Study design and setting

A cross sectional convergent mixed-methods study was conducted between September and November 2025, on the adult surgical wards of MNRH in Kampala, Uganda. It consisted of a point prevalence survey of inpatient antibiotic use (23^rd Sep-3^rd October 2025) and a retrospective analysis of microbiology data for patients admitted to MNRH surgical wards from June 2023 to August 2025. Qualitative IDIs (23^rd Sep to 30^th Nov 2025) were also conducted, exploring healthcare worker (HCWs’) perceptions and lived experiences on AMR, antibiotic prescriptions and antimicrobial stewardship practices. At the time of this study, antibiotic prescriptions both pre- and post-operatively were based on clinician discretion, antibiotics available in hospital pharmacy and the 2023 Uganda Clinical Guidelines (UCG) [9].

Study participants

The point prevalence survey targeted adult surgical patients present on the ward by 8 am on the morning of the survey and received an antibiotic within the last 24 hours. The microbiological data analysis focused on results of specimens collected from the surgical wards and processed at the MNRH Microbiology Laboratory or the Makerere University Clinical Microbiology Laboratory (MUCML) between June 2023 and August 2025. Results that had critical or high priority pathogens including; Acinetobacter baumannii, Pseudomonas aeruginosa, Staphylococcus aureus, Escherichia coli, Proteus mirabilis, Klebsiella pneumoniae and Citrobacter freundii were extracted [10].

The IDIs focused on HCWs in the surgical wards including; Surgical specialists, Resident doctors, Medical Officers, Pharmacists or Nurses. HCWs had to have worked on the ward for at least one month by the time of study start and were present on the ward at least 3 days each week to actively engage in prescription, dispensing or drug administration while on duty. We excluded HCWs who were unable to be physically present for the interviews.

Study procedures

AMR survey

Microbiological data (sample type, isolate identified and susceptibility profiles) was extracted along with patient identification, ward, age and sex from the electronic laboratory information systems of MNRH Microbiology and MUCML laboratories and captured it onto the WHO PPS microbiology data form. Both laboratories utilized automated incubation and monitoring for blood cultures using BD BACTEC FX40 system (Becton Dickinson and Company, Sparks, MD, USA). For positive blood cultures and majority other specimens collected, they used incubation on solid media, microscopy and biochemical techniques for isolate identification and then antimicrobial susceptibility testing using the Kirby Bauer disc diffusion method according to the Clinical and Laboratory Standards Institute (CLSI) guidelines 2023–2025 [11–13]. Both laboratories also utilized identical laboratory standard operating procedures for culture and sensitivity, handling contaminants, and utilized known American Type Culture Collection (ATCC) controls among other quality control techniques in the culture and sensitivity process. Records of patients aged 18 + , who had been admitted to the surgical wards were included. The extracted data was entered into Microsoft Excel and cleaned. Missing data entries were cross-checked from the corresponding laboratory workbooks and entered into a Microsoft Excel sheet.

HCWs IDIs

An open-ended interview guide with probes developed by the authors, was used to guide the IDIs using key thematic questions relevant to the study objectives. It also consisted of prompts designed to explore participants’ perceptions and experiences. The interview guide was pretested on five participants similar to the study population. Eligible participants were approached in-person and provided with detailed information about the study and informed consent forms. Participants were interviewed by a female study staff (DA (Nurse) in a quiet study room, at their convenient time for 45–60 minutes. Interviewing study staff had undergone prior training on qualitative methodology and data collection and didn’t have any relationship with participants or interest in research topic. All interviews were audio recorded securely stored on password-protected devices. Notes were used to document major themes arising during the interviews. Interviews were transcribed verbatim (FMM, DA) in English language and cross-checked by the principal investigator and qualitative specialist (FMM, AT) for accuracy.

Statistical methods

Point Prevalence Survey (PPS) data analysis

Antibiotic use patterns constituted the primary unit of analysis and were linked to individual patients. Patient-level characteristics were summarized using frequencies and percentages for categorical variables, and medians with interquartile ranges (IQR) for continuous variables. The prevalence of antibiotic use was defined as the proportion of surveyed patients receiving at least one systemic antibacterial agent on the day of the survey. Antibiotic use patterns were described overall, and key indicators included the median number of antibiotics prescribed per patient and the distribution of prescriptions according to the WHO Access, Watch, and Reserve (AWaRe) classification. Antibiotic classes and individual agents were summarized using frequencies and percentages. Prescribing quality indicators recommended in the WHO PPS protocol were assessed, including compliance with Uganda Clinical Guidelines 2023 (UCG), and the duration of SAP. Guideline compliance was defined and calculated as the proportion of antibiotic agent choices made as indicated in the UCG (antibiotic choice only). All PPS indicators were presented overall and stratified by prescriber cadre, and AWaRe category where appropriate. All statistical analyses were conducted using R version 4.2.2 and STATA 19.

Antimicrobial Resistance (AMR) data analysis

Antimicrobial resistance (AMR) data was summarized as the percentage of each species of organisms that were non-susceptible (classified as resistant or intermediate on susceptibility testing) to each test antibiotic as per CLSI guidelines [16]. Non-susceptibility patterns were presented graphically to highlight antimicrobial agents with high prevalence of non-susceptibility. All statistical analyses were conducted using R version 4.2.2 and STATA 19. To prevent data inflation from repeat sampling, a strict de-duplication protocol was applied: only the first isolate per patient per unique bacterial species within the same clinical episode was included. Duplicate isolates of the identical species from the same patient were excluded. Organism-antimicrobial combinations with intrinsic resistance or without established interpretive criteria were either excluded from analysis or reported as non-recommended (NR) in accordance with CLSI guidelines.

Qualitative data analysis

Thematic content analysis was conducted as per the six-phase framework by Braun and Clarke (2006) [17]. After transcribing the audio recordings and typing them in word documents, we (FMM, DA, AT) conducted a thematic analysis using a systematic, multi-step approach to ensure depth and rigor. First, immersive reading of all interview transcripts was utilized, to become thoroughly familiar with the data, carefully marking significant statements and writing analytic memos to capture emerging ideas and reflections. In the second step, line-by-line open coding was performed (FMM, AT) on 25% of the selected transcripts to inductively generate an initial set of codes. These preliminary codes were developed manually on paper and used to construct a coding framework that guided subsequent analysis. During the second cycle of coding, all transcripts were imported into NVivo version 12 to facilitate systematic data organization and retrieval. The codes were then reviewed, compared, and grouped into broader categories based on patterns, similarities, and conceptual linkages, leading to the development of overarching themes. Finally, representative quotations corresponding to each theme were identified, extracted, and organized to support and illustrate the findings presented in the results narrative.

Study variables

The primary PPS outcome was prevalence and patterns of antibiotic use. Prevalence of antibiotic use was defined as the proportion of patients receiving at least one systemic antibacterial agent within the last 24 hours. The unit of analysis was the antibiotic prescription, linked to individual patients. Key variables included ward-specific antibiotic prevalence, median number of antibiotics per patient, antibiotic class, individual agent, route and indication (therapeutic or prophylactic). Antibiotics were categorized using the WHO AWaRe classification. Prescribing quality indicators included compliance with UCG, proportion of SAP, duration of prophylaxis (≤24 hours versus >24 hours) and cadre of prescribing HCW.

The primary AMR survey outcome was non-susceptibility, defined as isolates classified as intermediate or resistant to each antibiotic according to CLSI criteria. Each laboratory-confirmed isolate represented a unique observation. Variables included specimen type, organism identified, antimicrobial agents tested, and susceptibility results. Non-susceptible proportions were calculated for each organism-antibiotic combination as the proportion of non-susceptible isolates among those tested, in line with WHO GLASS standards.

Integration of PPS, AMR and IDI results

Integration of quantitative and qualitative findings was performed using a convergent mixed-methods approach. Quantitative (PPS and AMR surveys) and qualitative IDIs data was collected and analysed independently and then brought together for interpretation. Findings were compared, contrasted and triangulated to assess areas of convergence, complementarity and divergence. Quantitative PPS patterns were examined alongside qualitative explanations to understand underlying drivers of prescribing behaviour. Patterns of antibiotic use were interpreted considering AMR data to determine whether prescribing aligned with local susceptibility profiles. Where discrepancies arose, qualitative findings were used to explore contextual and systemic explanations. A joint display matrix was developed to align key quantitative indicators with corresponding qualitative themes.

Ethical considerations

This study was conducted according to the Declaration of Helsinki, Good Clinical Practice Guidelines and guidelines by the Uganda National Council of Science and Technology. Ethical approval was sought from the Research and Ethics Committee of Makerere University School of Public Health (SPH-2025–899) and administrative clearance was obtained from MNRH in Uganda (MHREC 2967). This study was registered with the Uganda National Council of Science and Technology (HS6321ES). A waiver of informed consent was obtained for the cross-sectional aspect of the study as all procedures involved data abstraction from patient files and there was no interaction with patients whatsoever. For the qualitative aspect, only HCWs who provided voluntary written informed consent were enrolled on the study. HCWs enrolled on the study received a time compensation of USD 15 after each interview.

PPS results

Characteristics of study population.

Among the 303 surgical inpatients present on the wards during PPS, 193/303 (63.7%) (95% CI: 58.0–69.12) had at least one antibiotic prescribed and were included in the analysis, of these, 125/193 (64.8%) were male (Table 1). The median age was 40 years (IQR 27–58) with 106/193 (54.9%) being 18–44 years of age. A total of 135/193 (70.0%) had undergone surgery since current admission, 133/193 (68.9%) had peripheral vascular catheter in situ and 62/193 (32.1%) had comorbidities.

Download:

• PNG
  larger image
• TIFF
  original image

Table 1. Sociodemographic and clinical characteristics of the PPS participants.

https://doi.org/10.1371/journal.pone.0352983.t001

Prevalence of antibiotic use

The median number of antibiotics prescribed per patient was one (IQR:1–2) with 109/193 (56.5%) patients being on one antibiotic and 78/193 (40.4%) on two antibiotics. Of all the prescriptions reviewed, only 136/281 (48.4%) were compliant to UCG (Fig 1).

Download:

• PNG
  larger image
• TIFF
  original image

Fig 1. Prevalence of antibiotic use, AWaRe and antibiotic agents used.

https://doi.org/10.1371/journal.pone.0352983.g001

Antibiotics prescribed

281 antibiotics were prescribed in total to 193 patients, with 89/281 (31.7%) (95% CI: 26.27–37.46) being ceftriaxone, 66/281 (23.5%) (95% CI: 18.66–28.89) metronidazole and 24/281 (8.5%) (95% CI: 11.61–20.44) being levofloxacin (Fig 1). Of all prescriptions, 152/281 (54.1%) (95% CI: 48.08–60.08) were in the Watch category and 129/281 (45.9%) (95% CI: 39.9–51.9) were in Access category. Antibiotics were prescribed for SAP 256/281 (91.1%), with 239/256 (93.4%) of these SAP antibiotics given as multiple doses beyond first 24 hours (Table 2). Prescribing was done by Resident doctors 120/281 (42.7%) and Surgical specialists 79/281 (28.1%). Directed antibiotic therapy was at 10/281 (3.6%).

Download:

• PNG
  larger image
• TIFF
  original image

Table 2. Summary of antibiotic use characteristics.

https://doi.org/10.1371/journal.pone.0352983.t002

AMR results

Among the 1,879 microbiology tests conducted between June 2023 and August 2025, 480/1879 (25.5%) were conducted on adult surgical inpatients and were included in this study (S1 Table). The median age of the participants was 38 years (IQR: 28–53.5), with 304/480 (63.3%) being 18–44 years and male 329/480 (68.5%). The sample types processed for culture and sensitivity testing included pus 335/480 (69.8%) and urine 36/480 (7.5%) among others (S2 Fig).

Bacterial pathogens Isolated on culture

Of the 482 bacterial pathogens isolated from 480 specimens, 151/482 (31.3%) (95% CI: 27.35–35.60) was Escherichia coli, followed by Staphylococcus aureus at 133/482 (27.6%) (95% CI: 23.79–31.75), Citrobacter freundii at 73/482 (15.1%) (95% CI: 12.22–18.62) and Klebsiella pneumoniae at 67/482 (13.9%) (95% CI: 11.10–17.28) (Fig 2).

Download:

• PNG
  larger image
• TIFF
  original image

Fig 2. Bacterial pathogens isolated on culture across sample types (N = 482).

https://doi.org/10.1371/journal.pone.0352983.g002

*Other = other sample types include: ascitic, pleural, joint and endocardial fluid aspirations.

Drug susceptibility profiles of isolated bacterial pathogens

Non-susceptibility to individual antibiotics was observed across all isolates. Escherichia coli showed 142/151 (94.0%) non-susceptibility to Ceftriaxone, 124/151 (82.1%) to Cefuroxime and Ciprofloxacin, 113/151 (74.8%) to Cefepime, 81/151 (53.6%) to Amoxicillin-Clavulanic acid, 32/151 (21.2%) to Imipenem and 9/151 (6.0%) to Meropenem (Table 3). Citrobacter freundii showed 66/73 (90.4%) non-susceptibility to cefuroxime, 63/73 (86.3%) to Ceftriaxone, 61/73 (83.6%) to Ciprofloxacin and 22/73 (30.1%) to Imipenem. Klebsiella pneumoniae showed 63/67 (94.0%) non-susceptibility to Cefepime, 61/67 (91.0%) to Cefuroxime, 59/67 (88.1%) to Ceftriaxone, 53/67 (79.1%) to Ciprofloxacin, 41/67 (61.2%) to Amoxicillin-clavulanic acid, 25/67 (37.3%) to Imipenem and 15/67 (22.4%) to Meropenem. Staphylococcus aureus showed 44/133 (33.1%) non-susceptibility to Cefoxitin, 109/133 (82.0%) to Ciprofloxacin, 71/133 (53.4%) to Gentamicin and 77/133 (57.9%) to chloramphenicol.

Download:

• PNG
  larger image
• TIFF
  original image

Table 3. Antibiogram showing percentage susceptibility across bacterial pathogens (%S=(S)/(I + R + S)*100).

https://doi.org/10.1371/journal.pone.0352983.t003

Qualitative results

We screened 49 HCWs and enrolled 29 including: 3 specialist surgeons, 12 resident doctors, 2 pharmacists and 12 nurses into this arm of the study (S2 Table). Eighteen were male and median duration in service was 4.0 years (IQR: 2.5–5.0). 20 eligible HCWs declined participation in the study as they couldn’t afford a one-hour sit-down for the IDIs. All IDIs were conducted once without repetition. Three major themes arose from the qualitative data analysis: 1) Determinants of antibiotic use, 2) Perceptions on AMR and 3) perceptions on AMS.

Determinants of antibiotic use

Across IDIs, several respondents (10/29) described prescription practice as largely driven by experience, training, and drug availability rather than standardized protocols. Many respondents explicitly stated that no formal, written, ward specific antibiotic use protocols existed, even where national guidance was acknowledged, implementation was uncertain. In the absence of local standard operating procedures (SOPs), prescribing relied on inherited norms based on school experience and personal knowledge:

“...I wouldn’t say there’s a written guideline… It’s not there”. IDI-Lgit-029. “They are the Uganda clinical guidelines… I really don’t know if we really go in so much to follow the step-by-step prescription they are suggesting”. IDI-Thor-001. “...We use what we know works based on experience and knowledge that has been passed down to us”. IDI-Lgit-029

Clinicians relied on clinical judgment and experience to escalate to broad-spectrum antibiotics for last resort antibiotics like meropenem or vancomycin when severe infections or high-risk patients were encountered, even if this departed from UCG.

“Sometimes you find the patient is having a very bad infection… you start with an antibiotic which is high in the ladder because you think that if you don’t start it right early, we may lose the patient. Even some patients, we start… meropenem or vancomycin.” IDI-Neur-018

Broad-spectrum antibiotics, particularly ceftriaxone, metronidazole, meropenem, and levofloxacin were reported by majority respondents (24/29) as frequently used for both prophylactic and therapeutic purposes in the adult surgical wards of MNRH, grounded in “we know what works, based on experience and knowledge that has been passed down to us” IDI-Lgit-029. These were chosen because of their availability, affordability, proven effectiveness, and ease of administration.

“The most commonly prescribed are cephalosporins, that is, ceftriaxone, cefixime, and then metronidazole. We also prescribe ciprofloxacin and levofloxacin, mainly because those are the most available, and at times, those are the ones that the bacteria respond to”. IDI-Ugit-028

Ward environment lapses like overcrowded patient areas, unhygienic conditions and past experiences with infections drove clinicians to use antibiotics more broadly and pre-emptively, even when UCG suggested that single prophylactic use would suffice. Prophylaxis was framed as standard surgical principle and justified by anticipated contamination.

“...in our setting, we have a fear that the patient’s hygiene is poor, the environment where we work is not very safe, sometimes we work in theatres that are very hot, people are sweating, because we fear that the risk of wound infection is high. We tend to always put the patients on antibiotics when actually you feel that in an ideal situation, they don’t need antibiotics… for fear that in case there’s a contamination.” IDI-SHO-Orth-013

Delays in conducting culture and sensitivity sometimes up to two weeks further complicated timely directed therapy and prior antibiotic exposure from previous hospital admissions often rendered commonly used antibiotics to be perceived by majority clinicians (26/29) as ineffective.

“Actually, on that ward, they normally don’t perform culture and sensitivity, it’s common. So, you find a patient, these patients are always in hospital for three months and above, they usually get hospital acquired infections, so you give the patient an antibiotic for a week, seven days later, no change and the patient is deteriorating. That’s when they decide to do a culture.” IDI-Throc-003

Availability of free services also shaped prescription choices. Clinicians stated that even when tests like cultures and drugs were “...entirely free of charge,” IDI-Phar-Neur-027, the supply was inconsistent and the cost of antibiotics thus limited optimal prescribing. The price of antibiotics influenced which drugs were chosen, sometimes favouring cheaper but less ideal options.

Perceptions on AMR

Across IDIs, several respondents (26/29) narrated antimicrobial resistance (AMR) as a persistent and escalating clinical challenge marked by treatment failure, limited antibiotic options, financial strain, and difficult prescribing decisions. Respondents described repeated antibiotic failures. Ineffectiveness of antibiotics due to resistance was frequently associated with prior antibiotic use. Respondents described prolonged empirical treatment before culture guidance. Some (6/29) HCWs narrated that patients reportedly arrived after multiple prior antibiotic exposures: “They come… when they have already had a cocktail of antibiotics” IDI-NO-Urol-010. Multidrug resistance among elderly patients was described as severe: “All drugs are almost ineffective” IDI-SHO-HDU-017. AMR was also perceived by many HCWs (19/29) to result in delayed recovery and stunted wound healing. Clinically, resistance led to treatment failure, prolonged infections, complications, and preventable deaths. Respondents recounted cases where patients failed to respond to multiple antibiotics and died. Resistant infections often forced clinicians to escalate to costly “big gun” antibiotics, sometimes without culture results.

“...Sometimes you find we have gone for some big antibiotic… you choose to save the patient… but you are risking the future.” IDI-Neur-018

HCWs also acknowledged that hospital-acquired infections and poor adherence to infection prevention and control (IPC) practices, such as inadequate hand hygiene or improper glove use, facilitated the spread of resistant organisms (super bugs) from patient to patient.

“If I interact with a patient with a bug and I don’t do proper hand washing… transfer the infection… there’s an emergence of more resistant bugs due to spread from patient to patient.” IDI-SHO-Trau-007

Perceptions on AMS

Several respondents (28/29) narrated basic awareness of core AMS practices, particularly rational prescribing, correct dosing, culture-guided therapy, and avoidance of unnecessary antibiotics. However, understanding was often individualized and experiential rather than institutionalised. Across Respondents, antimicrobial stewardship (AMS) was generally understood as the responsible, rational, and evidence-based use of antibiotics to prevent resistance, though the depth and clarity of understanding varied. Many respondents described stewardship in terms of vigilance and responsibility. One consultant noted it involves;

“...being keen and observant and responsible to ensure you prevent resistance… avoid unnecessary prescription… monitor… do culture and sensitivity” IDI-Cons-Lgit-029.

While some Respondents acknowledged the presence of a hospital-based formal antimicrobial stewardship (AMS) subcommittee, awareness, orientation, and operational clarity varied across cadres. Formal stewardship structures were perceived as limited or inactive. At consultant level, there was recognition of a formal stewardship presence within the hospital; at management level, but not on the ward. One consultant acknowledged a hospital-level team, but noted, “in the unit we don’t have a specific team.” IDI-Cons-Lgit-029 and another reported, “...I have not been oriented to one [AMS]...”. IDI-NO-LGIT-014. One respondent also stated, “I know it is there”, and described its expected role.

“...being at the forefront of determining what bugs are available… monitoring the organisms… championing standard operating procedures… [and] giving feedback to all the teams.” IDI-CONS-UGIT-026.

Integration of quantitative and qualitative results

Quantitative results showed a prevalence of antibiotic use, of antibiotics in the Watch (54.1%) category compared to the Access category (45.9%), with Ceftriaxone (31.7%), being the most prescribed. SAP was given using these antibiotics, as multiple doses beyond first 24 hours (93.4%). The commonest bacterial pathogens isolated from among surgical patients were Escherichia coli (31.3%) and Staphylococcus aureus (27.6%), which pathogens had 80–100% non-susceptibility to Ceftriaxone.

Qualitative data explained the factors affecting prescription patterns: there was no ward SOP on antibiotic prescription, the hospital environment was overcrowded and clinicians perceived IPC practices to be inadequate. Rational antibiotic use was mainly determined by availability of drugs, cost and turnaround time of microbiology tests and drugs, lack of structured, documented guidance on antibiotic choice and lack of knowledge on AMR and proper AMS practices. Some clinicians understood AMR and AMS but they were not aware of the existence of the hospital level antimicrobial stewardship subcommittee or its roles. There were no ward specific personnel for AMS (S3 Table).