Children in low- and middle-income countries (LMICs) are receiving an average of 25 antibiotic prescriptions during their first five years of life, an excessive amount that could harm the children’s ability to fight pathogens as well as increase antibiotic resistance worldwide, according to a new study from the Swiss Tropical and Public Health Institute (Swiss TPH) and Harvard T.H. Chan School of Public Health.
“We knew children in LMICs are sick more often, and we knew antibiotic prescription rates are high in many countries.
What we did not know was how these elements translate into actual antibiotic exposure – and the results are rather alarming,” said Günther Fink, lead author of the study and head of the Household Economics and Health Systems Research unit at Swiss TPH.
The study – the first to look at total antibiotic prescribing in children under the age of five in LMICs – will be published on 13 December in The Lancet Infectious Diseases.
Global health threat
Antimicrobial resistance is one of today’s biggest threats to global health and development, according to the World Health Organization. One factor contributing to this global health threat is the excessive use of antibiotics worldwide.
Previous studies have shown that antibiotics are overprescribed to children in many countries. In Tanzania, for instance, several studies have shown that over 90% of children who visit a health facility receive an antibiotic, although only in about 20% of the cases treatment was actually required.
The research team from Swiss TPH and Harvard Chan School analyzed data from 2007-2017 from health facilities and household surveys from eight countries: Haiti, Kenya, Malawi, Namibia, Nepal, Senegal, Tanzania, and Uganda.
The study found that, on average, children received 25 antibiotic prescriptions through age five – a “remarkable” estimate, the authors wrote, given that two antibiotic prescriptions per year is considered excessive in many high-income settings. Results showed that antibiotics were administered in 81% of cases for children with a respiratory illness, in 50% for children with diarrhea, and in 28% for children with malaria.
The researchers found that the number of antibiotic prescriptions in early childhood varied from country to country: While a child in Senegal received approximately one antibiotic prescription per year in the first five years of life, a child in Uganda was prescribed up to 12. In comparison, a prior study showed that children under five in Europe receive less than one antibiotic prescription per year on average.
“This number is still high given that the vast majority of infections in this age group are of viral origin,” said Valérie D’Acremont, a study co-author and head of the Management of Fevers group at Swiss TPH.
“What is unique about this study is that it provides a much more comprehensive picture of pediatric antibiotic exposure in LMICs than what has been reported previously.
It combines both household data on where and when children are brought for care with data from direct observations of health care workers caring for sick children,” said Jessica Cohen, the Bruce A. Beal, Robert L. Beal, and Alexander S. Beal Associate Professor of Global Health at Harvard Chan School and senior author of the study.
Impact on children
“The consequences of antibiotic overprescription not only pose a huge threat to global health, but can also result in a concrete health impact for these children,” said Valérie D’Acremont. “Excess antibiotic use destroys the natural gut flora which is essential to fighting pathogens.”
A Swiss TPH research project is underway to better comprehend the health impact of overusing antibiotics on children.
“Understanding the concrete impact on individual children is crucial to achieve a policy change,” said Fink. His research team is currently comparing policies at a country level to identify best practices that lead to lower antibiotic prescription rates.
Globally, antibiotic use is increasing, up 65% between 2000 and 2015 [1]. The majority of this increase has been caused by rapid expansion of use in low- and middle-income countries (LMICs). Consumption rates in many LMICs are catching up with higher-income countries [1] but usage is often unnecessary due to lack of supportive regulation and widespread informal use [2]. While access to antibiotics is of clear public health benefit, the inappropriate use of antibiotics is a major driver of antibiotic resistance [3] and requires urgent attention. To inform context-appropriate policy in LMICs, we conducted a review of supply side interventions and their impact on unnecessary and inappropriate use of antibiotics in human and animal health systems. The aim was to determine the evidence base for changing the practices of antibiotic prescribers and gatekeepers in LMICs.
Previous systematic reviews have found evidence to support a range of effective and safe interventions to reduce antibiotic prescribing in human health settings. Interventions that combine restrictive and enabling policies have shown better results [4]. Interventions that are interactive and multi-faceted have stronger effects, for example education with feedback and monitoring mechanisms [5,6]. However, the majority of studies included in these reviews are from high-income countries, and within that a limited range of settings (e.g., in-patient prescribing) and methods. One review [6] looked specifically for evidence on doctor’s prescribing in primary care in upper middle-income countries. However, their review only included eight studies and the authors noted that the quality of evidence was limited. We are not aware of any reviews of prescribing interventions in animal health settings. Taken together, these reviews have found encouraging results but there are questions about how generalizable they are beyond the specific targeted groups (e.g., doctors) and homogenous prescribing contexts (e.g., formal providers in relatively well organised human health systems in high-income countries). LMICs, as defined by World Bank, include a variety of health system arrangements, which makes generalizations difficult. However, in many, the human and animal healthcare contexts can be described as “pluralistic”, where people consult a variety of public, private, traditional, biomedical, formal and informal providers, and where the capacity of governments to regulate these providers is variable, and is increasingly recognised as a major concern [7]. Effective action is complicated by high rates of infection due to poor preventative public health combined with inconsistent and inequitable access to essential medicines. As such, research on LMICs interventions and contexts is urgently needed and is the focus of this review.
The monitoring of unintended consequences requires particular attention as the supply and distribution of antibiotics has the potential to include competing objectives and is sensitive to interventions not specifically targeting antibiotic use. For example, the use of rapid diagnostic tests to improve malaria diagnosis has led to an increase in antibiotic use in some settings [8,9].
The review set out to answer the following questions (for human and animal health systems):
- -What supply-side interventions have been tried in LMICs to reduce inappropriate antibiotic prescribing and sale?
- -Where have interventions been implemented e.g., country and kind of health care settings?
- -Which kinds of interventions have shown promise in reducing inappropriate prescribing?
- -What information exists on the contexts of interventions and attempts to influence antibiotic use and prescribing?
Results
We present summary statistics of the intervention studies included in the review. Table 3 shows the number of studies included from each country. The largest number of studies are from China (25 out of 70 studies), followed by India (6 studies). The high number of studies in China may be in response to the major health reforms that took place there, e.g., in 2009, which saw the implementation of Essential Medicine Lists, and, during 2011–2013, when there was a national campaign on improving the use of antibacterial drugs. There are two cross-country studies—one in Brazil and Mexico, and the other in Vietnam and Thailand. The rest of the studies were conducted in a single-country setting.
Table 3
Country settings.
Country, by World Health Organisation Region | |
---|---|
African region | 15 |
Kenya | 3 |
Mali | 1 |
Malawi | 1 |
Nigeria | 3 |
South Africa | 2 |
Sudan | 2 |
Tanzania | 3 |
Region of the Americas | 5 |
Argentina | 1 |
Brazil | 2 |
Brazil and Mexico | 1 |
French Guiana | 1 |
European region | 7 |
Republic of Srpska, Bosnia and Herzegovina | 1 |
Serbia | 2 |
Turkey | 4 |
Eastern Mediterranean region | 3 |
Pakistan | 1 |
Iran | 2 |
South-East Asian region | 13 |
India | 6 |
Indonesia | 1 |
Nepal | 3 |
Thailand | 2 |
Thailand and Vietnam | 1 |
Western Pacific Region | 27 |
China | 25 |
Vietnam | 2 |
Total | 70 |
Intervention Settings
We found no interventions addressing antibiotic prescribing in animal health. All intervention studies were from human healthcare settings. The overwhelming majority of the reviewed studies evaluated interventions at public prescribing facilities rather than private (see Figure 1). This trend is particularly pronounced among studies conducted in primary care settings, where 15 out of 22 studies examined the public sector. The exception was for pharmacies where seven out of nine studies were conducted in the private sector. This is likely because pharmacies and drug shops are usually private enterprises.
In the hospital settings, 16 and 4 studies were conducted in public and private hospitals, respectively, while in the remaining 16 it is unclear. This was due to the fact that studies mentioned hospitals’ names but we were unable to judge whether the hospitals concerned are public or private from available sources. Three other studies are not included in the figures: two were from mixed public and private sector settings, and one was not clear.

Sector where intervention was tested, by type of health setting.
There was also a majority of studies from urban areas, as shown in Figure 2. In total, 41 out of 70 studies looked at interventions in urban areas. In particular, studies conducted at hospitals and primary care mostly evaluated urban settings. Two of the studies carried out in mixed health care settings were from rural areas, while the remaining one study addressed both rural and urban areas.

Location (urban or rural) where intervention was tried, by type of health setting.
Discussion
The review confirms previous studies [4,5,6] that have found that multiplex interventions combining different strategies to influence behaviour tend to have a higher success rate than interventions based on single strategies. Another similarity to other reviews [4] is that many of the interventions which worked well combined restrictive and enabling strategies, i.e., educational techniques combined with forms of monitoring. To the best of the authors’ knowledge, this is the largest review to document these patterns in LMIC countries.
The inclusion of evidence on prescribing contexts provides important additional insight. These studies highlighted a web of interacting influences on prescriber behaviour that cannot be reduced to simple motivations, such as profit. Instead, health system quality and availability, education, perceptions of patient demand, bureaucratic processes, competition, and cultures of care all play a role.
Physician perceptions of patient demand [15,17,19,20,22,26,27,28] requires more attention as observations do not always support this [28]. Indeed, elsewhere studies that have observed clinical interactions have shown how doctors prescribe irrespective of patient demand [37] and that patient satisfaction is not necessarily linked to whether a prescription was received [38], although these studies are old. A more recent study from multiple European settings [39] found that, while patients frequently hope for and expect antibiotics, it is less common for them to ask for them explicitly.
Patients were prescribed antibiotics considerably more often than they asked for them (54% compared to 10%, respectively), and more than was clinically necessary. The same study found that, in general, patients were satisfied with their consultations, whether prescribed antibiotics or not. Assessing the level and influence of patient demand is difficult as research is often based on reported behaviour and perceptions rather than observations.
It is striking the extent to which different providers blame others within the system: health workers blame patients, pharmacists blame doctors and their incorrect prescriptions, and drug sellers blame competitors saying patients will just “go somewhere else”. Such blame shifting highlights a more general importance of interactions between people within the health system: pharmacists fear challenging doctors’ prescriptions, drug sellers fear losing their customers, and many doctors and pharmacists get a large proportion of their information about antibiotics from drug detailers who also offer incentives.
In discussions of interventions to address AMR, the temptation is to focus on awareness raising and changing the behaviour of individuals, be they healthcare workers including prescribers and pharmacists, or patients. However, the stewardship, supply and use of antibiotics in human and animal healthcare is best viewed with a broad systemic lens [2,40]. This encompasses an appreciation of how the behaviour of individuals is mediated by a range of structural and contextual factors operating at different scales, which can affect individual agency and decision-making.
Scholars have advocated for a consideration of behavioural change interventions within the context of everyday social practices [41]; for a whole system approach (at macro, meso and micro levels) to analysing policy outcomes in the health sector [42]; for a consideration of formal and informal health markets for drugs and services for both people and animals [7,43]; for an appreciation of incentive structures in drug supply chains and the effects on over-the-counter drug retailers and informal providers [44]; and for an approach to healthcare worker behavioural change interventions that appreciates individual rationales as well as the contextual settings in which specific behaviours are practiced [45]. The overall view is also one of decentred governance and regulation, which departs significantly from traditional “command and control” models [46].
It is notable therefore that while a majority of studies we identified in this review implemented multiplex interventions, most of these took place within one health setting, e.g., a stewardship programme in a hospital. Very few studies targeted different kinds of healthcare provider and different kinds of health settings. One exception was by Hoa et al [47] who attempted an educational programme for all providers in one district, public, private, formal and informal and primary or secondary care.
In this example, however, there was a particularly high drop-out rate among informal providers illustrating the challenges of engaging hard to reach groups in interventions. Moreover, important interactions between types of providers are not being addressed and it is rare to find examples of whole system approaches which have been recommended [2,40].
China provides one of the few examples of system-wide interventions at the national and regional level. There is a need for coordinated action and research across multiple settings and actors throughout and between the human and animal supply chain.
Significantly, our study was unable to identify a single study on interventions that addressed prescribing behaviour for veterinary use of antibiotics, and only one context study on agricultural use. Evidence is emerging that antibiotics critical for human health are being used in animal farming [48] including colistin [49,50].
Research needs to address agricultural use of antibiotics and its overlap with human health system in terms of the antibiotics used, healthcare personnel, and transmission pathways of resistant pathogens. Measures to improve essential medicine use provides a precedent for success [51].
The National Action Plan process provides an opportunity to facilitate this. In particular, employing the “One Health” lens [52,53], taking the intersection among human, veterinary, and environmental health into account, can help address the multi-sector nature of AMR and develop systematic strategies to tackle this challenge.
In addition to the major lack of evidence documenting the effectiveness of interventions to improve veterinary prescribing, this review has identified some gaps in the human evidence. Studies in hospitals were most common, possibly as hospitals are more observable and controllable environments. The overwhelming majority of studies evaluated interventions at public facilities.
There was also a majority of studies from urban areas, with less from rural areas, possibly because hospitals are over represented in the data and tend to be in urban areas. Private providers, and in particular those in primary care settings, is another gap. Informal private providers are especially underrepresented. Evidence from LMICs suggests that rural and/or poor people extensively rely on healthcare services provided by informal practitioners and suppliers of drugs [54,55].
Therefore, understanding the antibiotic prescribing practices of the informal providers, and improving them, has important implications for health of many marginalized people. There were no interventions in this review involving drug retailers or the pharmaceutical companies despite their prominent role in the contextual studies.
Public understanding and demand is a critical piece of the puzzle and requires action but it must be based on accurate understandings of lay knowledge and demand otherwise messaging and educational interventions will fail. Research is needed into public views and their influence on demand and prescriber behaviour. This should form part of a holistic view of supply and demand.
Our study has a number of limitations. While the literature search was extensive, it may not be exhaustive and studies prior to 2000 were not included. The large number of heterogeneous studies meant we were unable to assess quality of the evidence in detail and we did not compare or combine the effects of selected studies.
Further research is needed to assess the strength and magnitude of effects for promising interventions. A majority of studies (n = 20) reported positive effects while far fewer reported mixed, negative or no effects (n = 13, 4 and 3 respectively). This indicates that publication bias may be an issue. Nevertheless, the review was able to identify trends and gaps in the evidence base.
Conclusions
The evidence collected in this review comes from a range of health care settings, for example hospitals, primary care, pharmacists or drug shops, and from interventions targeting different types of health providers including doctors, nurses, and drug sellers. This review has identified evidence on interventions to improve antibiotic use among providers in LMIC settings, which, in the authors’ opinion, has not been well represented in previous reviews.
The review found that multiplex interventions that combine different strategies to influence behaviour tend to have a higher success rate than interventions based on single strategies. However, the evidence base is uneven with hospital and urban contexts over-represented for interventions.
There is much less evidence on private providers, especially in primary care settings. Informal private providers who play a major role in drug distribution in LMICs are especially underrepresented. Furthermore, there were no interventions involving drug detailers or the pharmaceutical companies despite their prominent role in the contextual studies. Strikingly, no study was identified that addressed veterinary prescribing of antibiotics.
Evidence on prescribing contexts highlights a web of interacting influences on prescriber behaviour including health system quality and availability, education, perceptions of patient demand, bureaucratic processes, profit, competition, and cultures of care.
These contextual studies underscore the importance of interactions between different people within the health system. Although a majority of studies implemented multiplex interventions, most of these took place within one health setting, e.g., a stewardship programme in a hospital. Very few studies targeted different kinds of health provider and interactions across different kinds of health setting.
There is an urgent need for coordinated multi-actor studies including multiple settings, by taking a One Health approach, including agricultural settings, and actors throughout the supply chain.
More information: “Antibiotic exposure among children under age five: A cross-sectional analysis of nationally representative facility and household surveys in 8 low- and middle-income countries,” Günther Fink, Valérie D’Acremont, Hannah Leslie, Jessica Cohen, The Lancet Infectious Diseases, December 13, 2019, DOI: 10.1016/S1473-3099(19)30572-9 , http://www.thelancet.com/journals/laninf/article/PIIS1473-3099(19)30572-9/fulltext