Clinical profiles of patients with antibiotic-resistant infections at select tertiary hospitals in Uganda: a retrospective study of clinical records Authors: Conrad Tumwine, Flavia Dhikusooka, Reuben Kiggundu, Joseph Musaazi, Ronald Galiwango, Elly Nuwamanya, Mackline Hope, Alex Andema, Nathan Onyachi, Andrew Odur, Stephen Obbo, Peter Mukobi, Deus Twesigye, Francis Oriokot, James Elima, Alfred Yayi, Ben Watmon, Filbert J Nyeko, Charles Olaro, Dathan M Byonanebye, Francis Kakooza, Andrew Kambugu, Jonathan Mayito Published: 2026 ABSTRACT AbstractBackgroundAntibiotic resistance (ABR) remains a critical global health threat, with disproportionate burden in low-resource settings.MethodsThis retrospective study analyzed the clinical records of 599 patients across nine Uganda tertiary hospitals between March 2023 and April 2024. We described ABR rates and used multivariable mixed effects modified Poisson regression to determine the factors associated with ABR. We defined non-multidrug-resistance (non-MDR) as resistance to less than three antibiotic classes, MDR as resistance to at least three antibiotic classes, extremely drug-resistance (XDR) as resistance to at least one antibiotic in all but two or less antibiotic classes and pan drug resistance (PDR) as resistance to all antibiotics classes.ResultsThe 325 samples from patients with culture confirmed infections yielded 378 pathogens. Of the pathogens tested against three or more antibiotic classes (362), 126/362 (34.8%) were non-MDR and 236/362 (65.2%) MDR. Among the MDR, 165 (47.3%) were XDR, and 18 (5.2%) PDR. The prevalence of ABR was highest on surgical 168 (92%), maternity 30 (91%), and ICU 24 (89%) wards. Regression analysis identified ward type as a key predictor: surgical and “other” wards (gynecology, private, and orthopedic) had a 22% (PR, 1.22, 95% CI 1.09–1.37, p < 0.001) and 26% (PR, 1.26, 95% CI 1.05–1.52 p = 0.016) higher ABR risks, respectively. Antibiotics prescription on admission was associated with a reduced risk for ABR while age, sex, referral status, and Human Immunodeficiency Virus status were not associated with ABR. Overall ABR associated mortality was 12.7%, higher among inpatients admitted for shorter hospital days (median 8 vs. 20 days, p < 0.001), intubated patients, and those admitted to the ICU. Patients with resistance stayed in the hospital five days longer than those with no resistance. Escherichia coli, Klebsiella spp. and Staphylococcus aureus, the most isolated pathogens, were highly resistant to commonly used antibiotics, including extended-spectrum beta-lactamase (ESBL) and methicillin-resistant Staphylococcus aureus (MRSA) phenotypes.ConclusionThese findings highlight the clinical impact of ABR in Uganda, especially in high-risk wards such as the ICU and among intubated patients. Strategies for strengthening antibiotic stewardship, infection prevention and diagnostics to combat ABR in Uganda are urgently needed.Trail registrationN/ASupplementary InformationThe online version contains supplementary material available at 10.1186/s12879-026-13753-1.Keywords: Antimicrobial, Antibiotic resistance, Antimicrobial Stewardship, Infection prevention and control, Sub-Saharan Africa, Hospital acquired infection Introduction Bacterial antimicrobial resistance (AMR) accounted for 1.14 million deaths in 2021 [1], a slight decline from 1.27 million in 2019 [2], indicating that bacterial AMR is among the most critical global health threats. The estimates further show that AMR-attributable deaths will reach 1.9 million annually by 2050 if the trend is left unabated [1]. While the AMR threat is global, it is disproportionately borne by low- and middle-income countries (LMICs). In 2019, of the 1.27 million deaths attributable to AMR, 250,000 were from the WHO African region, with western sub-Saharan Africa having the highest overall all-age mortality at 27 × 3 deaths per 100 000 [3]. In the same year, Uganda experienced 7,100 deaths attributable to AMR, ranking 40th among 204, with the highest age-standardized mortality rate per 100,000 people [4] Antimicrobial resistance is usually associated with severe disease with the need for admission to intensive care units (ICUs), high mortality, increased average length of hospital stay (ALOS), and high disability adjusted life years (DALYS) [5]. Infectious syndromes such as blood stream infections (BSIs) are notorious by nature, with BSIs caused by drug-resistant organisms resulting in worse clinical outcomes [6]. For example, carbapenem-resistant Klebsiella spp. and methicillin-resistant Staphylococcus aureus (MRSA) BSIs have 9- and 2-fold higher mortality rates than susceptible pathogens do, whereas BSIs caused by vancomycin-resistant Enterococci (VREs) are more likely to be admitted to the ICU and have longer ALOSs and higher treatment costs [7–9]. These hospital-acquired infections (HAIs), which are an increasing concern in Uganda, lead to avoidable deaths and hamper the delivery of basic and specialized medical treatment, such as chemotherapy and high-risk surgeries and transplants [10, 11]. The clinical characteristics and outcomes of infectious syndromes, including those caused by drug-resistant pathogens, are influenced by underlying health risks. The health risks may include immune-compromising conditions such as cancer, diabetes, and HIV infections as well as healthcare infrastructure [12]. Despite these being disproportionately distributed between high- and low-income countries, the majority of the current data defining the clinical characteristics and outcomes are largely from high-income countries. Moreover, the usually available data from LMICs are from routine laboratory-based surveillance, which has limited accompanying clinical data [13]. While this type of data is important for disease surveillance and the design of diagnostic protocols, clinical characteristics and outcome data such as mortality, disability, ALOS, and cost are required to inform clinical protocols, planning, resource allocation, and policy formulation. This study therefore aimed to describe the clinical profiles of patients with bacterial AMR presenting at select tertiary hospitals in Uganda. This information is critical to the AMR response, whose emerging surveillance data under the national action plan reveal an escalating problem [14–16]. Methods We conducted a retrospective study of the clinical records of patients with infectious syndromes with at least one culture and sensitivity test admitted at nine Uganda Regional Referral Hospitals (RRHs): Arua, Gulu, Lira, Soroti, Mbale, Jinja, Masaka, Mbarara, and Kabale RRHs between 1st March 2023 and 5th April 2024. Regional referral hospitals in the Uganda healthcare system are the first level with specialist services in surgery, internal medicine, obstetrics and gynecology, and pediatrics. A RRH serves a population of approximately 2 million people across several districts, usually three to five or more [17] The microbiology laboratory capacity of these RRHs has been developed with the support of development partners such as the Fleming Fund [18]. The laboratories receive continuous mentorship in quality management systems on the basis of strengthening laboratory management toward accreditation (SLMTA) and stepwise laboratory quality improvement process toward accreditation (SLIPTA) systems, and the majority are SANAS accredited (Stars 3–5). Qualified clinical and laboratory staff at the study sites collected and analyzed the samples. These individuals receive continuous in-service training and mentorship in diagnostic stewardship, AMR surveillance, bacteriology techniques, and laboratory quality management systems (LQMS). Sample processing followed standard laboratory procedures; in-house reconstituted culture media, including 5% sheep blood agar, MacConkey, and chocolate agar, were used for primary isolation. Bacterial identification utilizes biochemical methods, including colonial appearance and Gram stain reactions. The culture media used were quality controlled for sterility and the ability to support growth via American Type Culture Collection (ATCC) reference bacterial strains. Mueller–Hinton agar (MHA) media performance and the potency of antibiotic discs were assessed via Escherichia coli ATCC 25,922, Staphylococcus aureus ATCC 25,923, and Pseudomonas aeruginosa ATCC 27,853 [19]. Antimicrobial susceptibility testing (AST) was performed using the Kirby–Bauer disc diffusion method according to the pathogen antibiotic combination and breakpoints of the updated Clinical and Laboratory Standards Institute (CLSI) M100, 32nd, 33rd and 34th editions [20–22]. Specimens were collected from patients with community- or hospital-acquired infections. The samples included both Global Antimicrobial Resistance and Use Surveillance System (GLASS) priority samples—blood, urine, urogenital, and cerebrospinal fluid (CSF)—and non-GLASS samples—wound swabs and pus aspirates. Infectious syndrome and mortality determination The diagnoses in the study were those made by the attending clinicians and documented in the clinical records. However, the diagnoses were verified, categorized, and standardized on the basis of the International Statistical Classification of Diseases 11th Revision (ICD-11) codes for infections [23] by two independent medical doctors. Disagreements among the doctors were resolved through discussion, with the supervising physician, who also provided oversight throughout the entire process. The AST results were those recorded and filed in the clinical records or from laboratory records if they were missing in the clinical records. The clinical records of the patients who died were retrieved, and the clinical and laboratory information was extracted. The resistance reported was that present during the patient’s illness or at the time of death but not determined as cause of death postmortem; hence, we report mortality associated with bacterial AMR. Procedure of data abstraction A data abstraction form was designed, pretested, and digitalized in REDCap [24] with inbuilt checks to ensure accurate data collection. Data was collected by a team of clinicians who were trained on the study protocol, data abstraction tool, and data collection standard operating procedures prior to data collection. The RRHs have similar filing systems, which are paper-based, in which clinical files are kept in shelves on the basis of the year, ward, and serial patient admission hospital number. Each record retrieved was assessed for the diagnosis and AST results to determine eligibility for the study. For records without AST results but for which the request for tests was recorded in the clinical records, the results were obtained from the laboratory records. Abstracted data were reviewed by supervisors and a data manager to ensure completeness. The variables extracted included referral status, prior antibiotic use, admission date, admission diagnosis, sampling date, comorbidities, AST results, antibiotics prescribed after culture, treatment outcome, ALOS, and discharge date or date of death, among others. Bacterial antimicrobial-resistant infection was considered as non-multidrug resistance (Non-MDR) if the causative pathogen was resistant to less than three antibiotic classes, MDR infection (MDR) if resistant to at least three antibiotic classes, extremely drug-resistant infection (XDR) if resistant to at least one antibiotic in all classes but two or one antibiotic categories, and pan resistant infection (PDR) if resistant to antibiotic classes in all categories [25]. The XDR and PDR phenotypes were derived where at least four antibiotics appropriate for a particular pathogen were tested. Sample size statement We abstracted data from 599 clinical records of patients with infectious syndromes available in the study period of interest, from which 325 patients had bacterial growth and were included in the study analyses. Because of the retrospective nature, no sample size calculation was done. Statistical analysis Statistical analysis was conducted using R software version 4.3.3 [26]. Descriptive statistics were used to summarize patient characteristics, with categorical variables presented as frequencies and percentages and continuous variables summarized as medians and interquartile ranges (IQRs). Group comparisons were conducted using Pearson’s chi-square test for categorical variables and Fisher’s exact test for variables with expected values less than 5. The Mann‒Whitney U test was used to compare medians for the continuous variables. A mixed-effects modified Poisson regression model with robust standard errors was used to assess factors associated with antibiotic resistance and to estimate adjusted incidence rate ratios (IRRs) with corresponding 95% confidence intervals. The modified Poisson regression model was considered over the logistic regression model because of the common outcome (prevalence of ABR = 88.3%) therefore the logistic regression model would overestimate the relative risk. The modified Poisson approach with robust standard errors was chosen to provide reliable parameter estimates in the presence of potential overdispersion and variance misspecification. A random intercept for regional referral hospital was included to account for clustering and correlation of outcomes among patients treated within the same hospital. Multicollinearity among independent variables was assessed using the variance inflation factor (VIF), with VIF values greater than 10 considered indicative of substantial multicollinearity. Model adequacy was evaluated by examining the distribution of residuals using quantile–quantile (Q–Q) plots and kernel density plots. Statistical significance was considered at 5% level of significance. Results Among the 599 patients included in the study, 54.3% (325/599) had bacterial growth, and 45.7% (274/599) had no growth (Supplementary Table S1, Table S2). A total of 378 isolates were recovered from the 325 samples with growth, of which 262 samples yielded 1 organism, 58 samples yielded 2 organisms, and 4 samples yielded 3 organisms. Of the pathogens tested on three or more antibiotic classes (362), 126/362 (34.8%) were non-MDR and 236/362 (65.2%) MDR. Among the MDR, 165(47.3%) were XDR, and 18 (5.2%) PDR, these being derived where at least four antibiotics appropriate for a particular pathogen were tested (Supplementary Table S3). The clinical and demographic characteristics of patients are summarized in Table 1. Table 1Characteristics of patients with bacteria resistant to at least one class of antibioticsCharacteristicOverall,N (Col%)**Bacteria with resistancep valueNo ABRABRn(Row%)n(Row%)Overall32538 (11.7%)287 (88.3%)Age in complete years0.050C 0–1470 (21.6)14 (20.0)56 (80.0) 15–34114 (35.2)10 (8.8)104 (91.2) 35+140 (43.2)14 (10.0)126 (90.0)Sex0.821C Female174 (53.5)21 (12.1)153 (87.9) Male151 (46.5)17 (11.3)134 (88.7)Referred from other health facility0.875C No236 (72.6)28 (11.9)208 (88.1) Yes89 (27.4)10 (11.2)79 (88.8)Ward0.015F ICU Unit27 (8.3)3 (11.1)24 (88.9) Maternity ward33 (10.2)3 (9.1)30 (90.9) Medical ward35 (10.8)8 (22.9)27 (77.1) Pediatric ward28 (8.6)8 (28.6)20 (71.4) Surgical ward183 (56.3)15 (8.2)168 (91.8) Others*19 (5.8)1 (5.3)18 (94.7)Duration of symptoms (N = 299)0.947C <2 weeks215 (71.9)25 (11.6)190 (88.4) >2 weeks84 (28.1)10 (11.9)74 (88.1)HIV Status (N = 139)0.249F Negative100 (71.9)10 (10.0)90 (90.0) Positive39 (28.1)7 (17.9)32 (82.1)Intubation (N = 152)0.693F No134 (88.2)13 (9.7)121 (90.3) Yes18 (11.8)2 (11.1)16 (88.9)Comorbid conditions0.753F No250 (76.9)30 (12.0)220 (88.0) Yes75 (23.1)8 (10.7)67 (89.3)Antibiotics prescribed at time of admission (N = 322) ¶0.387F No14 (4.3)0 (0.0)14 (100.0) Yes308 (95.7)38 (12.3)270 (87.7)Length of hospital stay in days (Median (IQR))19 (9, 34)14 (6, 26)19 (9, 35)0.073MTreatment Outcome0.786C Died38 (13.0)5 (13.2)33 (86.8) Recovered255 (87.0)29 (11.4)226 (88.6)ABR-Antibiotic resistance, where Col % represents the column percentage; Row % represents the percentage calculated across the row; IQR represents the interquartile range; and n represents the absolute frequency. M represents the Mann‒Whitney U test, C represents the Pearson chi-square test, and F represents Fisher’s exact test. These were lumped because of their small numbers. **The binary outcome variable was obtained by grouping patients who had a pathogen that was resistant to at least one class of antibiotics. *For Ward, others include gynecology, private wing and orthopedic wards Characteristics of patients with bacteria resistant to at least one class of antibiotics ABR-Antibiotic resistance, where Col % represents the column percentage; Row % represents the percentage calculated across the row; IQR represents the interquartile range; and n represents the absolute frequency. M represents the Mann‒Whitney U test, C represents the Pearson chi-square test, and F represents Fisher’s exact test. These were lumped because of their small numbers. **The binary outcome variable was obtained by grouping patients who had a pathogen that was resistant to at least one class of antibiotics. *For Ward, others include gynecology, private wing and orthopedic wards Demographic characteristics The prevalence of ABR among patients with bacterial growth was highest among individuals aged 15–34 years at 91.2% (104/114), followed by those aged 35 and above at 90.0% (104/114), and lowest among children aged 0–14 years at 80.0% (56/70). However, this difference was only marginally statistically significant (p = 0.050). In contrast, sex and referral status were not significantly associated with ABR incidence: 87.9% of females (n = 153/174) and 88.7% of males (n = 134/151) exhibited ABR (p = 0.821). 88.1% of patients not referred from other facilities (n = 208/236) and 88.8% of those referred (n = 79/89) had resistant infections (p = 0.875). (Table 1). Clinical characteristics The prevalence of ABR was 88.8% (79/89) in patients referred from other facilities and 88.1% (208/236) in those not referred, and these were not statistically different (p = 0.875). There were significant differences in resistance rates across hospital wards (p = 0.015), with the highest prevalence of ABR being observed in the surgical wards (91.8%, 168/183), followed by the maternity wards (90.9%, 30/33) and the ICU (88.9%, 24/27). The lowest ABR was in the pediatric wards (71.4%, 20/28). There was no significant difference in resistance between patients with symptoms for less than 2 weeks (88.4%, 190/215) and those with symptoms for more than 2 weeks (88.1%, 74/84). HIV-negative patients had a slightly higher but non-significant resistance rate (90.0%, 90/100) versus HIV-positive patients (82.1%, 32/39). The ABR in patients with comorbidities was 89.3% (67/75), whereas it was 88.0% (220/250) in those without comorbidities (p = 0.753), and these were not significantly different. (Table 1). Treatment outcomes and length of hospital stay The median ALOS in non-ABR patients was 14 days [IQR: 6–26] versus 19 days [IQR: 9–35] in ABR patients. For clinical outcomes, there was no significant difference in the prevalence of ABR among patients who died (86.8%) and those who recovered (88.6%). (Table 1). Pathogen recovery and resistance profiles The 378 pathogens recovered from 325 samples with growth were used in this analysis. Two (2) pathogens were excluded because of insufficient growth. Escherichia coli, 26.5% (100/378), was the most isolated pathogen, followed by Staphylococcus aureus, 20.4% (77/378), and Klebsiella spp., 15.9% (60/378) (Fig. 1). Other notable pathogens recovered included Proteus spp., Citrobacter spp., Acinetobacter spp., and Pseudomonas spp. Pathogen recovery from 325 samples collected. The yellow bars represent isolates from the study that are on the WHO bacterial Priority pathogens list 2024 [27] Escherichia coli showed high resistance across several antibiotic classes, with the highest resistance against extended-spectrum penicillin (ampicillin, 96%), followed by cotrimoxazole (88%), ceftriaxone (82%), and cefepime (79%), among others. Notably, it had 73 and 52% resistance against amoxicillin-clavulanic acid and piperacillin-tazobactam, respectively, and the lowest resistance was against meropenem (0.03%). Figure 2. Percentage resistance of Escherichia coli was calculated for isolates with interpretable susceptibility results for antibiotic-organism combination. This was calculated as resistant/ (intermediate + resistant + susceptible) *100. The denominator (N), number of Escherichia coli isolates, was 100 The resistance pattern of Klebsiella spp. against cotrimoxazole (83%), ceftriaxone (82%), cefepime (75%), and amoxicillin-clavulanic acid (62%), among others, was similar to that of Escherichia coli. However, it had lower resistance against piperacillin‒tazobactam (33%) and higher resistance against meropenem (20%). Figure 3 Percentage resistance of Staphylococcus aureus was calculated for isolates with interpretable susceptibility results for antibiotic-organism combination this was calculated as resistant/ (intermediate + resistant + susceptible) *100. The denominator (N), number of Staphylococcus aureus isolates, was 77 Percentage resistance of Klebsiella pneumoniae was calculated for isolates with interpretable susceptibility results for antibiotic-organism combination. This was calculated as resistant/ (intermediate + resistant + susceptible) *100. The denominator (N), number of Klebsiella pneumoniae isolates, was 33. Staphylococcus aureus had marked resistance against erythromycin (67%), azithromycin (62%), and cotrimoxazole (61%). Notably, resistance against methicillin (methicillin-resistant Staphylococcus aureus – MRSA) was 54%. Figure 4 Factors associated with antibacterial resistance (ABR) The mixed-effects modified Poisson regression model identified ward type and antibiotics prescribed at admission as independent predictors of ABR, whereas demographic factors were not significantly associated with ABR (Table 2). Table 2Mixed effects modified Poisson regression model for factors associated with antibiotic resistance (ABR)CharacteristicABR n/N (%)UnadjustedPR (95% CI)p valueAdjustedPR (95% CI)p valueAge categories 0–1456/70 (80.0)ReferenceReference 15–34104/114 (91.2)1.14 (0.96–1.35)0.1311.07 (0.89–1.29)0.467 35+126/140 (90.0)1.13 (0.97–1.30)0.1171.06 (0.89–1.26)0.506Sex Female153/174 (87.9)Reference Male134/151 (88.7)0.99 (0.90–1.09)0.8450.99 (0.89–1.09)0.845Referred from another health facility No208/236 (88.1)ReferenceN/AN/A Yes79/89 (88.8)1.01(0.91–1.11)0.887N/AN/AWard ICU24/27 (88.9)1.15 (0.87-0.1.52)0.3201.25 (0.87–1.79)0.230 Maternity ward30/33 (90.9)1.18 (0.98–1.41)0.0801.20 (0.97–1.48)0.093 Medical ward27/35 (77.1)ReferenceReference Pediatric ward20/28 (71.4)0.93(0.78–1.12)0.3951.03 (0.79–1.36)0.815 Surgical ward168/183 (91.8)1.19 (1.06–1.33)0.0031.22 (1.09–1.37)<0.001 Others*18/19 (94.7)1.23 (1.04–1.45)0.0141.26 (1.05–1.52)0.016Symptom duration <2 weeks190/215 (88.4)1.01 (0.91–1.12)0.951 >2 weeks74/84 (88.1)ReferenceHIV Status Negative90/100 (90.0)Reference Positive32/39 (82.1)0.89 (0.77–1.09)0.298Intubation No121/134 (90.3)Reference Yes16/18 (88.9)0.98 (0.82–1.18)0.865Comorbid conditions No220/250 (88.0)Reference Yes67/75 (89.3)1.02 (0.94–1.10)0.707Antibiotics prescribed at time of admission No14/14 (100.0)ReferenceReference Yes270/308 (87.7)0.88 (0.83–0.93)<0.0010.84 (0.78–0.91)<0.001Length of hospital stay in days (per unit log increase)19 (9, 35)1.04 (1.00–1.09)0.0531.03 (0.98–1.08)0.199ABR-Antibiotic resistance. PR denotes prevalence ratios, and 95% CIs were obtained by fitting a mixed-effects modified Poisson regression model to complete cases (N=318), excluding 7 observations which were missing on age(n=1), Antibiotics prescribed at time of admission(n=3), and length of hospital stay (n=3). Factors were included in the adjusted model if the p value was <0.2. Prior antibiotic use was not considered for the adjusted model despite having a p value <0.2 at unadjusted because only 44% of the participants had a known status of prior antibiotic use, and the rest were unknown. *For Ward, others include gynecology, private wing and orthopedic wards. These were lumped because of their small numbers. The random intercept variance is estimated to be 4.82e-37(Standard error: 5.51e-36)) Mixed effects modified Poisson regression model for factors associated with antibiotic resistance (ABR) ABR-Antibiotic resistance. PR denotes prevalence ratios, and 95% CIs were obtained by fitting a mixed-effects modified Poisson regression model to complete cases (N=318), excluding 7 observations which were missing on age(n=1), Antibiotics prescribed at time of admission(n=3), and length of hospital stay (n=3). Factors were included in the adjusted model if the p value was <0.2. Prior antibiotic use was not considered for the adjusted model despite having a p value <0.2 at unadjusted because only 44% of the participants had a known status of prior antibiotic use, and the rest were unknown. *For Ward, others include gynecology, private wing and orthopedic wards. These were lumped because of their small numbers. The random intercept variance is estimated to be 4.82e-37(Standard error: 5.51e-36)) Patients in surgical wards had a 22% greater ABR prevalence (adjusted PR = 1.22, 95% CI: 1.09–1.37; p < 0.001) than those on medical wards. Patients admitted to “other wards” (e.g., gynecology, private wing, and orthopedic wards) had a 26% higher ABR prevalence (aPR = 1.26, 95% CI: 1.05–1.52; p = 0.016) compared to those on medical wards. In addition, patients admitted to the maternity wards had a 20% higher ABR prevalence than those admitted to the medical wards, and this association was marginally statistically significant (aPR = 1.20, 95% CI: 0.97–1.48, p = 0.093). Patients who had antibiotics prescribed at admission had a 16% lower ABR prevalence (aPR = 0.84, 95% CI: 0.78–0.91; p < 0.001) than those not prescribed antibiotics. (Table 2). Demographic factors such as age and sex had no significant associations with ABR in the adjusted model. Similarly, clinical and treatment factors such as symptom duration (>2 weeks), HIV status, intubation, and comorbidities had no statistically significant associations with ABR (p > 0.05) in the unadjusted model and were therefore not included in the adjusted model. Each log increase in ALOS had no significant association with ABR (aPR = 1.03, 95% CI: 0.98–1.08; p = 0.199). (Table 2). At model diagnostics: Antibiotics prescribed at admission exceeded the pre-defined threshold for multicollinearity of 10 (VIF = 14.5). At further exploration we found out that antibiotic prescription at admission was significantly associated with ward (chi-square test p-value = 0.032). Despite this, both variables were retained in the model because of their clinical relevance. For model fit, the Q-Q plots indicated that the residuals deviated from normality, which is likely attributable to insufficient number of data points and covariates, given the analysis was performed on data abstracted from hospital records rather than prospectively collected observations. Although a mixed-effects model was employed to account for the clustering of patients within facilities, the estimated random effect was negligible (variance: 4.82 × 10− 3 7; SE: 5.51 × 10− 3 6), suggesting minimal facility-level effect on the outcome. Nevertheless, the mixed-effects model was retained because the data had a hierarchical structure, with patients nested within hospital facilities, and the approach appropriately accounted for potential intra-facility correlation and dependence of observations. Clinical outcomes of ABR patients Among the 287 patients with ABRs, 259 patients had known clinical outcomes, of whom 33 (12.7%) died and 226 (87.3%) recovered (including clinical improvement and recovery with or without complications). Admission to the ICU, intubation, and a shorter ALOS were associated with high ABR-related mortality. (Table 3). Table 3Characteristics of patients with ABR stratified by clinical outcomesCharacteristicOverall, N(col%)Died, n(Row%)¥Recovered, n(Row%)p valueOverall25933 (12.7%)226 (87.3%)Ward<0.001F ICU Unit22 (8.5)8 (36.4)14 (63.6) Maternity ward29 (11.2)0 (0.0)29 (100.0) Medical ward26 (10.0)6 (23.1)20 (76.9) Pediatric ward18 (6.9)4 (22.2)14 (77.8) Surgical ward148 (57.1)14 (9.5)134 (90.5) Others16 (6.2)1 (6.3)15 (93.8)Intubation0.002F No102 (87.2)7 (6.9)95 (93.1) Yes15 (12.8)6 (40.0)9 (60.0)Comorbid conditions0.785C No201 (77.6)25 (12.4)176 (87.6) Yes58 (22.4)8 (13.8)50 (86.2)Antibiotics prescribed at time of admission0.064F No13 (5.1)4 (30.8)9 (69.2) Yes243 (94.9)28 (11.5)215 (88.5)Length of hospital stay in days (Median (IQR))19 (9, 33)8 (5, 19)20 (10, 35)<0.001 MDiagnosis0.439C Documented infections125 (48.3)18 (14.4)107 (85.6) Documented infections coupled with underlying non-infectious comorbidities134 (51.7)15 (11.2)119 (88.8)where Col % represents the column percentage, Row % represents the row percentage, IQR represents the inter quartile range, and n represents the absolute frequencies. Mstands for the Mann-Whitney U test, Cstands for Pearson Chi-square, and Fstands for Fisher’s Exact test. ¥-For treatment outcomes: The Recovered category includes patients that recovered with and without complications, as well as clinical improvement. Unknown outcomes include those that were lost to follow-up and those that were referred or self-referred. * For Ward, others include gynecology, private wing and orthopedic ward. These were lumped because of their small numbers. Participants who had no treatment outcomes (28) were excluded from this analysis Characteristics of patients with ABR stratified by clinical outcomes where Col % represents the column percentage, Row % represents the row percentage, IQR represents the inter quartile range, and n represents the absolute frequencies. Mstands for the Mann-Whitney U test, Cstands for Pearson Chi-square, and Fstands for Fisher’s Exact test. ¥-For treatment outcomes: The Recovered category includes patients that recovered with and without complications, as well as clinical improvement. Unknown outcomes include those that were lost to follow-up and those that were referred or self-referred. * For Ward, others include gynecology, private wing and orthopedic ward. These were lumped because of their small numbers. Participants who had no treatment outcomes (28) were excluded from this analysis ABR-associated mortality varied significantly by ward (p < 0.001), with the ICU having the highest mortality (36.4%), followed by medical wards (23.1%), and pediatric wards (22.2%). The surgical wards had the lowest mortality rate (9.5%), whereas no deaths occurred in the maternity wards. Compared with non-intubated patients, intubated patients had a greater risk of ABR-associated death (40.0% vs. 6.9, p = 0.002). Patients who were not prescribed antibiotics at admission had higher ABR-associated mortality (30.8% vs. 11.5%, p = 0.064) compared to those who were prescribed antibiotics, but this difference was nonsignificant. The median ALOS was significantly shorter for those who died than for those who recovered (8 days [IQR: 5–19] vs. 20 days [IQR: 10–35], p < 0.001). There was no significant difference in mortality between patients with documented infections (14.4%) and those with documented infections plus underlying non-infectious comorbidities (11.2%, p = 0.439). Prescription of antibiotics at time of admission and comorbid conditions were higher in patients who recovered but were not significantly associated with higher risk of ABR- associated mortality. (Table 3). Discussion Unlike prior evaluations that have focused on the prevalence of individual resistant pathogens or their prevalence in particular syndromes, we sought to characterize the wider clinical burden and impact of ABR on clinical outcomes. We report a high ABR prevalence, 34.8% non-MDR versus 65.2% MDR, among inpatients with bacterial growth in nine referral hospitals in Uganda. Resistance was high in all wards but higher in the ICU, maternity ward, and surgical ward. Moreover, ABR was associated with ward type. Admission to surgical or other non-major wards (gynecology, private, and orthopedic) was independently associated with a higher risk for ABR, whereas the prescription of antibiotics at admission was associated with a reduced risk for ABR. Age, sex, referral status, and HIV status were not associated with ABR. Overall, ABR-associated mortality was 12.7%, which was higher among inpatients with meningitis, those at extreme ages (<4 and >45 years), females, intubated patients, and those admitted to the ICU. Patients with resistance stayed 5 days longer in the hospital than those with no resistance, but ALOS was not associated with ABR. Patients who died stayed in the hospital for a significantly shorter period than those who survived. Escherichia coli, Klebsiella spp., and Staphylococcus aureus, the most isolated pathogens, were highly resistant to commonly used antibiotics, including extended-spectrum beta-lactamase (ESBL) and methicillin-resistant Staphylococcus aureus (MRSA) phenotypes. The high ABR prevalence in hospital wards, especially ICU, maternity, and surgical wards, is consistent with results from a prior study in the same setting, which reported that resistance prevalence varied by ward type and was higher in specialized units such as surgical and ICU units [16, 28, 29]. The ICU is characterized by invasive procedures, high broad-spectrum antibiotic use, and the ability to handle critically ill patients [30]. Similarly, maternity and surgical wards highly use antibiotics for the prevention and treatment of surgical site infections resulting from high-risk obstetric and surgical procedures or inadequate infection prevention [31]. Therefore, the clinical environment of these wards favors hospital-acquired infections (HAIs), which are commonly associated with AMR [32], whereas high antibiotic use imparts high selection pressure, leading to the emergence of resistant pathogens [33]. The prescription of antibiotics at admission was associated with a reduced risk and mortality associated with ABR. As most of the encountered clinical resistance is selected resistance due to inappropriate antibiotic use [34], effective therapy based of appropriate guidelines or microbiology results mitigates the propagation of such resistance, which is the goal of antimicrobial stewardship. While diagnostic–antimicrobial stewardship interventions such as adherence to empirical treatment guidelines or targeted prescriptions on the basis of microbiology results, which are practiced at the study sites may have had an effect [35], other factors such as antibiotic administration before sample collection or disease severity that were not assessed could have also contributed to this result. ABR not being associated with age, sex, referral status, or HIV status, contrasts findings of a prior evaluation in the same setting, which revealed that AMR varied significantly with age and HIV status, whereas resistance to erythromycin was greater in females than in males [16]. Other studies have also shown higher rates of AMR in immunocompromised patients, such as HIV-positive patients and cancer patients [36], and in individuals of different ages [1, 2]. The lack of association in our study could reflect the relatively homogenous population of patients across these nine hospitals or the broad range of antibiotic-resistant pathogens involved, which may not follow predictable patterns on the basis of patient demographics alone. The RRHs also function as the primary point of contact with a health facility for surrounding communities as well as distant communities referred for specialized or advanced care [37]. Over the last two decades, AMR-related deaths have shown an increasing trend, with 1.27 million and 1.14 million deaths globally occurring in 2019 and 2021, respectively [1]. Indeed, ABR had a profound impact on patient outcomes in our study: the overall mortality rate was 12.7%, which was higher in patients in the ICU, those with short hospital stays, and intubated patients. These findings are consistent with prior literature, where patients with AMR admitted to the ICU were at a higher risk of mortality [38]. These patients are more commonly intubated and remain in the hospital for a short time due to early death [38]. Despite not being statistically significant, ABR-associated mortality in our study was greater among those aged 0–14 and 35+ years than among those aged 15–35 years. This was similar to the findings of the 2019 evaluation of the global AMR burden, in which mortality was highest in those under 5 years and over 50 years of age [1]. However, this trend changed in 2021, with mortality increasing in those over 50 years of age and decreasing in those under 5 years of age, except for those with sepsis, who had a 31.1% mortality related to AMR [1]. Severe infectious syndromes such as BSIs, lower respiratory tract infections, and other deep organ infections are associated with high AMR-related deaths [39]. Our study, however, did not report any difference in mortality between patients with documented infections and those with documented infections plus underlying non-infectious comorbidities. Emerging clinical studies in the same setting indicate that sepsis, pneumonia, and meningitis are the leading causes of death due to AMR [40]. On the other hand, an extended hospital stay for patients with resistant infections implies increased clinical burden and resource utilization. Although not statistically significant, patients with resistance stayed longer in the hospital by five days, whereas those who died stayed in the hospital for a significantly short period. This mirrors findings in Ghana, where AMR patients stayed five days longer than did patients with susceptible infections [41], but contrasts the situation in France and Australia, where AMR patients stayed only 1.6 days and 2.9 days longer [42, 43]. This might result from differences in the resources and facilities used to diagnose and treat AMR. Escherichia coli, Klebsiella spp., and Staphylococcus aureus, the most isolated pathogens in our study, had high resistance to commonly used antibiotics. These results closely relate with findings from a recent evaluation in the same setting, which also revealed that the three pathogens were the most isolated pathogens and had similar resistance patterns [16]. Of most concern is the high prevalence of notorious resistant phenotypes like ESBL, MRSA, and emerging carbapenem-resistant organisms (CROs). The high amoxicillin clavulanic acid resistance in Escherichia coli, presumably indicates a role for AmpC beta lactamases whereas the high resistance of Klebsiella spp. to imipenem/meropenem suggests presence of carbapenamase producers. However, laboratory confirmation for ESBL and carbapenamase-producing organisms was not routinely done due to logistical gaps. Prior studies have reported an ESBL prevalence of >80% and a MRSA prevalence between 46 and 56%. The MRSA prevalence was comparable to that reported in neighboring countries, Kenya (53.4%) [44] and Tanzania (43.3%) [45]. Vancomycin-resistant Enterococcus (VRE) was not tested. This situation underscores the urgent need to strengthen stewardship and infection control measures to curb the spread of these resistant pathogens. The strengths of our study lie in the multi-hospital design and the inclusion of a wide variety of hospital wards and patient demographics, which constitute one of the first attempts to describe the clinical burden of AMR. However, several limitations should be considered, including the retrospective nature of the observational study, which led to some missing variables and limited the ability to define the causality between ABR, clinical outcomes and the variables assessed. Additionally, the study did not include information on the specific antibiotic regimens used, which could have provided deeper insight into the relationships between antibiotic therapy and resistance patterns. The study also worked with only samples with growth and in some instances only those with resistance to characterize the clinical profile of resistance. This introduced some selection bias and therefore limits the generalization of the results to the general patient population. Finally, much as the study measured several factors contributing to resistance, other significant factors like timing of antibiotic exposure versus specimen sampling were not evaluated. Antibiotic exposure before sampling significantly reduces bacteria growth and could lead to false no growth cultures results, which could have led to underestimation of AMR [46]. Conclusion In conclusion, our findings highlight the critical public health threat posed by AMR in Uganda and the substantial clinical burden it imposes on the hospital systems. The high prevalence of resistant pathogens, especially in high-risk wards such as the ICU and surgical areas, as well as the impact of ABR on mortality and hospital stay, calls for urgent and comprehensive strategies to combat AMR, including strengthening antibiotic stewardship programs, improving infection control practices, and enhancing diagnostic capacities. Electronic supplementary material Below is the link to the electronic supplementary material. Supplementary Material 1 Supplementary Material 2 Acknowledgements The authors would like to thank the research team for their time in conducting this study. Abbreviations Antibiotic Resistance Average Length of Hospital Stay Antimicrobial Resistance Extended-spectrum beta-lactamase Human Immunodeficiency Virus Intensive Care Unit Multidrug resistance Methicillin Resistant Staphylococcus Aureus Extremely Drug Resistant Author contributions Conceptualization: JM, CT, FD, RG, EN, HM; Methodology: JM, CT, FD, RK; Formal analysis and investigation: JM, RG, FD; Writing - original draft preparation: CT, JM, FD; Writing - review and editing: RK, AA, SO, PM, DT, FO, JE, AY, NO, AO, BW, FJN, CO; Funding acquisition: FK, AK; Supervision: JM, DMB. All the authors read and reviewed the manuscript and approved the final version. Funding This research was funded by the Welcome Trust CAMO-Net Programme [226692/Z/22/Z]. For the purpose of open access, the author has applied a CC BY public copyright license to any Author Accepted Manuscript version arising from this submission. Data availability Data supporting this study are available from the IDI AMR Data Warehouse via https://amrdb.idi.co.ug/datasets/card/clinical_outcomes. Access to the data is subject to approval and a data sharing agreement due to legal or ethical reasons. Declarations Ethical approval This study was approved by the Infectious Diseases Institute (IDI) Research Ethics Committee (IDI-REC-2023–67), which also granted a waiver of participant consent in accordance with the Declaration of Helsinki. The study was also approved by the Uganda National Council for Science and Technology (UNCST - HS3690ES). No participants’ identification information was used to disseminate or publish the study results. Informed consent Data collected during this study was retrospective and didn’t include any participant identifiers as such, the study was provided with a waiver of consent by the research ethics committee. Consent for publication Not applicable Competing interest The authors declare no competing interests. Footnotes Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. References Associated Data Supplementary Materials Supplementary Material 1 Supplementary Material 2 Data Availability Statement Data supporting this study are available from the IDI AMR Data Warehouse via https://amrdb.idi.co.ug/datasets/card/clinical_outcomes. Access to the data is subject to approval and a data sharing agreement due to legal or ethical reasons.