parameters not specified in the survey) and diarrhoea (frequent loose or liquid stools) were defined as occurrence of the symptoms in the last 2 weeks. Dysentery was ascertained as the presence of diarrhoea with bloody stools. Occurrence of these symptoms was based on maternal/care‐giver recall. Care seeking was defined by whether the mother sought advice or treatment for the illness from any healthcare facility. Antibiotic treatment was assessed by asking the mother if the child had taken any drugs during the illness and if so, whether this consisted of antibiotic pills, syrups or injections. Rates of antibiotic use were calculated with the total under‐five population as the denominator to reflect antibiotic consumption at the population level. A full list of definitions can be found in Table S1. A descriptive analysis was carried out on the survey reports and datasets to identify changes from 2006 to 2016 with regard to demographics, disease prevalence and antibiotic use. A bivariable analysis was performed to evaluate both direct (AMR‐specific) risk factors (age, sex, wealth, location, maternal education, household size) and indirect (AMR‐sensitive) risk factors (maternal smoking, use of clean fuel, sanitation, water source, nutritional status, access to health care and vaccinations) for antibiotic use. Variables were chosen based on existing literature and expert opinion [10, 11]. Bivariable models were included to provide a comprehensive overview of all the risk factors being explored, including those that were not significant. The consistently significant factors in the bivariable analysis (age, wealth and location) were then included in the multivariable analysis using logistic regression. Factors not associated with the outcome in bivariable models were omitted from adjusted models to avoid data sparsity. Each survey year’s data set was modelled separately and results then compared across time periods. Sample weights provided by the DHS data were applied to account for over and undersampling of particular regions, and adjustments were made for clustering of data using Taylor‐linearised variance estimation. Health care‐seeking behaviours, the source of antibiotics and appropriate use in accordance with the WHO IMCI guidelines were also examined. Data management and analysis were carried out using Stata version SE 12. All data were publicly available and anonymised with ethical approval covered under the original data collection.
