Introduction

Drug resistant infections are increasing worldwide and there is global consensus on the need for urgent action to address antimicrobial resistance (AMR) [1]. Optimising the use of antibiotics is one of the central pillars of the World Health Organization (WHO) global action plan to address AMR [2]. The ramifications of AMR are being more acutely felt in low-and-middle-income countries (LMICs), where there is a higher burden of infectious diseases and health systems are weaker [3]. In low-income contexts where routine diagnostic microbiology facilities are scarce, data on the epidemiology and burden of AMR is limited [4–6].

In the past two decades, academic and policy research has shown a substantially increased use of antibiotics in LMICs [7,8]. Due to contextual factors including weaker regulatory frameworks and more fragile health systems, antibiotics are often bought over-the-counter, or obtained through informal networks [9,10]. However, research conducted in LMICs has demonstrated that antibiotic use practices, including self-medication, are heterogeneous [11–13]. Do and colleagues (2021) undertook a survey in six countries in Asia and Africa, and found substantial differences in self-medication of antibiotics, with rates lower in Mozambique (8·0%) and South Africa (1·2%) than in Bangladesh (45·7%) and Ghana (36·1%) [14]. Torres and colleagues (2020) did a systematic scoping review of self-medication with antibiotics in 10 low- and middle-income countries, and found a prevalence of self-medication in the preceding year that ranged from 8.1% to 93% across countries [15]. Both these studies found that factors shaping self-medication were complex and included access to, and cost and quality of care at health care facilities [16,17]. The diversity in the prevalence of self-medication of antibiotics in different countries underscores the need for accurate national data to guide antibiotic use and AMR policy and practice.

Global antibiotic consumption assessments have primarily relied on pharmaceutical sales records, but this data is frequently unavailable or unreliable, especially in the African region [18–20]. In LMICs, most academic research has been conducted in secondary or tertiary care facilities. While researchers are increasingly investigating antibiotic use in primary health care and community contexts [21–23], there remain significant gaps in knowledge on antibiotic use practices in community settings, with knowledge particularly limited in countries in Southern and Eastern Africa [24,25]. Given the role reduction in antibiotic use (ABU) is expected to play in reducing AMR, generating robust ABU surveillance data is a necessary first step to inform context-specific interventions [26]. This paper aimed to generate data on current antibiotic use in three residential areas in Blantyre, Malawi.

Materials and methods

Study design and population

Data were collected between April and July 2021. We undertook a cross-sectional survey in two densely-populated neighbourhoods (Chilomoni and Ndirande) and one peri-urban neighbourhood (Chileka) in Blantyre, Malawi. Blantyre is a major commercial centre located in the Southern Region of Malawi. The total population of Blantyre District was estimated to be approximately 1 million people in the 2018 Malawi National Census [20], and adult HIV prevalence is estimated to be 18% [21]. Chilomoni and Ndirande are well-established urban neighbourhoods, with poor access to municipal services, and high rates of household poverty. Chileka is located on the periphery of Blantyre and is characterised by a mixture of recently established peri-urban households and households engaged in subsistence farming.

Participants eligible to take part in the survey were adults (aged 18 years or older) resident in households in either Chilomoni, Chileka, or Ndirande. We excluded households where there was no adult available to complete questionnaires, or where the household head declined consent to participate.

The unit of sampling was the household. To establish a household sampling frame for Chilomoni and Ndirande, we obtained neighbourhood boundaries and used georeferenced household data from a previous population census of these neighbourhoods conducted as part of a cluster-randomised trial of TB and HIV interventions in 2015 (Fig 1) [22,23]. Households were randomly selected within each neighbourhood, with a replacement list for when households declined participation or were not identified. As Chileka was not part of the 2015 trial census, we selected random sets of GPS coordinates from within the neighbourhood boundaries.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Location of households sampled for interview.

(A) Malawi boundary data from the United Nations Office for the Coordination of Humanitarian Affairs, Humanitarian Data Exchange (https://data.humdata.org/dataset/geoboundaries-admin-boundaries-for-malawi). (B) Map tile data from OpenStreetMap (https://www.openstreetmap.org/copyright).

https://doi.org/10.1371/journal.pgph.0001946.g001

Results

In total, 217 households were interviewed and included in this analysis (Table 1). A total of 1051 household members were identified by respondents. Numbers of household residents ranged between 1 and 13, with a median of 5 (interquartile range: 3–6). Numbers of household residents were similar between the three study sites (p = 0.771).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Characteristics of households and household respondents, by study site.

https://doi.org/10.1371/journal.pgph.0001946.t001

The majority of household respondents were female (80.2%), and the mean respondent age was 38 years. Household and household respondent characteristics were similar in Chilomoni and Ndirande, but there were substantial differences between these neighbourhoods and Chileka. Whereas in Ndirande (76.3%) and Chilomoni (74.6%) most households had electricity supplied to the dwelling, in Chileka only 8.1% had electricity (p<0.001). 78.4% of household respondents in Chileka reported that they were literate, compared to 91.0% in Chilomoni and 82.9% in Ndirande (p = 0.119). Reported coverage of HIV testing was very high and similar across neighbourhoods, with 93.1% of household respondents reporting having ever previously tested for HIV (p = 0.639 for comparison between sites). Self-reported HIV-positive status was 11.1% overall, and was substantially (but not statistically significantly) lower in Chileka (5.4%) compared to Chilomoni (10.4%) and Ndirande (17.1%), p = 0.131. Antiretroviral therapy coverage was 100% among HIV-positive respondents across all three sites.

Antibiotic recognition

Using the drug bag method, household respondents recognised a median of five antibiotics (IQR: 3–6). The median number of antibiotics recognised in Chileka (4, IQR: 3–6) was slightly lower than in Chilomoni (5, IQR:3–6) or Ndirande (5, IQR: 4–6.25), but this was not statistically significant (p = 0.1176). The most common antibiotics recognised overall by household respondents were: amoxicillin tablets (202/217, 93.1%, 95% CI: 88.9–96.1%); cotrimoxazole tablets (184/217, 84.8%, 95% CI: 79.3–89.3%); and amoxicillin suspension (94/217, 43.3%, 95% CI: 36.6–50.2%). Patterns of antibiotic recognition were broadly similar between the three study sites (Fig 2).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Household respondent antibiotic recognition, by study site.

https://doi.org/10.1371/journal.pgph.0001946.g002

Associations with recognition of antibiotics

In univariable regression analysis, household respondents from Ndirande (relative risk [RR]: 1.18, 95% CI: 1.02–1.36, compared to Chileka), and from larger households (RR: 1.03, 95% CI: 1.00–1.06 per person increase in household size) recognised a significantly greater number of antibiotics (Table 2). Respondents with unknown HIV status (RR: 0.76, 95% CI: 0.58–0.98) and male household respondents recognised significantly fewer antibiotics than women (0.88, 95% CI: 0.75–1.03) although this was not statistically significant. In multivariable regression, male sex (RR: 0.83, 95% CI: 0.69–0.99) and having unknown HIV status (RR: 0.76, 95% CI: 0.57–0.99) were significantly associated with recognition of fewer antibiotics.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Associations with number of antibiotics recognised by household respondent.

https://doi.org/10.1371/journal.pgph.0001946.t002

Antibiotic use

Of the 1051 household members, a total of 67 (6.4%, 95% CI: 5.0–8.0%) were reportedly currently taking one of the antibiotics identified within the drug bag (Table 3). Estimates of household members’ current antibiotic use were similar between Chileka (22/355, 6.2%, 95% CI: 3.9–9.2%), Chilomoni (20/321, 6.3%, 95% CI: 3.8–9.5%), and Ndirande (25/375, 6.7%, 4.4–9.7%). Only five antibiotics (counted by active ingredient) were identified as being currently taken by household members (Table 2). The antibiotic formulation most commonly reported as currently used was cotrimoxazole tablets (39/1051, 3.7%, 95% CI: 2.7–5.0%), followed by amoxicillin tablets (13/1051, 1.2%, 95% CI: 0.7–2.1%), erythromycin tablets (5/1051, 0.5%, 95% CI: 0.2–1.1%) and amoxicillin suspension (5/1051, 0.5%, 95% CI: 0.2–1.1%).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Antibiotic use by household members.

https://doi.org/10.1371/journal.pgph.0001946.t003

Household respondents reported that 440 of the 1051 household members (41.9%) had taken an antibiotic in the last 6-months. The antibiotic formulations that had been reportedly most frequently taken in the preceding 6-months were: amoxicillin tablets (126/1051, 12.0%, 95% CI: 10.1%-14.1%); cotrimoxazole tablets (116/1051, 11.0%, 95%CI: 9.2–13.1%); and amoxicillin suspension (38/1051, 3.6%, 95% CI: 2.6–4.9%). Patterns of reported antibiotic use in the preceding 6-months were broadly similar between the three study sites.

Discussion

The main findings of this detailed cross-sectional survey of community antibiotic use practices in Blantyre Malawi were that current antibiotic use was (67/1051, 6.4%) and recent use (last 6 months) was (440/1051, 41.9%). Both current and recent antibiotic use were limited to a small number of antibiotic formulations, with amoxicillin and cotrimoxazole being the most frequently recognised and used. Both antibiotics are classified as “access” on the WHO’s Access, Watch, Reserve (AWaRe) list, meaning they are commonly used to treat infections and should be clinically available at all times [26]. Previous studies in Malawi have shown rapidly emerging bacterial resistance to these antibiotics, threatening clinical and public health utility [27,28]. In contexts such as Malawi, cotrimoxazole is commonly used for chemoprophylaxis for people living with HIV, and with HIV prevalence in the study populations being high, it is not surprising that this was one of the most commonly used and recognised antibiotic. In contrast, the number of antibiotics recognised was significantly lower among people less likely to access formal health services (such as people with unknown HIV status, a strong indicator of delayed healthcare seeking and poor access where testing is widely available), and men, who frequently do not participate in care-seeking activities for families.

Taken together, these findings suggest that, contrary to popular narratives, use of a wide range of antibiotics and antibiotics on the watch or reserve lists was not widespread in these urban and periurban African settings; rather access to a wide range of antibiotics is likely to be limited for the majority of the population, and potentially not a major contributor to so-called “antibiotic overuse”. Nevertheless, antibiotics such as cotrimoxazole, amoxicillin, and erythromycin have broad spectrum action, and dependence on these antibiotics for community management of illness is associated with generation of resistance. Therefore, ABU and AMR guidance must recognise the healthcare and sociological context within which recommendations are made; improvements in universal healthcare access (including treatment of infection) should be accompanied by high quality surveillance and antibiotic use data, supported by interrogation of social narratives around the determinants of antibiotic use.

In Malawi and similar settings in Southern and Eastern African countries, primary health care facilities are an important source of community care, with services provided for free [29], but delivered under severe resource constraints with high rates of stockouts of many essential medicines [30,31]. International donor programmes play a significant role in funding services, and infectious diseases have received comparably more resources [32]. The provision of cotrimoxazole by the HIV programme at primary care clinics may explain why it is well recognised and used within households [30]. A multi-site study exploring household antibiotic use in Zimbabwe, Malawi and Uganda identified differences in the profile of ABU, with cotrimoxazole being the most frequently used in rural Malawi, amoxicillin in Harare, Zimbabwe, and metronidazole within informal settlements in Uganda [33]. Most differences in use reflected differences in the configuration of the health system and antibiotic supplies.

In Malawi, gender power relations shape household care seeking practices, with men often presenting at health care facilities critically ill, and having delayed diagnosis of diseases of major public health importance, particularly HIV and tuberculosis [34]. Women spend considerably more time than men seeking care and supporting household members to seek care [35]. These factors are likely to explain why women had better recognition and use of antibiotics. While government health care facilities provide care without user fees, care seeking incurs costs for transportation and loss of time from other livelihood activities [36]. Improving access and removing economic barriers to care seeking, are likely to be important interventions to improve the community management of febrile illness, both to improve recognition of severe illness requiring antibiotics with referral to hospital where required, and to restrict unnecessary ABU.

Global public engagement campaigns to improve awareness of antimicrobial resistance have focused predominantly on communicating the need to stop people from indiscriminately using antibiotics with insufficient attention paid to the issues of access [37]. Our findings demonstrate frequent use of a very limited number of antibiotic formulations in the preceding six months, however in a context where febrile illness is common and presentation to health care facilities is late, and diagnostics are few, this may represent appropriate prescribing. Indeed, this may instead speak to the need to improve access to and quality of health care services, not limit access, particularly for the most economically disadvantaged groups.

The limitations of this study include the potential for social desirability bias, whereby household respondents may state that they recognised or used antibiotics to meet interviewer’s expectations. We mitigated against this by using our innovative and previously validated “drug bag” methodology and experienced field interviewers. The drug bag uses a wide range of antibiotics available in the community; however it is possible that we did not include some important formulations, or that packaging designs have changed, limiting recognition. The concept of an “antibiotic” is not well understood in Malawi and has no specific Chichewa word, potentially hindering accurate recall of use and recognition. We randomly sampled households from a community sampling frame to minimise sampling bias; nevertheless, women were over-represented, perhaps because men may not have been available for interview, or at work. We may therefore have underestimated antibiotic use and recognition. Neighbourhoods were purposively sampled on the basis of availability of sampling frames from previous surveys and demographic/poverty profiles; we did however randomly sample households within neighbourhoods, and this may limit generalisability. We did not collect data on where participants obtained antibiotics from; this should be a priority for future research projects and surveillance.

In conclusion, we found patterns of current and recent antibiotic use and recognition among randomly-sampled household members in Blantyre, Malawi to be limited to a small number of broads-spectrum antibiotic formulations. People known to have poorer access to healthcare reported recognising fewer antibiotics. So-called “antibiotic overuse”–specifically of antibiotics on the watch list–is unlikely to be a major driver of antimicrobial resistance and drug resistance infections in this and similar settings. Rapidly reducing generation of antibiotic resistance and drug resistant infection in low-income settings in Africa will require a shift of focus, away from narratives that blame people in precarious living conditions for “antibiotic overuse” and instead towards holistic approaches that address the underlying systemic drivers of AMR, whilst recognising and supporting antibiotic access within well-functioning health systems.

Acknowledgments

We are grateful to all study participants who took part in this study and to the DRUM Consortium.