Measuring antibiotic availability and use in 20 low- and middle-income countries

Abstract Objective To assess antibiotic availability and use in health facilities in low- and middle-income countries, using the service provision assessment and service availability and readiness assessment surveys. Methods We obtained data on antibiotic availability at 13 561 health facilities in 13 service provision assessment and 8 service availability and readiness assessment surveys. In 10 service provision assessment surveys, child consultations with health-care providers were observed, giving data on antibiotic use in 22 699 children. Antibiotics were classified as access, watch or reserve, according to the World Health Organization’s AWaRe categories. The percentage of health-care facilities across countries with specific antibiotics available and the proportion of children receiving antibiotics for key clinical syndromes were estimated. Findings The surveys assessed the availability of 27 antibiotics (19 access, 7 watch, 1 unclassified). Co-trimoxazole and metronidazole were most widely available, being in stock at 89.5% (interquartile range, IQR: 11.6%) and 87.1% (IQR: 15.9%) of health facilities, respectively. In contrast, 17 other access and watch antibiotics were stocked, by fewer than a median of 50% of facilities. Of the 22 699 children observed, 60.1% (13 638) were prescribed antibiotics (mostly co-trimoxazole or amoxicillin). Children with respiratory conditions were most often prescribed antibiotics (76.1%; 8972/11 796) followed by undifferentiated fever (50.1%; 760/1518), diarrhoea (45.7%; 1293/2832) and malaria (30.3%; 352/1160). Conclusion Routine health facility surveys provided a valuable data source on the availability and use of antibiotics in low- and middle-income countries. Many access antibiotics were unavailable in a majority of most health-care facilities.


Introduction
The reliable availability of affordable, high-quality antibiotics remains a major global concern. 1,2 Antibiotics are vital for preventing and treating bacterial infection, without which the risk of surgery becomes greater, managing noncommunicable disease becomes more difficult and universal health coverage becomes less attainable. Sustainable development goal (SDG) 3.8 includes the achievement of, "access to safe, effective, quality and affordable essential medicines and vaccines for all. " 3 However, although ensuring universal access to antimicrobials can save millions of lives, 4 excessive and inappropriate use must be limited to avoid the development of antimicrobial resistance.
An insight into the specific types of antibiotics available and used in different countries is vital. When the correct medication is not available, it may be substituted by an alternative, such as a broad-spectrum antibiotic, and patients may buy over-the-counter medicines that could be falsified or of a poor quality. These alternatives can be less effective, have more adverse effects and could drive the development of antimicrobial resistance. 5,6 The reduced effectiveness of antimicrobials and the increasing burden of antimicrobial resistance are particularly problematic in low-and middleincome countries, where multidrug-resistant pathogens (e.g. Escherichia coli and Salmonella species) are common. [6][7][8][9] Although it may be unrealistic and undesirable to achieve universal access to all antibiotics at all health facilities, it should be possible to ensure consistent access to key antibiotics. The AWaRe (access, watch, reserve) antibiotics categories (Box 1 and Table 1; both available at: http://www.who.int/bulletin/ volumes/98/3/19-241349) of the World Health Organization's (WHO's) 2019 list of essential medicines includes a core set of antibiotics that should be available everywhere (i.e. access antibiotics) because they are the first and second choice for treating common or severe clinical syndromes. 10,12 These antibiotics are generally narrow-spectrum agents with a low risk of resistance selection and of adverse effects. The other two AWaRe categories are: (i) watch antibiotics, which have a higher risk of toxicity or resistance development; and (ii) reserve antibiotics, which should be used as a last resort in specific clinical situations and whose effectiveness should be preserved. 10,12 Monitoring the progress of efforts to address antimicrobial resistance requires data on not only resistance patterns, but also on the availability and use of antibiotics, and how they are changing. External surveys of health facilities carried out as part of service provision assessments and service availability and readiness assessments include data on antibiotic availability, and potentially provide countries with an overview of the antibiotics available locally (Box 2; available at: http:// www.who.int/bulletin/volumes/98/3/19-241349). 13,14 Other approaches to monitoring the availability of drugs (including antibiotics) are the medicines monitoring tools used to generate data for monitoring SDGs (currently used in five countries and being extended to others). 15 Several countries also track Objective To assess antibiotic availability and use in health facilities in low-and middle-income countries, using the service provision assessment and service availability and readiness assessment surveys. Methods We obtained data on antibiotic availability at 13 561 health facilities in 13 service provision assessment and 8 service availability and readiness assessment surveys. In 10 service provision assessment surveys, child consultations with health-care providers were observed, giving data on antibiotic use in 22 699 children. Antibiotics were classified as access, watch or reserve, according to the World Health Organization's AWaRe categories. The percentage of health-care facilities across countries with specific antibiotics available and the proportion of children receiving antibiotics for key clinical syndromes were estimated. Findings The surveys assessed the availability of 27 antibiotics (19 access, 7 watch, 1 unclassified). Co-trimoxazole and metronidazole were most widely available, being in stock at 89.5% (interquartile range, IQR: 11.6%) and 87.1% (IQR: 15.9%) of health facilities, respectively. In contrast, 17 other access and watch antibiotics were stocked, by fewer than a median of 50% of facilities. Of the 22 699 children observed, 60.1% (13 638) were prescribed antibiotics (mostly co-trimoxazole or amoxicillin). Children with respiratory conditions were most often prescribed antibiotics (76.1%; 8972/11 796) followed by undifferentiated fever (50.1%; 760/1518), diarrhoea (45.7%; 1293/2832) and malaria (30.3%; 352/1160). Conclusion Routine health facility surveys provided a valuable data source on the availability and use of antibiotics in low-and middleincome countries. Many access antibiotics were unavailable in a majority of most health-care facilities.
stock-outs of medicines (usually including some antibiotics) at individual facilities as a performance metric for health systems.
The aim of this study was to determine whether external assessments of health facilities in low-and middleincome countries can provide data on antibiotic availability and use in general, and on the availability of key antibiotics in particular. To do this, we used data from service provision assessment and service availability and readiness assessment surveys in low-and middle-income countries to calculate the proportion of health facilities that held stocks of core antibiotics in each country, and the proportion of children prescribed antibiotics for key clinical syndromes.

Methods
The service provision assessment includes a cross-sectional, health facility survey developed by ICF International Inc. under the Demographic and Health Surveys (DHS) programme funded by the United States Agency for International Development. 13 The service availability and readiness assessment surveys are conducted by WHO using very similar methods (Box 2 and Table 2; both available at: http://www.who.int/bulletin/volumes/98/3/19-241349). 16 Full details of the surveys' procedures, methods and questionnaires are available online. 17,18 For both types of survey, health facilities were selected in each country from national facility lists, which included private, non-profit and faith-based hospitals, and health centres. 19 The surveys used nationally representative samples of the formal health system in all countries except Haiti, Malawi, Mauritania, Namibia, Rwanda and Uganda, where all or almost all facilities were included.
Each assessment was based on an inventory questionnaire completed by trained interviewers from WHO or the DHS programme during a visit to the health facility and provided an external validation of the facility's functioning. 20 Antibiotics were audited to determine if they were in stock at each facility on the interview day. In most service provision assessment surveys, interviewers also observed child consultations to assess adherence to standards of care provision and antibiotic prescription.
Details of the survey design are available online for most countries. 13,21 In addition, data sets for the service provision assessments are publicly available from the DHS programme and service availability and readiness assessment data sets are available from WHO. The surveys included in our study sample were those: (i) for which microdata were available (rather than just survey reports); (ii) that had been conducted after 2000 (studies completed between 1997 and 2000 were less comparable because different survey instruments were used); and (iii) that provided the most recent data set available for the country (available in the data repository). 22 For countries where several surveys had been performed, we used the latest survey that provided data on antibiotic use.

Antibiotic availability and use
An antibiotic was considered available at a facility if the medications in stock on the assessment day were within their usage dates and were, therefore, available for patients, as stipulated in WHO methods for measuring medicine availability. 23 Oral and intravenous formulations were assessed separately. Each survey questionnaire was country-specific and the number of antibiotics assessed varied slightly between countries. If the availability of a particular antibiotic was not assessed in a country, data for that antibiotic were classed as missing data. The availability of an antibiotic in a country was defined as the percentage of health facilities in that country where the antibiotic was available. The median and interquartile range (IQR) of the percentage availability across all countries were calculated. Availability is presented according to AWaRe categories ( Table 1).
The use of antibiotics for treating particular illnesses in children was assessed in service provision assessment surveys that included observations of child consultations. Trained interviewers asked health-care providers (e.g. a medical doctor, nurse, nonphysician clinical specialist or midwife) about the children's diagnoses and what treatment was prescribed or provided. Diagnoses were based on the children's medical history and physical examinations, except for malaria, where the diagnosis was based on a rapid diagnostic test, blood smear microscopy or clinical findings, depending on the services available -details are provided in the online observation protocol. 24 The percentage of children who were prescribed, or provided with, an antibiotic for each condition diagnosed was calculated for each survey country individually and overall. The diagnostic categories were: (i) pneumonia; (ii) asthma; (iii) other respiratory tract infection, including other upper respiratory infections and unknown respiratory illness; (iv) ear infection; (v) throat infection; (vi) diarrhoea; (vii) malaria; (vii) undifferentiated fever or measles; and (viii) any other illness. If a child was diagnosed with more than one condition, they were regarded as being diagnosed with the condition for which they were most likely to receive an antibiotic, this was determined using WHO's 2017 Model list of essential medicines for children. 25 The association between the availability and use of each antibiotic was assessed by multivariable logistic regression, which included adjustment for confounding variables, such as the child's sex and age, the survey country and year, the type of facility, the condition diagnosed, the facility's managing authority, the role of the health-care provider and season. Our study abided by WHO ethics and research committee rules and procedures on research involving human participants.

Results
We identified 65 (38 service availability and readiness assessment and 27 service provision assessment) surveys conducted between 1997 and 2017. Although other surveys may have been carried out, we were not able to obtain either data or reports for our analysis. Of the 65, we excluded 3 because they were conducted between 1997 and 2000, 19 because more recent data were available for the country surveyed and 22 because no microdata were available (available in the data repository). 22 The final sample included surveys from 20 locations (13 service provision assessment surveys and 8 service availability and readiness assessment surveys) conducted between 2004 and 2017, mainly in Africa. They covered a total of 13 561 health facilities (Table 3), of which 9111 (67.2%) were government facilities. The most common type of facility was the health centre, which comprised 39.1% (5302) of facilities. Overall, the    The total number of physician types may not equal the number of consultations observed because nonclinical personnel (such as laboratory officers) conducted the consultations in some cases. c The technical qualification of health-care assistants was country-specific; assistants included community health workers, medical assistants or officers and clinical assistants or officers.

Research
Assessing antibiotic use in low-and middle-income countries Rebecca Knowles et al.
surveys investigated the availability of 27 antibiotic formulations (19 access, 7 watch and 1 unclassified antibiotic); 17 were oral and 10 were intravenous. Ten of the service provision assessment surveys collected data on antibiotic use in a total of 22 699 children (Table 4): 99.4 % (21 604/21 715) of children whose ages were known were younger than 5 years of age.

Antibiotic availability
The median availability of all antibiotics at all health facilities in the surveys was 48.9%. The access antibiotics were most often investigated in surveys and were the most widely available at health facilities. Although no access antibiotic was universally available, the median proportion of facilities across countries with co-trimoxazole, metronidazole and amoxicillin available was 89.5% (IQR: 12.6%), 87.1% (IQR: 15.9%) and 83.8% (IQR: 26.4%), respectively (Fig. 1). Some access antibiotics (i.e. ampicillin, cloxacillin, amoxicillin with clavulanate, tetracycline, cefalexin and clindamycin) were available in a median of 30.9% or fewer health facilities, although cefalexin and clindamycin were assessed in only six and five surveys, respectively ( Table 5 available at: http://www.who. int/bulletin/volumes/98/3/19-241349, and data repository). 22 The surveys assessed seven watch antibiotics, which were less frequently available than access antibiotics. The most widely available watch antibiotic was erythromycin, which had a median overall availability of 65% (IQR: 34; Fig. 1, Table 5, and data repository). 22 Across all AWaRe categories, there were some large variations between and within countries; for example, the proportion of facilities with benzathine benzylpenicillin (an access antibiotic) in stock ranged from 4% in Sierra Leone to 96% in Benin (Table 5). In total, 17 access and watch antibiotics were, on average, stocked by fewer than 50% of facilities.

Research
Assessing antibiotic use in low-and middle-income countries Rebecca Knowles et al.

Discussion
Using the data from service provision assessment and service availability and readiness assessment surveys have great potential for informing countries about the pattern of antibiotic use at health facilities. In addition, the data can also be used by antimicrobial resistance coordination committees. As the majority of facilities surveyed in our study were health centres, clinics or dispensaries, it is appropriate that access antibiotics were more widely available than watch antibiotics. Of access antibiotics, several were available at most facilities, such as amoxicillin and co-trimoxazole. However, other access antibiotics were much less available: gentamycin, which is used for treating neonatal sepsis and other severe infections, 26 was available at only 56.6% of facilities. Moreover, most facilities had shortages of watch antibiotics,   Percentages were calculated using the number of children diagnosed with the condition in each country given in Table 4.
Assessing antibiotic use in low-and middle-income countries Rebecca Knowles et al.
which have a key therapeutic role in some infections. Although the quantity required may be small, they should be available at all health facilities.
Overall, we found that 60.1% of children whose consultations were observed were prescribed an antibiotic. Although there may be valid reasons for treating conditions, such as upper respiratory tract infection and diarrhoea, the high percentage of prescribing suggests antibiotics were being used inappropriately, as has been observed both anecdotally and in other studies. [27][28][29] Among diagnoses, respiratory conditions had the highest percentage of antibiotic prescription. However, many children were prescribed antibiotics for conditions for which they are not usually indicated, including undifferentiated fever, diarrhoea and malaria.
Globally there is considerable debate about the importance of access to antibiotics, but this frequently focuses on national supplies. Some studies have used pharmaceutical sales data, which reflect the antibiotic consumption of whole countries rather than individuals or communities. 5,30,31 Nevertheless, despite the different data sources used, the proportion of prescribed antibiotics that were access antibiotics was broadly similar across paediatric studies. One international study found that 76.0% of all antibiotics used were access antibiotics (compared with 72.4% in our study) and 30.7% were amoxicillin (compared with 33.9% in our study). 30 However, there are inconsistencies between health facility surveys and pharmaceutical sales studies because the antibiotics with the highest sales are not always available at health facilities. For example, in our study, amoxicillin with clavulanic acid was available at only 17.4% of health facilities, whereas recent global pharmaceutical sales data indicate it is used in almost equal amounts to amoxicillin, which was available at 83.8% of facilities in our study. 31 Similarly, cefixime is also one of the most commonly used antibiotics according to pharmaceutical sales, but was available at only 12.9% of facilities. 31 These discrepancies suggest there may be a divergence between supply and use in some countries. There might be differences in prescribing patterns between health facilities and other vendors who were not covered in our study, but whose antibiotic sales were reflected in pharmaceutical data.
The lack of high-quality data at the community level presents a bar-rier to understanding antibiotic access and use. One systematic analysis of antibiotic consumption in countries in the WHO African Region found that studies were frequently limited by their small sample size, a lack of adherence to WHO recommendations on reporting medicines and poor reporting of study details, 32 which illustrates the difficulty of obtaining meaningful data in lowand middle-income countries. Despite methodological difficulties, community studies in India, Nepal, Viet Nam and in countries in sub-Saharan Africa have reported high and possibly inappropriate percentages of antibiotic prescriptions in the range of 41.0-85.0%, similar to our finding in children. [33][34][35][36][37][38][39] The main advantage of these analyses is that they can identify the diseases for which antibiotics are most commonly used and they consistently showed most antibiotic prescriptions were for respiratory conditions. 33,34,36,37,40,41 They also documented high antibiotic use by individuals with malaria, despite rapid malaria diagnostic tests being available. The antibiotic prescription rates for diarrhoea and undifferentiated fever reported in several low-and middle-income countries were comparable with our findings. 34,39,[42][43][44][45][46] In our study, amoxicillin accounted for 33.9% of antibiotics prescribed to children, which was low relative to other community studies where amoxicillin made up over half of prescriptions. 34,40,47,48 We found the variation in availability between countries was greater for amoxicillin than co-trimoxazole, which is often widely distributed within countries because HIV programmes have made substantial investments in supply chain management to ensure its availability for prophylaxis. 49  Diagnoses were made by health-care providers in consultations observed during service provision assessment surveys. For Rwanda and Uganda, which recorded amoxicillin and co-trimoxazole in a single category, we assigned all entries in that category to co-trimoxazole because the availability of co-trimoxazole was greater than that of amoxicillin in both countries. Other respiratory tract infection includes unspecified upper respiratory infection and other unknown respiratory illnesses. Undifferentiated fever includes cases where the cause of the fever was not known or the diagnosis was measles. Diarrhoea includes diarrhoea and dysentery. An observation was classified as no diagnosis when no diagnosis was recorded but treatment was still given. The data for each country is available from the corresponding author. Countries included: Egypt, Haiti, Kenya, Malawi, Namibia, Nepal, Rwanda, Senegal, Uganda and United Republic of Tanzania.

Research
Assessing antibiotic use in low-and middle-income countries Rebecca Knowles et al.
the ready availability of co-trimoxazole can result in it being heavily prescribed even when inappropriate. 37,47 As expected, availability was correlated with use, though the association was weak, probably because patients were instructed to buy specific medicines elsewhere if they were not available at a facility. One strength of our study is that the surveys were reliably conducted in many countries, had large sample sizes and adopted a standard approach. 20 Both assessment surveys are well suited to assessing all aspects of health-care provision, and the physical and human resources required. Although their total cost is high, these surveys offer a more efficient and cost-effective way of obtaining basic information on antibiotic consumption than specific surveys. Moreover, they can help monitor the actions taken to manage antimicrobial resistance both nationally and globally.
There are some limitations, however. First, our study was primarily an exploratory analysis of the usefulness of health facility surveys in low-and middleincome countries. Second, the survey data did not cover drug sources, such as local pharmacies or informal providers and not all antibiotics were included (e.g. no reserve antibiotics were monitored). Third, the availability of a medication may not correlate with its use because: (i) some countries use drug availability as a performance indicator, which may encourage suppliers to keep key medicines in stock instead of dispensing them; and (ii) the cost of an antibiotic (which was not recorded in surveys) may have been high enough to prevent individuals accessing it. Fourth, surveys were crosssectional and thus reflected the status of facilities on one specific day, which may limit the generalizability of a survey's findings beyond the specific country and year in which it was conducted. In particular, as some surveys were conducted over 10 years ago (i.e. in Egypt, Guyana and Uganda), recent antibiotic availability may have been underestimated. Finally, health workers are more likely to prescribe in accordance with guidelines when being observed.
More surveys are planned and underway. Future surveys will also collect information on the price of essential medicines to patients. Although assessments of facilities may not be able to provide detailed information on the formulation of drugs or on prescribing behaviour, they will continue to give insights into antibiotic availability and use in primary and secondary care, where monitoring capacity is limited but antibiotic use is greatest. Future surveys would benefit from the inclusion of standard questions on antibiot-ics based on AWaRe categories. As data from more surveys become available, future research will be able to monitor changing patterns of use.
This study of service provision assessment and service availability and readiness assessment surveys of health facilities in low-and middle-income countries demonstrated that more data on antibiotic availability and use are available than previously reported. These data can help countries evaluate the risk of antimicrobial resistance. Both surveys provide an important and expanding resource that can be used to improve understanding of local and global antibiotic consumption patterns, without the need for collecting new data. Our study found that first-line access antibiotics were unavailable at many health facilities in some countries, investment in antibiotic supply chain management is therefore needed. We also found that antibiotics were used extensively in primary care, often for conditions for which they are not usually indicated. ■ Acknowledgements RK was an intern at WHO's Antimicrobial Resistance Secretariat in Geneva when this study was done.

Research
Assessing antibiotic use in low-and middle-income countries Rebecca Knowles et al.
Resultados Las encuestas evaluaron la disponibilidad de 27 antibióticos (19 de acceso, 7 de vigilancia, 1 sin clasificar). El cotrimoxazol y el metronidazol fueron los antibióticos con mayor disponibilidad, ya que se encontraban en existencias en el 89,5 % (rango intercuartil, IQR: 11,6 %) y el 87,1 % (IQR: 15,9 %)  Assessing antibiotic use in low-and middle-income countries Rebecca Knowles et al. None a Access antibiotics are the first and second choice for treating common or severe clinical syndromes, watch antibiotics have a higher risk of toxicity or resistance development and reserve antibiotics should be used as a last resort to preserve their effectiveness (Box 1). b This column lists only the one unclassified antibiotic that was included in surveys. c Surveys include the service availability and readiness assessments and service provision assessments.

Box 2. Service availability and readiness assessment surveys and service provision assessment surveys
Service availability and readiness assessment surveys and service provision assessment surveys are health facility surveys that assess the availability and readiness of different health services in a country with reference to accepted standards of care. In addition, service provision assessment surveys also include observations of patient care and evaluate client satisfaction with service delivery. Both survey tools generate indicators of service availability and readiness that provide reliable and regular information on: (i) service delivery processes and provisions, such as the availability of key human and infrastructure resources, basic equipment, basic amenities, essential medicines and diagnostic capacities; and (ii) the readiness of facilities to provide basic health-care interventions, such as family planning, child health services, basic and comprehensive emergency obstetric care, and the treatment of HIV infection, tuberculosis, malaria and noncommunicable diseases. Currently, service availability and readiness assessment surveys are being implemented in 32 countries and service provision assessment surveys are being implemented in 17. As the average time between surveys is 2 to 3 years, they are not intended to replace routine supervision and monitoring. Instead, they collect information that can provide an external validation of whether health systems are functioning as reported. In particular, they provide an ideal way of verifying service standards in countries where accreditation and certification systems are undergoing revision and improvement.
Assessing antibiotic use in low-and middle-income countries Rebecca Knowles et al. Table 2. Comparison of service provision assessment and service availability and readiness assessment surveys Characteristic Survey Service provision assessment 13 Service availability and readiness assessment 16 Survey conducted by DHS using the USAID-WHO inventory questionnaire WHO and USAID Background The survey was developed by updating the method used in service availability and readiness assessment surveys to cover more areas and to give a more comprehensive overview. The topics covered include equipment, amenities, essential medicines, diagnostic capacity and the readiness of health facilities to provide basic health-care interventions for family planning, child health, obstetric care, HIV infection, tuberculosis, malaria and noncommunicable diseases The survey was developed through a joint WHO-USAID collaboration. The health facility assessment tool was designed to assess and monitor service availability and the readiness of a country's health sector, and to generate evidence to support planning and management. The topics covered include equipment, amenities, essential medicines, diagnostic capacity and the readiness of health facilities to provide basic health-care interventions for family planning, child health, obstetric care, HIV infection, tuberculosis, malaria and noncommunicable diseases Survey elements (i) Inventory questionnaire (including data on antibiotic availability); (ii) observation protocols and interviews with clients leaving facilities about antenatal care, family planning and sick children (including data on antibiotic use in children); and (iii) Service provision assessment health worker and health-care provider interview questionnaire Inventory questionnaire (including data on antibiotic availability) Method and health facilities included (i) The survey typically includes 400-700 facilities (surveys can be carried out either as a census or as a representative sample of health facilities) selected from the country's master facility list; (ii) surveys are typically conducted by 10-15 teams, each comprising 3-4 interviewers (mostly health workers); and (iii) interviewers collect data from the people in charge or the most knowledgeable people at each facility using the inventory questionnaire, observe consultations and interview clients leaving facilities  Co-trimoxazole