Antimicrobial resistance (AMR) is a global challenge with major implications for African countries where severe bacterial infections are common, but access to antibiotics is often limited ^{1, 2} . Recent estimates suggest that the population mortality rate linked to bacterial AMR is highest in sub-Saharan Africa ³ . Bloodstream infections (BSI) caused by Gram-negative pathogens including Enterobacterales are important contributors to neonatal, child and adult morbidity and mortality in Africa ^{3– 6} . High prevalence of 3 ^rd generation cephalosporin (3GC) resistance amongst E. coli and K. pneumoniae causing BSI has been reported from many African countries, often substantially exceeding proportions found in high-income settings ^{3, 7– 10} . Prevalence of resistance to many other antibiotic agents in Enterobacterales in Africa is also thought to be concerningly high.

Many African hospitals lack local microbiological testing services and have relatively limited resources for preventative activities to minimize the impacts of bacterial AMR, including infection prevention and control (IPC) and antimicrobial stewardship. Furthermore, in many African hospitals, the availability of “watch” and “reserve” antibiotic agents ¹¹ for treating resistant infections is limited, though precise data describing day-by-day supply of particular agents are scarce ¹² . We are not aware of any previous comparisons between the availability of antibiotics with actual patient-level usage in hospital settings in Africa.

As part of a multi-centre cohort study across African hospitals designed to measure mortality impact effects associated with bacterial AMR (the MBIRA study), we aimed to describe the baseline circumstances in the participating hospitals in terms of their capacity to treat patients with resistant bacterial bloodstream infections caused by Enterobacterales.

Study setting: Eight public hospitals from eight different sub-Saharan African countries were invited to participate. These were Tygerberg Hospital, Cape Town, South Africa; Korle-Bu Teaching Hospital, Accra, Ghana; National Hospital, Abuja, Nigeria; Kilifi County Hospital, Kilifi, Kenya; University Teaching Hospital, Lusaka, Zambia; Kilimanjaro Christian Medical Centre, Moshi, Tanzania; Hiwot Fana Comprehensive Specialized Hospital, Harar, Ethiopia; Queen Elizabeth Central Hospital, Blantyre, Malawi. These hospitals were purposively selected, primarily on the basis of having well-established microbiology services including performance of blood cultures, with relevant indicators of laboratory quality control measures. Preliminary discussions were held with several other hospitals, but they were not formally invited to participate as they did not meet the inclusion criteria.

Data collection: at the start of the MBIRA study, each site gathered data at the hospital level. This included data on the number of general and intensive care beds in the hospital, assessing the level of intensive care beds with reference to a common international standard ¹³ . We categorized intensive care units from level 1 (lowest) to level 3 (highest) using objective criteria, including therapeutic capacity, personnel and monitoring capacity. Laboratories reported historic blood culture activity levels for 2019 (where available) and prospectively reported total monthly blood cultures performed during the period of active patient recruitment in 2020–22. Laboratories self-assessed which blood culture isolates represented contamination, using standardized recommendations ¹⁴ .

As part of this assessment, we used two externally created tools: the “WHO Infection Prevention and Control Assessment Framework at Facility Level” (hereafter “WHO-IPCAF”) ¹⁵ and the “Antimicrobial Stewardship programmes in health-care facilities in low and middle-income countries – a WHO practical toolkit” (hereafter “WHO stewardship toolkit”) ¹⁶ . The WHO-IPCAF questionnaire self-assesses facility-level IPC activity via questions in eight core components ( Table 2), each scored out of 100 for a maximum possible total of 800. According to the interpretation provided, a score of <200 represents an “inadequate” level of IPC, 200–400 represents a “basic” IPC level, 400–600 represents “intermediate” IPC level and >600 represents “advanced” IPC level. This WHO-IPCAF tool has previously been shown to have adequate internal consistency and good inter-observer reliability ¹⁷ . In practice, one large hospital in Ghana repeated the WHO-ICPAF survey with seven specialists from different departments across the facility; we used the arithmetic mean of the scores obtained as a facility-wide score. All other hospitals used a single scoring with local investigators identifying relevant hospital staff for different questionnaire elements.

The WHO stewardship toolkit includes a series of 31 yes/no questions grouped into six domains ( Table 2); these questions are designed for self-assessment of current implementation of antimicrobial stewardship activities at the facility level. Local investigators identified suitable hospital staff for response to different questions, who provided informed oral consent to participate. We summarized these questions by awarding 1 point for each “yes” response. We are not aware of any published validation studies of this tool, which appears to have been primarily designed for within-facility quality-improvement exercises. Both the WHO-IPCAF and the stewardship tools contain extensive descriptions of the intended approaches to data collection.

During the active period of patient recruitment in the main study, each MBIRA hospital team completed monthly cross-sectional pharmacy surveys on the availability of antibiotic agents suitable for the treatment of bloodstream infections caused by Enterobacterales. The survey was completed on a working weekday at a regular timepoint in each month, with a local pharmacist and a study investigator jointly assessing antibiotic availability by examining the drug stock physically held in the pharmacy at that time. Where a large hospital had more than one pharmacy, the current antibiotic availability in the main or central pharmacy was surveyed. We considered an antibiotic as “available” when any quantity of the drug of any formulation was found to be “in stock”, provided that it was within the manufacturer’s expiry date. We considered an antibiotic as “not available” in a particular month when the drug was reported as “not normally available” or “out of stock” on the particular day of the survey. We identified 11 antibiotics drugs that are all included in the WHO Essential Medicines List Antibiotic Book formulary (either under “Access” or “Reserve” classes of AWARE classification) to gather individual drug availability data. We summarized the results of these repeated surveys by drug and by hospital by taking an arithmetic average of these binary results across all relevant surveys, expressing the result as a percentage. Each hospital also provided a written response to a standardized question about the particular approaches or special circumstances that applied (in the experience of the most senior local clinical investigator) for gaining access to antibiotic drugs for treatment of antibiotic resistant infections, if suitable agents were not immediately available through routine supplies.

For patient-level antibiotic use information, hospital inpatients of all ages with culture-confirmed bloodstream infections (BSI) caused by Enterobacterales bacteria (including E. coli, K. pneumoniae and various other species, but excluding Salmonella spp.) were prospectively recruited as part of the main work of the MBIRA study. These patients (or their relatives for children) provided written informed consent to participate in the study and were followed up whilst inpatients, including recording of day-by-day antibiotic use up to 30 days after BSI onset. Full description of these BSI patients and exploration of the impacts of AMR is planned for a subsequent publication. Here, we plotted the proportion of study months in which these antibiotic drugs were available in the hospital pharmacies (x-axis) against the collapsed proportions of Enterobacterales BSI patients at each hospital over the same period who received at least one drug-day of treatment with the respective drug (y-axis). Data collection on antibiotic use for BSI patients was limited from the day prior to blood culture collection to the cessation of antibiotic treatment or 30 days after blood culture collection, whichever was sooner.

Contextual information on national-level income categories and Gross Domestic Product (GDP) per head were obtained from the World Bank using 2020 data (Purchasing Power Parity, current international $ ¹⁸ ) . Hospital-level and pharmacy-level data were entered into an online REDcap database. Statistical tests were performed using STATA (v16, Statacorp, USA).

Study approvals: Ethical approval was granted by the relevant health research ethics committee at each site as follows:

Permissions to access the laboratory and hospital registers were obtained from the relevant authorities in each hospital.

The eight hospitals participating in the MBIRA study collected the hospital-level data between October 2020 and March 2021. Pharmacy-level data was collected monthly in each hospital for between 8 and 15 months between November 2020 and February 2022, corresponding to the months of active participant recruitment to the main MBIRA study. Antibiotic use data was obtained from 996 patients with confirmed Enterobacterales bloodstream infection in these hospitals.

Hospital-level characteristics ( Table 1). These hospitals were all public-sector tertiary-level hospitals with some prior experience of clinical research. Seven of the eight hospitals were based in major urban areas – either in a city or the national capital. The eight hospitals were distributed across countries which reflect the substantial spread of income-levels across sub-Saharan Africa, ranging from the “upper-middle income” (South Africa) to “low income” (Ethiopia and Malawi). There was substantial variation in the number of intensive care beds available in relation to the size of the hospital. This ranged from no beds meeting an enhanced care definition available at all in a hospital in Kenya, to a modest number of Level 1 and Level 2 beds in hospitals in Tanzania, Ethiopia and Malawi, to more substantial numbers of Level 3 beds in hospitals in Nigeria and South Africa. In terms of blood culturing practices, in 2019 these hospitals all reported performing more than 100 blood cultures/1,000 patient admissions, where suitable data was available to calculate this metric. In 2021, the hospitals in Malawi and South Africa performed the largest absolute number of blood cultures, both averaging over 1,000 cultures per month. One hospital in Nigeria performed a relatively low number of blood cultures in 2021 (average of 90/month) – this was partly due to two prolonged national medical strikes in Nigeria that year. All other participating hospitals performed an average of more than 100 blood cultures per month in the MBIRA study period. Blood culture contamination rates varied widely, from 3.6% to 25%.

WHO-IPCAF toolkit ( Table 2). All hospitals self-completed the WHO-IPCAF questionnaire. The median total WHO-ICPAF score for these facilities was 428/800 with a range of 155 to 687.5 ( Table 2). Across all the hospitals, the two lowest-scoring IPC core components were “Hospital-acquired infection surveillance” (mean score 34/100) and “Workload, staffing and bed occupancy” (mean score 37/100). We found no evidence of a relationship between World Bank national income level categories and the hospital-level WHO-IPCAF scores (Kruskall-Wallis equality-of-populations rank test, p=0.83).

WHO stewardship toolkit ( Table 2). All hospitals completed responses for the 31 yes/no questions in the WHO stewardship toolkit. The range of self-assessed total scores varied very widely (2 to 23), with a median of 14.5 “yes” responses. Amongst the different categories, the “Leadership commitment” and “Reporting and Feedback” questions were least likely to have “yes” responses and the “Monitoring and feedback” questions were most likely to have “yes” responses. We found no evidence of an association between national income level and facility WHO stewardship toolkit overall score (Kruskall-Wallis equality-of-populations rank test, p=0.64). However, there was strong evidence of association between the hospitals’ scores for the WHO-IPCAF questionnaire and the WHO stewardship toolkit (Spearman’s rank correlation coefficient, p<0.01).

Cross-sectional antibiotic availability surveys ( Table 3). In terms of local availability of antibiotic drugs suitable for treatment of invasive Gram-negative infections, all the hospitals completed at least 8 monthly cross-sectional surveys (range 8–15) in their main hospital pharmacies. We found a marked variation in the availability of these antibiotic agents across the eight hospitals - see Table 3. In one hospital in South Africa (the only upper-middle income country in the study), all of these 11 agents were available in all rounds of the cross-sectional survey. In all of the other hospitals, there were one or more antibiotic agents that were only intermittently available, meaning that the drug was recorded as “not available” in at least one monthly cross-sectional survey. This phenomenon was especially marked in the study hospitals in Malawi (8 agents with intermittent availability) and Zambia (7 agents with intermittent availability). Finally, there were various antibiotic drugs that were “never available”, meaning that none of the cross-sectional surveys reported availability of that particular agent in that hospital. This occurred most commonly in the study hospitals in Ethiopia (3 agents never available) and Tanzania (2 agents never available).

We found that the antibiotics ceftriaxone and ciprofloxacin were the most consistently available antibiotic agents across these eight hospitals – these were available in almost all cross-sectional surveys in all these hospitals (both >95% availability across all survey rounds). The antibiotics amoxicillin, co-amoxiclav, gentamicin and co-trimoxazole were all found to be widely, but not universally, available (all available 70% to 90% across all survey rounds). The antibiotics piperacillin-tazobactam, meropenem, amikacin, cefotaxime and chloramphenicol were only partially available across these hospitals (all available 40% to 65% across all survey rounds). Few hospitals (2/8) described any availability of other β-lactam-β-lactamase-inhibitor agents or colistin for treating resistant Gram-negative infections.

All hospitals provided a description of the local special circumstances for accessing antibiotics for treating resistant infections – Table 4. We identified three common themes in these responses: 1) patients buying antibiotics from external private pharmacies using either out-of-pocket payments or personal/private insurance (mentioned at 7/8 hospitals); 2) additional provision of antibiotic access through national health insurance (mentioned at 3/8 hospitals); 3) other access routes using external funding such as charitable donations or research-associated supply (mentioned at 2/8 hospitals). Only the South African hospital described rarely experiencing any antibiotic non-availability, whilst the hospital in Malawi emphasized that difficulties in antibiotic access were “commonplace”.

In a scatter plot of the in-pharmacy availability of antibiotic drugs against percentage of local BSI patients with recorded use of each drug ( Figure 1), we found that for most antibiotics in most hospitals, there was a correlation between measured pharmacy availability and actual use for treating BSI patients. When specific antibiotics were infrequently available in hospital pharmacies, they were rarely used for treating patients with BSI. There were a small number of exceptions to this broad pattern, most notably the usage of meropenem in the hospital in Tanzania and of amikacin in the hospital Malawi, which were both commonly used despite rarely being available in the hospital pharmacy.

We found some variation in contextual circumstances for treatment and prevention of antibiotic resistant infections across eight public-sector hospitals in different African countries, across the range of national income levels in sub-Saharan Africa. Although these were all tertiary hospitals with some research orientation and mainly located in major cities, the levels of institutional capacity for dealing with antibiotic resistant infections varied markedly both at institutional level (IPC and antibiotic stewardship activities) and at individual-patient level (access to relevant “watch” and “reserve” antibiotics). We discuss both common themes and areas of variation.

Using the WHO-IPCAF tool, we found a median self-assessed facility-level score of 428/800, which corresponds to an “intermediate” overall level of IPC compliance, according to the developers of this tool. There was a wide variation in the self-assessed score achieved from these eight hospitals (range 155 to 687.5), but there were common themes of poor performance in particular domains (‘Healthcare associated infection (HAI) surveillance’ and ‘Workload, staffing and bed occupancy’). We compared our findings to other studies that used the WHO-IPCAF tool. A recent major study described use of this WHO tool in over 4,400 hospitals worldwide ¹⁹ . In that study, the weighted median (IQR) IPCAF score for 595 hospitals in Africa was 415/800 (290-582), with “HAI surveillance” also identified as an area of notable weakness (median score 32/100). These are extremely similar findings to our own results, suggesting that in terms of IPC activities, our 8 hospitals are broadly similar to hospitals across the continent. In terms of other published reports, a study of five hospitals in Pakistan reported a median WHO-IPCAF score of 117.5/800, with all hospitals scoring <200/800, representing substantially worse IPC scoring than we found. The authors attributed these low scores to non-existent IPC guidelines and HAI surveillance ²⁰ . A large study from Germany ²¹ included more than 700 hospitals and reported a median WHO-IPCAF score in the “advanced” level (677/800) of IPC compliance.

The low WHO-IPCAF scores found in our study might be partly explained by the resource constraints experienced in most public hospitals in Low- and Middle-Income Countries (LMIC), as others have found ¹⁹ . However, we did not find an association between the WHO-IPCAF scores and national-level income data; the highest WHO-IPCAF score was from a hospital in Tanzania, whereas the only hospital from an “upper-middle income” nation (South Africa) had a score close to our median value. This suggests it is not purely financial limitations that restrict the performance of IPC – it appears some African hospitals are able to implement significant IPC activities within limited budgets. As others have also found ¹⁹ , the range of IPCAF scores amongst these African hospitals was much wider than the range seen amongst much larger numbers of European hospitals in either Germany ²¹ or Austria ²² . This suggests that across sub-Saharan Africa, hospital IPC activities are typically both less advanced and much more variable than within high-income nations.

The variation in the responses to the WHO stewardship toolkit were extreme with “yes” responses to the questions ranging from 2 to 23/31. The median score (14.5 “yes” responses) was less than 50% of the possible maximum score. This suggests antimicrobial stewardship activities in these eight hospitals were limited and highly variable. However, this tool was not designed with between-facility comparisons in mind but was rather intended to support national and facility-level development of stewardship activities through an iterative process of self-assessment. We have not found any published examples of multi-site studies using these questions from this WHO toolkit in this way.

The variation found in access to antibiotic agents through repeated cross-sectional surveys was substantial, ranging from universal access to all 11 drugs surveyed (in one hospital in South Africa) to intermittent access to the majority of these drugs (in the hospitals in Ghana, Malawi and Zambia) to a consistent-but-limited pattern of antibiotic access (in one hospital in Ethiopia). No two hospitals displayed the same pattern of results in these cross-sectional antibiotic access surveys, though the antibiotics ceftriaxone and ciprofloxacin were consistently available in all hospitals. We are not aware of any previous research that has investigated the availability of agents for treating invasive Enterobacterales infections in LMIC in this way. There was limited access to agents such as piperacillin-tazobactam, amikacin and meropenem that are all vital drugs for treating resistant Enterobacterales infections. As emphasized in a recent review, access to antimicrobials in LMIC hospitals is a serious issue ²³ . Intermittent drug availability in these tertiary hospitals is highly concerning as it suggests that health systems across the African continent are ill-equipped to treat severe antibiotic-resistant infections.

Apart from one hospital in South Africa, investigators at all hospitals described that patients were given prescriptions to buy antibiotics from private pharmacies outside the hospital using out-of-pocket expenditure. Additionally, for hospitals in Tanzania, Zambia and Ghana, patients could access some additional antibiotics from external private pharmacies for free if they were enrolled in the national insurance schemes. However, when comparing pharmacy drug availability against the actual use of antibiotics by patients with confirmed Enterobacterales BSI in these institutions, we found that there was a clear correlation between pharmacy availability and use of drug in individual patients. This suggests that in practice, external purchase of antibiotics for treating drug-resistant BSI took place relatively rarely in most of these hospitals during the study period.

This study has several notable strengths through use of original approaches. This is, to the best of our knowledge, the first study to report use of the WHO stewardship toolkit from hospitals in multiple sub-Saharan African countries. We are not aware of any prior publication describing availability of antibiotic agents for treating infections in repeated pharmacy-based surveys in hospitals across multiple sub-Saharan African nations, nor of any prior work comparing pharmacy availability against patient-level use in this region. Infection control and access to antibiotics are both subjects of intense international interest as attention turns to how we can avert future deaths from bacterial AMR.

In terms of study limitations, firstly, these eight hospitals are unlikely to be representative of all hospitals in sub-Saharan Africa. These hospitals are mainly located in major urban areas with well-developed microbiology services and some participation in research studies. The hospitals in Kenya and Malawi both have long-standing relationships with international research organizations, unusual amongst African public-sector hospitals. This introduces a selection bias to our sample, which we believe is likely to give an overly-optimistic picture of services across the region. Secondly, while these data were collected by senior healthcare staff with research experience using standardized materials, we remain concerned about the possibility of response bias (in particular, a social-desirability bias) in these self-reported questionnaires, especially the WHO stewardship toolkit. This form of bias could lead to either over-optimistic or over-pessimistic responses, in comparison to an idealized external assessor.

Across eight tertiary hospitals in sub-Saharan Africa, we found wide variation in the intensive care facilities, the self-assessed IPC activities, stewardship activities and the objectively-assessed profile of access to antibiotic drugs for treating invasive Enterobacterales infections. This means that the same, or highly similar, antibiotic resistant pathogens could plausibly have very different human-level impacts between these hospitals. The ongoing MBIRA study is gathering patient-level outcome data from these same eight hospitals combined with individual-level antibiotic use data for treatment of Enterobacterales BSI. We hope that these and further data will provide the basis to better understand the human impacts of antibiotic resistance across sub-Saharan African nations.

This study is published with the permission of: the Director of KEMRI, Kenya; NIMR, Tanzania; and NHRA, Zambia