Mortality outcomes and economic burden of antimicrobial-resistant bacterial bloodstream infections in a Nigerian tertiary hospital

Abstract

Background

Antimicrobial resistance (AMR) poses a global public health and economic challenge, significantly contributing to poor outcomes of bacteria bloodstream infections (BSI). There is a paucity of local data on outcomes of AMR BSIs in Nigeria. This prospective cohort study determined outcomes and healthcare costs of BSIs stratified by resistance status.

Method

Data were prospectively collected from enrolled participants (neonates, children and adults) with susceptible (cohort 1) and resistant (cohort 2) BSIs caused by Enterobacterales, Pseudomonas spp, Acinetobacter spp., Staphylococcus aureus, and Enterococcus spp, and a control group without BSI (cohort 3), at the University of Port Harcourt Teaching Hospital between April and November 2025. Outcomes (in-hospital mortality and extra length of hospital stay, LOS) were compared between the three cohorts using time-dependent and multi-state regression models that adjusted for patients’ characteristics, time-varying cofounders and competing risks, while healthcare costs were compared using Students’ t’ test and analysis of variance.

Results

A total of 952 patients were recruited, with 330 neonates, 271 children and 351 adults. The overall prevalence of bacteria-confirmed BSI was 18.1%, with neonatal BSI of 27.3%, childhood BSI of 12.5% and adult BSI of 13.7% (p < 0.0001). In-hospital mortality was higher for healthcare onset (HO) compared to community onset (CO) BSI (_sdHR:1.7 (0.65–4.5)], MRSA compared to MSSA BSI [_sdHR: 4.1 (0.49–33.5)], and clindamycin-resistant compared to clindamycin-susceptible S. aureus BSI (_sdHR: 6.8 (0.36–129.4)]. Multistate modelling showed that HO-BSI caused extra LOS of 2.28 days compared to CO-BSI, resistant BSI caused extra LOS of 1.72 days compared to susceptible BSI, MRSA-BSI caused extra LOS of 2.83 days compared to MSSA-BSI, and ESBL-BSI caused extra LOS of 4.28 days compared to non-ESBL-BSIs. Healthcare costs, catastrophic expenditures and impoverishment from out-of-pocket spending were high for all groups but generally higher for patients with HO and resistant BSIs.

Conclusion

Compared to susceptible bacterial BSIs, resistant bacterial BSIs are associated with higher mortality, extra length of hospital stays and excess healthcare costs. Policies to reduce the burden of resistant BSIs should focus on improving infection prevention and control and antimicrobial stewardship practices in both hospital settings and the community.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s12866-026-05348-2.

Introduction

Antimicrobial Resistance (AMR) has emerged as a global public health and economic threat of serious concern. In 2019, AMR was associated with an estimated 4.95 million deaths, with 1.27 million deaths directly attributable to bacterial AMR worldwide [1]. If current trends persist, annual deaths associated with and attributable to AMR are projected to rise to 8.2 million and 1.9 million, respectively, by 2050 [2].

The global economic burden of AMR is substantial, driven by increased length of hospital stay, rising healthcare costs, and productivity losses. A recent study estimated that hospital-related costs associated with AMR reached US$693 billion globally in 2019, while productivity losses due to AMR-related deaths were estimated at US$194 billion [3].

Although all regions of the world are affected by AMR, the burden is estimated to disproportionately affect low-and middle-income countries (LMICs) compared to high-income countries (HICs). In 2019, Western sub-Saharan Africa recorded the highest AMR burden, with 114.8 deaths per 100,000 population associated with AMR and 27.3 deaths per 100,000 population directly attributable to it [1].

Antimicrobial resistance in bloodstream infections (BSIs) contributes significantly to the impact of AMR on mortality in LMICs and has been reported to be associated with a 58% increase in BSI death rates in LMICs [4]. In the WHO Africa region, of the 250,000 deaths attributable to AMR in 2019, BSIs accounted for 22% of the estimated burden of death, constituting the second largest fatal AMR burden after lower respiratory and thoracic infections [5].

In Nigeria, the reported prevalence of bloodstream infections varies widely, ranging from 6.6% to 49% (Nigerian Centre for Disease Control) [6]. This high and variable burden suggests a significant contribution of BSIs to AMR-related mortality in the country. Consequently, deaths attributable to AMR in Nigeria are estimated to exceed those from several major disease groups, including enteric infections, respiratory infections, tuberculosis, maternal and neonatal disorders, neglected tropical diseases, malaria, and cardiovascular diseases.

However, effectively tackling the issue of AMR requires a clear understanding of the true burden and the context-specific factors driving it. Recent efforts at comprehensively estimating the burden of AMR in LMICs have largely depended on modelling, which may have underestimated the burden of AMR in these countries [1]. Paucity of accurate empirical data on the actual burden of AMR in LMICs, including data on both phenotypic and genotypic AMR profiles from BSIs, has been recognized as an important limitation of the findings from previous studies [1, 4, 5]. Among the few studies which have attempted to use empirical data to evaluate the excess mortality and economic burden attributable to AMR in LMICs, important limitations in terms of study methodology adopted [7], study population evaluated [8], or pathogen-drug combinations assessed [9] may have impacted the robustness of the findings of these studies. In order to eliminate bias in the estimation of the mortality attributable to AMR BSIs, the use of an appropriate cohort study design with comparison of two counterfactual scenarios and exposure density sampling to match cases of AMR BSI with the comparison groups with antimicrobial-sensitive BSI or ‘no BSI’ has been recommended [10]. In addition, accurately estimating the economic impact of AMR BSIs requires the adoption of the appropriate perspective and time horizon in estimating costs associated with AMR BSIs [11]. An accurate, actionable estimate of the AMR burden is needed to prioritise resource allocation for strategies to reduce AMR, including infection control and antimicrobial stewardship, particularly in LMICs like Nigeria.

This study therefore assessed mortality outcomes in patients with BSIs, extra length of stay (LOS) and healthcare costs associated with AMR BSIs in a tertiary hospital in Nigeria. To aid potential future cost-effectiveness evaluations and intervention prioritization based on the economic burden of BSIs for patients and governments, stratified by resistance status, this study estimated the costs of BSIs in terms of out-of-pocket and total expenses, and the implications for patients on catastrophic healthcare expenditure and impoverishment.

Materials and methods

Study setting and period

The study was conducted in the University of Port Harcourt Teaching Hospital (UPTH), a public tertiary healthcare facility located in Port Harcourt, Rivers State, South-south Nigeria. Rivers State has an estimated population of 9.6 million [12]. The hospital is a 900-bed capacity facility with 45 acute care and 12 intensive care unit (ICU) beds, and approximately 12,095 inpatient admissions yearly. The hospital serves as a referral center for the adjoining States in the region. The study was conducted between April and November 2025.

Ethical considerations

This study was conducted in compliance with the principles of the Helsinki Declaration. Ethical approval for this study was obtained from the Health Research Ethics Committee of the University of Port Harcourt Teaching Hospital (UPTH/ADM/90/S.11/VOL.XI/1822) before data collection. Written informed consent was obtained from all participants, and for participants younger than 16, consent to participate was obtained from their parents or legal guardians. Confidentiality of all study participants was ensured and maintained during and after the study.

Study design and population

The study is a prospective observational cohort design that estimated mortality and healthcare costs of BSIs in hospitalized neonates (1–28 days), children (1 month-17 years of age) and adults (≥ 18 years of age) in acute care departments of the hospital for selected antimicrobial-resistant bacterial pathogens, which included Enterobacterales, Pseudomonas species, Acinetobacter species, Staphylococcus aureus, Enterococcus species, and Streptococcus pneumoniae.

Study participants and eligibility criteria

During the period of the study, patients of all age groups (neonates, children, and adults) hospitalized as inpatients for suspected BSI, using distinctly defined clinical criteria [13–15], and requiring at least one overnight stay, were eligible for inclusion. Active case finding of all eligible patients was done by trained research assistants during routine hours in all acute care wards and intensive care units of the hospital. All patients with suspected BSI were referred for blood culture sample collection immediately, before administration of any antibiotic treatment.

Blood culture collection and microbiological analysis

For adult patients, 10 ml each of venous blood was aseptically collected by trained phlebotomists into aerobic and anaerobic Bact/Alert culture bottles. For neonates and children, an age-appropriate volume of blood (1–4 mL) was collected into an aerobic Paediatric Bact/Alert bottle [12]. Blood culture specimens were transported immediately to the clinical microbiology laboratory of the hospital for incubation in the automated Bact/Alert blood culture system [16].

Subculture from positively flagged culture bottles was made on Blood and MacConkey agar plates, and on CHROM GNB, CHROM Pseudo and CHROM Acinetobacter plates (Liofilchem, Italy) [17]. Preliminary identification of bacteria was based on colony morphology and Gram stain reaction performed on the colonies. Gram-positive isolates were identified to species level using conventional biochemical tests that included catalase, coagulase, bacitracin, optochin, bile solubility and bile aesculin [18]. Gram-negative isolates were identified to species level using conventional biochemical tests (lactose fermentation, oxidase, indole and motility) and Analytical Profile Index (API) and Enterosystem 18R.

Antimicrobial susceptibility test (AST)

Antimicrobial susceptibility test (AST) of each isolate to a panel of antibiotics was performed by the modified Kirby Bauer disk diffusion method, and result interpreted as sensitive, intermediate or resistant using the Clinical and Laboratory Standards Institute (CLSI) guidelines [19]. The antibiotics tested against both Gram-positive and Gram-negative isolates were penicillins (ampicillin, ampicillin-sulbactam, amoxicillin-clavulanate, piperacillin-tazobactam), cephalosporins (cefoxitin, cefotaxime, ceftriaxone, ceftazidime, ceftazidime-avibactam), carbapenems (imipenem, meropenem, ertapenem), aminoglycosides (gentamicin, amikacin), fluoroquinolones (ciprofloxacin, levofloxacin), and amphenicols (chloramphenicol). In addition, Gram-positive isolates were tested against macrolides (erythromycin, clindamycin), phosphonic acid (fosfomycin), oxazolidinones (linezolid) and glycopeptides (vancomycin, by E-test), while Gram-negative isolates were additionally tested against monobactam (aztreonam), glycylcyclines (tetracycline, minocycline, doxycycline, tigecycline) and polymyxin (colistin, by E-test). However, Acinetobacter spp were not tested against ampicillin and chloramphenicol, and Pseudomonas spp were not tested against ampicillin, ampicillin-sulbactam and chloramphenicol.

Antibiotic resistance detection in isolates

Selected resistance was detected in each isolate was by phenotypic methods and defined in line with CLSI guideline [19]; Methicillin resistant Staphylococcus aureus (MRSA) was defined as S. aureus resistant to cefoxitin by the standard AST method; Vancomycin resistant S. aureus (VRSA) was defined as S. aureus resistant to vancomycin as tested by E-test method; Vancomycin-resistant Enterococci (VRE) was defined as Enterococci resistant to vancomycin by standard AST method; Inducible clindamycin-resistant S. aureus was defined as S. aureus resistant to erythromycin by standard AST and exhibiting inducible resistance to clindamycin in the ‘D test’. ESBL-producing Gram-negative bacilli (GNB) was defined as any GNB that is non-susceptible (i.e. resistant or intermediate) to ceftazidime, cefotaxime, ceftriaxone, or cefepime in standard AST and confirmed as ESBL-producer by the combination E-test method; Carbapenem-resistant GNB was defined as any GNB resistant to imipenem, meropenem, doripenem, or ertapenem by standard AST; Colistin-resistant GNB was defined as any GNB resistant to colistin by standard colistin E-test method; Multi-drug resistant (MDR) Acinetobacter/Enterobacterales was defined as any Acinetobacter spp or Enterobacterales that is non-susceptible (i.e., resistant or intermediate) to at least one agent in at least 3 of these 6 antimicrobial classes (β-lactam & β-lactam/β-lactamase inhibitor combination, Aminoglycosides, Carbapenems, Fluoroquinolones, Cephalosporins, Sulbactam) [19]; MDR Pseudomonas was defined as any Pseudomonas that is non-susceptible (resistant or intermediate) to at least one agent in at least 3 of these 5 antimicrobial classes (β-lactam & β-lactam/β-lactamase inhibitor combination, Aminoglycosides, Carbapenems, Fluoroquinolones, Cephalosporins) [20].

Quality control

Quality control for antimicrobial susceptibility testing was performed using standard reference strains, including Escherichia coli ATCC 25,922, Staphylococcus aureus ATCC 25,923, and Pseudomonas aeruginosa ATCC 27,853, in accordance with CLSI guidelines.

For ESBL detection, quality control was performed using Klebsiella pneumoniae ATCC 700,603 (ESBL-positive control) and Escherichia coli ATCC 25,922 (ESBL-negative control). For colistin susceptibility testing, Escherichia coli ATCC 25,922 and Pseudomonas aeruginosa ATCC 27,853 were used as quality control strains, and testing was conducted in accordance with recommended guidelines.

All quality control strains used for antimicrobial susceptibility testing, ESBL detection, clindamycin resistance, and colistin susceptibility testing produced results within the acceptable ranges specified by CLSI guidelines, confirming the validity and reliability of the test procedures.

Enrollment of eligible participants and the control group for follow-up

Three cohorts of participants were enrolled for follow-up until discharge (dead or alive) or for 30 days post enrollment. The cohorts included patients with drug-resistant BSIs of each target bacteria (cohort 1), drug-susceptible BSIs of each target bacteria (cohort 2), and patients with no BSI of the target bacteria/culture-negative suspected infected group (cohort 3). Participants were included only once in each cohort. If a participant had multiple episodes of the same bacterial BSIs during the study period, only the first bacterial isolate was considered in line with the WHO GLASS methodology [21]. The time of blood specimen collection was utilised as a proxy for the onset of infection for cohort 1 or cohort 2 participants [21]. Community-onset BSI (CO-BSI) exposed patients were defined as those for whom the date of specimen collection was within the first two days of hospital admission [22, 23]. If the specimen date of collection was greater than two days post-admission, the exposure was classified as hospital-onset BSI (HO-BSI). Using exposure density sampling, participants with HO-BSI (cohorts 1 and 2) were matched to uninfected, unexposed controls (cohort 3). Matching was performed at the time of risk set sampling based on length of hospital stay (± 1 day), hospital location at blood collection, and admission period (± 1 week) to ensure temporal and clinical comparability.

Data collection

At enrollment, socio-demographic and clinical information were collected by trained research assistants from participants in all three cohorts using a composite participant enrollment questionnaire and from clinical case notes. During follow-up, additional clinical information on effect mediators, such as empirical and targeted antibiotic therapy (agent, dates, route, dose, and adequacy) were collected and recorded. At discharge or 30 days post enrollment, healthcare costs data were collected. If at day-30 after enrolment, a participant is already discharged, life status was ascertained through out-patient mobile phone contact with the participant or a family member. Data to estimate healthcare costs and catastrophic health expenditure were collected through hospital bills and exit surveys from participants/guardians/relatives in the three cohorts. All data collection was done electronically using standardized data collection Google forms. Data were transcribed from Google Forms into secure REDCAP software for storage and validation prior to analysis. The primary outcome was in-hospital mortality, while secondary outcomes included excess length of stay in hospital, healthcare costs and 30-day mortality.

Data analysis

There were no missing data in the final dataset, as continuous data validation procedures were implemented. Incomplete or inconsistent entries were identified and addressed in real time before data cleaning and analysis. The data cleaning and subsequent statistical analyses were performed using R statistics software version 4.3.2. Descriptive statistics were summarised using median and interquartile range (IQR) for continuous variables, and proportions (percentages) for categorical variables. Chi-square and Fisher Exact tests were used to determine the association between categorical variables, while the student’s ‘t’ test and analysis of variance (ANOVA) were used to determine associations between continuous variables, with p < 0.05 considered statistically significant. For binary outcomes, Hazard Ratios, Odds Ratios (ORs), or Risk Ratios (RRs) > 1 indicated increased odds or risk of the outcome associated with the exposure. For continuous outcomes such as length of hospital stay, a positive mean difference indicates an excess burden in the exposed group. Statistical significance for outcomes was assessed using 95% confidence intervals (CIs), with association considered statistically significant if the 95% CI value does not include the null value (specifically, 1 for ORs or RRs, and 0 for mean difference).

Prevalence and incidence of BSI

The prevalence of bacteria-confirmed BSI was determined by dividing the number of participants who tested positive for bacterial pathogens of interest by the total number of patients screened for sepsis in the study population. This was stratified with respect to age category, sex, types of BSI (CO-BSI or HO-BSI) and resistance profiles. The hospital incidence of pathogen-specific BSIs was expressed per 1,000 patient-days and stratified into relevant resistance profiles and patient groups (neonates, children and adults). Incidence data were displayed per month for the 7-month period of study to visualise possible fluctuation. The monthly incidence data were related to the facility IPC, HH and AMS assessment scores at inception and completion of the study.

In-hospital mortality

The exposures of interest (BSIs: resistant & susceptible) and the primary outcome (in-hospital mortality) are both time-dependent. Therefore, survival analysis (time-to-event) methods were applied while considering competing events, i.e., hospital discharge was a competing event for observing in-hospital mortality [24]. Cog regression model with sub-distribution relative hazard (_sdRH) that incorporated competing events was applied, which included BSI as a ‘time-fixed’ (time zero = onset of BSI) and ‘time-dependent’ exposure (time zero = admission) and provided estimates of the impact of BSIs (resistant & susceptible) on the outcome of interest as well as the competing event. Marginal structural Cox regression model was used to account for potential ‘time-varying’ confounders such as severity of illness and intravenous devices. In the analysis of hospital-acquired bloodstream infections (HO-BSI), variables such as mechanical ventilation and central venous catheterization were defined as binary exposures based on their presence within the 48-hour period prior to blood culture collection. As such, these variables were treated as time-fixed covariates rather than time-dependent exposures.

Excess length of hospital stays

Overall, excess length of hospital stays (LOS) associated with BSIs (resistant & susceptible) was determined for categories of patients. For CO-BSIs, the full length of stay was assumed attributable to the infection, and this was directly compared between resistant and susceptible cohorts. For HO-BSIs, multistate modelling was applied to ensure that only LOS after the BSI exposure was considered attributable to the infection (resistant or susceptible) [25]. This model prevented time-dependent bias by modelling the true timing of events of individual patients and estimating transition probabilities between states, comparing patients with resistant and susceptible BSI.

Healthcare costs, catastrophic expenditure and impoverishing out-of-pocket expenses

To aid potential future cost-effectiveness evaluations and intervention prioritization based on the economic burden of BSIs for patients and governments, stratified by resistance status, we estimated the costs of BSIs in terms of out-of-pocket and total expenses, and the implications for patients on catastrophic healthcare expenditure and impoverishment. The median (interquartile range IQR) healthcare costs of patients with CO-BSI and HO-BSI were compared with the control (no BSI) using ANOVA and ‘t’ test, and this was stratified for selected resistant and susceptible BSIs.

Results

The reporting of our study findings was done in accordance with Strengthening the Reporting of Observational studies in Epidemiology (STROBE) guidelines [26].

Sociodemographic characteristics of patients with BSI

A total of 952 patients with suspected sepsis were recruited, of whom 172 were culture-positive for bacterial pathogens, indicating a bloodstream infection prevalence of 18.1%. The age category of the 172 patients with BSI shows that 90 (52.3%) were neonates, 34 (19.8%) were children (aged 1 month-17 years), and adults were 48 (27.9%).

There are slightly more males (n = 91, 52.9%) than females (n = 81, 47.1%). The majority (77.9%) of the adults and parents of neonates/children live in urban areas. More than half (56.3%) of the adult participants with BSI were married, with secondary level education (52.1%) and were ‘sale service’ by occupation (59.0%). The majority of the adult participants with BSI were employed (81.1%), and all (100.0%) were Christian by religion Table 1.

Table 1.

Sociodemographic characteristics of patients with bloodstream infection caused by bacterial pathogens of interest

Variable	Characteristic	Freq (%)
Age (years): mean+/- SD	15.6+/-25.4
Age category	Neonates (1–28 days)	90 (52.3)
	Children (1 month-17 years)	34 (19.8)
	Adults (≥ 18 years)	48 (27.9)
Sex	Male	91 (52.9)
	Female	81 (47.1)
Locality	Urban	134 (77.9)
	Peri-urban	22 (12.8)
	Rural	16 (9.3)
Marital status (adult only; n = 48)	Married	27 (56.3)
	Others	12 (25.0)
	Single	9 (18.8)
Occupation group (adult only; n = 48)	Sale service	23 (59.0)
	Govt civil servant	6 (15.4)
	Professional	6 (15.4)
	Transporter	2 (5.1)
	Military security	1 (2.6)
	Skilled	1 (2.6)
Education level (adult only; n = 48)	None/primary	2 (4.2)
	Secondary	25 (52.1)
	Tertiary	21 (43.8)
Employment status (adult only; n = 48)	Employed	39 (81.3)
	Not employed	9(18.8)
Ethnicity (adult only; n = 48)	Others	28 (58.3)
	Igbo	17 (35.4)
	Yoruba	2 (4.2)
	Hausa	1 (2.1)
Religion (adult only; n = 48)	Christianity	48 (100.0)

This table presents the demographic characteristics of patients with BSI, including age distribution, sex, locality, and selected socioeconomic variables among adult patients

Prevalence of bloodstream infection in the study population

The prevalence of BSI was significantly higher in neonates (27.3%) than in children (12.5%) and adults (13.7%) (p < 0.0001). This significant difference also holds for both males (p < 0.0001) and females (p = 0.0119). The prevalence of community-onset BSI (12.2%) in the study population was higher than that of healthcare onset BSI (5.9%) (Table 2), and this was significantly higher in neonates (20.6%) compared to children (7.7%) and adults (7.7%) (p < 0.0001). However, the prevalence of HO-BSI was not significantly different across age categories (p = 0.622) Table 2.

Table 2.

Prevalence of bloodstream infection caused by bacterial pathogens of interest stratified by age category and types of infection

Age category	Screened (%)	CO-BSI positive (%)	HO-BSI positive (%)	Total BSI positive (%)
Neonates (1–28 days)	330	68 (20.6)*	22 (6.7)	90 (27.3)*
Children (1 month-17 years)	271	21 (7.7)	13 (4.8)	34 (12.5)
Adults (≥ 18 years)	351	27 (7.7)	21 (6.0)	48 (13.7)
Total	952	116 (12.2)	56 (5.9)	172 (18.1)
x 2		33.475	0.9496	29.046
p value		< 0.0001*	0.6220	< 0.0001*

x² = Chi square; ^*= Statistically significant at p < 0.05

Incidence of bloodstream infection in the study population

The overall incidence of BSI was 3.68/1000 patient days, with that of HO-BSI being 1.34/1000 patient days and that of CO-BSI being 5.73/100 patient admissions.

Frequency of bacterial pathogens of bloodstream infection

The frequency distribution of bacterial pathogens of interest in the study, as presented in Fig. 1, shows predominance of Gram-negative bacteria (GNB) (73.3%) compared to Gram-positive bacteria (GPB) (26.7%). With Klebsiella spp being the most frequent GNB (22.1%), while S. aureus is the most frequent GPB (22.1%).

Fig. 1.

Frequency of bacterial pathogens of interest causing bloodstream infection

Antibiotic susceptibility of bacterial isolates of bloodstream infections

The invitro antimicrobial susceptibility test result in Table 3 shows that more than 50% of the Gram-positive isolates (S. aureus and Enterococcus spp) were resistant to ampicillin (82.6%), amoxicillin-clavulanate (56.5%), all the third generation cephalosporins including combinations (range 76.1% − 97.8%), ciprofloxacin (67.4%), levofloxacin (63.0%), gentamicin (65.2.0%), amikacin (63.0%), all carbapenems (range 69.6% − 84.8%) and fosfomycin (65.2%), but less than 50% were resistant to piperacillin-tazobactam (45.7%), chloramphenicol (39.1%), clindamycin (47.8%) and linezolid (23.9%), while all (100%) isolates were sensitive to vancomycin.

Table 3.

Antibiotic susceptibility of Gram-positive bacterial isolates to selected antibiotics

Isolates	Antibiotic class
	Penicillin			Cephalosporin					Aminoglycoside		Fluoroquinolone		Amphenicol	Carbapenem			Macrolide		Glyco- peptide	Oxazolidinone	Phosphonic acid
	AMP (%)	AMC (%)	PEP (%)	FOX (%)	CRO (%)	CFX (%)	CAZ (%)	CAZ-AVB (%)	GEN (%)	AMK (%)	CIP (%)	LVX (%)	CHL (%)	IMP (%)	ETP (%)	MEM (%)	ERY (%)	CLI (%)	VAN (%)	LNZ (%)	FOS (%)
Staphylococcus aureus S R	5 (13.2)	15 (39.5)	21 (55.3)	11 (28.9)	1 (2.6)	3 (7.9)	2 (5.3)	3 (7.9)	16 (42.1)	16 (42.1)	14 (36.8 )	15 (39.5)	22 (57.9)	29 (76.3)	6 (15.8)	14 (36.8)	11 (28.9)	22 (57.9)	38 (100 )	29 (76.3)	12 (31.6)
	33 (86.8 )	23 (60.5)	17 (44.7)	27 (71.1)	37 (97.4)	35 (92.1)	36 (94.7)	35 (82.1)	22 (57.9)	22 (57.9)	24 (63.2)	23 (60.5)	16 (42.1)	9 (23.7)	32 (84.2)	24 (63.2)	27 (71.1)	16 (42.1)	0	9 (23.7)	26 (68.4)
Enterococcus spp S R	3 (37.5)	5 (62.5)	4 (50)	0	0	0	0	1 (12.5)	0	1 (12.5)	1 (12.5)	2 (25.0)	6 (75.0)	4 (50)	1 (12.5)	0	0	2 (25.0)	8 (100)	6 (75.0)	4 (50)
	5 (62.5)	3 (37.5)	4 (50)	8 (100)	8 (100)	8 (100)	8 (100)	7 (87.5)	8 (100)	7 (87.5)	7 (87.5)	6 (75.0)	2 (25.0)	4 (50)	7 (87.5)	8 (100)	8 (100)	6 (75.0)	0	2 (25.0)	4 (50)
Total GPB S R	8 (17.4)	20 (43.5)	25 (54.3)	11 (23.9)	1 (2.2)	3 (6.5)	2 (4.3)	4 (8.7)	16 (34.8)	17 (37.0)	15 (32.6)	17 (37.0)	28 (60.9)	33 (71.7)	7 (15.2)	14 (30.4)	11 (23.9)	24 (52.2)	46 (100)	35 (76.1)	16 (34.8)
	38 (82.6)	26 (56.5)	21 (45.7)	35 (76.1)	45 (97.8)	43 (93.5)	44 (95.7)	42 (91.3)	30 (65.2)	29 (63.0)	31 (67.4)	29 (63.0)	18 (39.1)	13 (28.3)	39 (84.8)	32 (69.6)	35 (76.1)	22 (47.8)	0	11 (23.9)	30 (65.2)

AMP Ampicillin, AMC Amoxicillin-clavulante, FOX Cefoxitin, CRO Ceftriaxone, CFX Cefotaxime, CAZ Ceftazidime, CAZ-AVB Ceftazidime-avibactam, GEN Gentamicin, AMK Amikacin, CIP Ciprofloxacin, LVX Levofloxacin, CHL Chloramphenicol, IMP Imipenem, ETP Ertapenem, MEM Meropenem, ERY Erythromycin, CLI Clindamycin, VAN Vancomycin, LNZ Linezolid, FOS Fosfomycin, S Sensitive, R Resistant

In Table 4, more than 50% of the Gram-negative isolates were resistant to the penicillins including combination (range 54.8% − 95.9%), all the third generation cephalosporins including combinations (range 54.% − 91.3%), ciprofloxacin (62.7%), gentamicin (77.0%), chloramphenicol (51.5%), aztreonam (84.1%), ertapenem (54.0%), tetracycline (68.3%), doxycycline (57.1%), and minocycline (61.9%), but less than 50% were resistant to amikacin (31.0%), levofloxacin (36.5%), imipenem (28.6%), meropenem (40.5%), tigecycline (35.7%) and colistin (16.7%).

Table 4.

Antibiotic susceptibility of Gram-negative bacterial isolates to selected antibiotics

Isolates	Antibiotic class
	Penicillin			Cephalosporin				Aminoglycoside		Fluoroquinolone		Amphenicol	Aztreonam	Carbapenem			Cycline				Polymyxin
	AMP (%)	AMC (%)	PEP (%)	CRO (%)	CFX (%)	CAZ (%)	CAZ-AVB (%)	GEN (%)	AMK (%)	CIP (%)	LVX (%)	CHL (%)	AZM (%)	IMP (%)	ETP (%)	MEM (%)	TET (%)	DOX (%)	MNX (%)	TGY (%)	COL (%)
Enterobacterales S R	4 (4.1)	11 (11.3 )	39 (40.2)	10 (10.3)	9 (10.2)	17 (17.5 )	47 (48.5)	22 (22.7)	71 (73.2 )	31 (32.0)	63 (64.9)	47 (48.5)	14 (14.4)	75 (77.3)	55 (56.7)	62 (63.9)	30 (30.9)	31 (32.0 )	29 (29.9)	61 (62.9)	85 (87.6)
	93 (95.9 )	88 (88.7)	58 (59.8)	87 (89.7)	88 (89.8)	80 (82.5)	50 (51.5)	75 (77.3 )	26 (26.8)	66 (68.0)	34 (35.1 )	50 (51.5)	83 (85.6)	22 (22.7)	42 (43.3)	35 (36.1)	67 (69.1)	66 (68.0)	68 (70.1)	36 (37.1)	12 (13.4)
Pseudomonas spp S R	NT	3 (13.0)	14 (60.9)	4 (17.4)	2 (8.7)	10 (43.5)	8 (34.8)	6 (26.1)	14 (60.9)	12 (52.2)	13 (56.5)	NT	6 (26.1)	12 (52.2)	3 (13.0)	10 (43.5)	8 (34.8)	18 (78.3)	17 (73.9)	15 (65.2)	15 (65.2)
	NT	20 (87.0)	9 (39.1)	19 (82.6)	21 (91.3)	13 (56.5 )	15 (65.2)	17 (73.9)	9 (39.1)	11 (47.8)	10 (43.5)	NT	17 (73.9 )	11 (47.8 )	20 (87.0)	13 (56.5)	15 (65.2)	5 (21.7)	6 (26.1)	8 (34.8)	8 (34.8)
Acinetobacter spp S R	NT	2 (33.3)	4 (66.7)	0	0	4 (66.7)	3 (50)	1 (16.7)	2 (33.3)	4 (66.7)	4 (66.7)	NT	0	3 (50)	0	3 (50)	2 (33.3)	5 (83.3)	2 (33.3)	5 (83.3)	5 (83.3)
	NT	4 (66.7)	2 (33.3)	6 (1001)	6 (100)	2 (33.3)	3 (50)	5 (83.3)	4 (66.7)	2 (33.3)	2 (33.3)	NT	6 (100)	3 (50)	6 (100)	3 (50 )	4 (66.7)	1 (16.7)	4 (66.7)	1 (16.7)	1 (16.7)
Total GNB S R	4 (4.1)	16 (12.7)	57 (45.2)	14 (11.1)	11 (8.7)	31 (24.6)	58 (46.0)	29 (23.0)	87 (69.0)	47 (37.3)	80 (63.5)	47 (48.5)	20 (15.9)	90 (71.4)	58 (46.0)	75 (59.5)	40 (31.7)	54 (42.9)	48 (38.1)	81 (62.3)	105 (83.3)
	93 (95.9)	110 (87.3)	69 (54.8)	112 (88.9)	115 (91.3)	95 (75.4)	68 (54.0)	97 (77.0)	39 (31.0)	79 (62.7)	46 (36.5)	50 (51.5)	106 (84.1)	36 (28.6)	68 (54.0)	51 (40.5)	86 (68.3)	72 (57.1)	78 (61.9)	45 (35.7)	21 (16.7)

AMP Ampicillin, AMC Amoxicillin-clavulante, FOX Cefoxitin, CRO Ceftriaxone, CFX Cefotaxime, CAZ Ceftazidime, CAZ-AVB Ceftazidime-avibactam, GEN Gentamicin, AMK Amikacin, CIP Ciprofloxacin, LVX Levofloxacin, CHL Chloramphenicol, AZM Aztreonam, IMP Imipenem, ETP Ertapenem, MEM Meropenem, TET Tetracycline, DOX Doxycycline, MNX Minocycline, TGY Tigecycline, COL Colistin, NT Not Tested, S Sensitive, R Resistant

Table 5 shows the frequency distribution of the bacteria pathogens with respect to selected resistance profiles, with 71.1% of the GNB being ESBL producers, 59.2% were carbapenem resistant, 16.0% were colistin resistant, and 96.0% were multi-drug resistant (resistant to 3 or more classes of antimicrobial agents). For GPB, 71.1% were methicillin-resistant S. aureus (MRSA), 12.9% were inducible clindamycin-resistant S. aureus, and 95.7% of the GPB were MDR.

Table 5.

Descriptive statistics of bloodstream infection caused by bacterial pathogens of interest with specific resistance detected

Specific resistance		Bacterial isolates (%)			Total
Gram-negative BSI		Enterobacterales (n = 97)	Acinetobacter (n = 6)	Pseudomonas (n = 23)	GNB (%)
ESBL (n = 121)	Positive	65 (70.7)	4 (66.7)	17 (73.9)	86 (71.1)
	Negative	27 (29.3)	2 (33.3)	6 (26.1)	35 (28.9)
Carbapenem (n = 125)	Sensitive	48 (50.0)	1 (16.7)	2 (8.7)	51 (40.8)
	Resistant	48 (50.0)	5 (83.3)	21 (91.3)	74 (59.2)
Colistin (n = 125)	Sensitive	85 (88.5)	5 (83.3)	15 (65.2)	105 (84.0)
	Resistant	11 (11.5)	1 (16.7)	8 (34.8)	20 (16.0)
MDR GNB (n = 126)	Yes No	94 (96.9) 3 (3.1)	5 (83.3) 1 (16.7)	22 (95.7) 1 (4.3)	121 (96.0) 5 (4.0)

Gram-positive BSI		Staphylococcus aureus (n = 38)	Enterococcus spp (n = 8)
MRSA (n = 38)	MRSA	27 (71.1)	-		-
	MSSA	11 (28.9)	-		-
Inducible clindamycin (S. aureus only n = 31) MDR GPB (n = 46)	Resistant	4 (12.9)	-		-
	Sensitive Yes No	27 (87.1) 36 (94.7) 2 (5.3)	- 8 (100.0) 0		- 44 (95.7) 2 (4.3)

MDR Multi-drug resistance, ESBL Extended spectrum b-lactamase, GNB Gram-negative bacteria, GPB Gram-positive bacteria, MRSA Methicillin resistant Staphylococcus aureus

Comparative characteristics and outcomes of patients with BSI and ‘no BSI’ control

Table 6 presents the comparative clinical characteristics and outcomes of patients with different BSI onset (community and healthcare onset) and those with ‘no BSI/culture negative BSI’ that served as a control. The length of hospital stays (LOS in days) was significantly longer in patients with HO-BSI (27.9 ± 26.7 days) compared to patients with CO-BSI (12.0 ± 10.5 days) and the ‘no BSI’ control (19.4+/-15.1) (p < 0.0001). In-hospital mortality was significantly higher in patients with CO-BSI (23.3%) and HO-BSI (21.4%) than ‘no BSI’ control (7.1%) (p = 0.0374). Similarly, 30-day mortality was also higher in CO-BSI (23.3%) and HO-BSI (21.4%) than ‘no BSI’ control (8.9%), although this difference was not statistically significant (p = 0.0673).

Table 6.

Descriptive summary of patients with BSI and no BSI controls

Characteristic	CO-BSI (n = 116)	HO-BSI (n = 56)	No BSI (n = 56)	p value
Age (years): mean+/-SD	13.6+/-25.0	19.6+/-26.1	17.3+/-22.6	0.2976
Age category
Neonates (1–28 days)	68 (58.6)	22 (39.3)	23 (41.1)
Children (1 month-17 years)	21 (18.1)	13 (23.2)	12 (21.4)
Adults (≥ 18 years)	27 (23.3)	21 (37.5)	21 (37.5)
Sex
Male	61 (52.6)	30 (53.6)	30 (53.6)
Female	55 (47.4)	26 (46.4)	26 (46.4)
Discharge status
Alive	89 (76.7)	43 (76.8)	51 (91.1)
Dead	27 (23.3)	12 (21.4)	4 (7.1)	0.0374*
Status at day 30
Alive	87 (75.0)	44 (78.6)	51 (91.1)
Death	27 (23.3)	12 (21.4)	5 (8.9)	0.0673
Unknown	2 (1.7)	0	0
Length of hospital stay	12.0+/-10.5	27.9+/-26.7	19.4+/-15.1	< 0.0001*
qSOFA score (adult only; n = 69)
≥2	14 (51.9)	13 (61.9)	9 (42.9)	0.4657
< 2	13 (48.1)	8 (38.1)	12 (57.1)
uCCI score (adult only; n = 69)
≥ 4	2 (7.4)	1 (4.8)	0	0.4676
< 4	25 (92.6)	20 (95.2)	19 (100.0)
Pitts score (adult only; n = 69)
≥ 4	4 (14.8)	3 (14.3)	2 (9.5)	0.8468
< 4	23 (85.2)	18 (85.7)	19 (90.5)

Values are presented as mean ± standard deviation (SD) or frequency (percentage)

p-values indicate statistical significance between groups; p  < 0.05 considered significant statistical significant data*

Mortality outcomes and length of hospital stay of patients with BSI

In-hospital mortality was significantly lower in neonates with BSI (14.4%) than in children (26.5%) and adults (35.4%) with BSI (χ2 = 8.201, p = 0.0166). The mean LOS is also shorter in neonates (12.3 ± 9.9 days) than in children (19.1 ± 16.2 days) and adults (male 24.4 ± 27.2 days and female 25.4 ± 30.1 days), and this difference was statistically significant for neonates compared to adult females (p = 0.0011) and to adult males (p = 0.0018). Further outcome in-hospital mortality was slightly lower (although insignificant, p = 0.8497) in patients with HO-BSI (21.4%) compared to CO-BSI (23.3%), but LOS was longer in HO-BSI (median LOS of 18 days) than in CO-BSI (median LOS of 10 days). Also, in-hospital mortality was higher (although insignificant, p = 0.6819) in patients with GNB (23.8%) compared to GPB (19.6%) BSI, but LOS was similar in both.

Modelling of the impact of resistance BSI on outcomes

By causal analysis using counterfactual approaches to determine the impact of susceptible and resistant BSI on in-hospital mortality, as shown in Table 7, Resistant BSI was associated with an additional 7 deaths per 100 BSI patients compared to susceptible BSI (although this was not statistically significant). For HO-BSI patients, Resistant infection was associated with an additional 10 deaths per 100 cases compared to “no BSI counterfactual; however, this finding was also not statistically significant. Due to the small sample size, it was not possible to reliably estimate the effect of susceptible HO-BSI compared to the ‘no BSI’ counterfactual.

Table 7.

Causal analysis using counterfactual approaches to determine the impact (in-hospital mortality) of resistant and susceptible BSIs

BSI	Test type	Estimate	95% CI
Resistant-BSI vs. Susceptible-BSI on in-hospital mortality	Counterfactual framework N = 171	RD: 0.071	-0.07, 0.21
Resistant HO-BSI vs. ‘no BSI’ on in-hospital mortality	Counterfactual framework N = 103	RD: 0.10	-0.11, 0.97
Susceptible HO-BSI vs. ‘no BSI’ on in-hospital mortality	Counterfactual framework N = 60	Did not converge due to a small sample size and sparse events. Estimates are therefore not reported

Excess risk of in-hospital mortality for resistant and susceptible BSI

With discharge alive as competing event, Cog regression model with sub-distribution hazard ratio (_sdHR) showed that excess risk of in-hospital mortality for HO-BSI was almost twice (HR = 1.7) that for CO-BSI after adjusting for co-variates (age, sex, qSOFA, Pitts score and uCCI in adults). For resistant versus susceptible BSI, excess risk of in-hospital mortality for MRSA BSI was almost 4 times (HR = 4.1) that for MSSA-BSI, and for clindamycin-R S. aureus BSI, was almost 7 times (HR = 6.8) that for clindamycin-S S. aureus BSI (although not statistically significant). However, there was no excess risk in-hospital mortality for ESBL, carbapenem-R and colistin-R BSI over their susceptible counterparts.

Extra length of hospital stays in resistant and susceptible BSI

Using Cog regression with a multi-state model, patients with HO-BSI had an extra LOS of 2.28 days compared to patients with CO-BSI. Resistant BSI caused extra LOS of 1.72 days compared to susceptible BSI.

Healthcare cost outcomes

The median monthly incomes were comparatively similar in patients with BSI (CO-BSI and HO-BSI) and the ‘no BSI’ control (p = 0.5813). However, healthcare costs were significantly higher in patients with HO-BSI (median cost of 125,500) than those with CO-BSI (median cost of 70,800) and the ‘no BSI’ control (median cost of 75,000) (p < 0.0001). Catastrophic health expenditure was determined at three different thresholds: 20%, 30% and 40%. At 20% threshold, 98.2% and 96.3% of patients with ‘no BSI’ and HO-BSI, respectively, had catastrophic health expenditures, that was significantly higher than those of patients with CO-BSI (89.7%) (p = 0.0128). At 30% threshold, 90.7% of patients with HO-BSI had catastrophic health expenditures, which was significantly higher than the ‘no BSI’ control (82.1%) and CO-BSI (75.9%) (p = 0.0458). At 40% threshold, 87% of patients with HO-BSI had catastrophic health expenditures, which was higher (insignificant, p = 0.1075) than the ‘no BSI’ control (71.4%) and CO-BSI (62.9%) Table 8.

Table 8.

Comparative statistics of catastrophic health expenditures in patients with and without BSI

Type of BSI	Number	40% threshold		30% threshold		20% threshold
		% with CHE	% without CHE	% with CHE	% without CHE	% with CHE	% without CHE
CO: BSI	116	73 (62.9)	43 (37.1)	87 (75.0)	29 (25.0)	104 (89.7)	12 (10.3)
HO: BSI	54	47 (87.0)	7 (13.0)	49 (90.7)	5 (9.3)	52 (96.3)	2 (3.7)
No BSI	56	40 (71.4)	16 (28.6)	46 (82.1)	10 (17.9)	55 (98.2)	1 (1.8)
Total	226	160 (70.8)	66 (29.2)	182 (80.5)	44 (19.5)	211 (93.4)	15 (6.6)

For a threshold of 20%, the number with CHE was significantly higher for the control and HO-BSI compared to CO-BSI (p = 0.0128) For a threshold of 30%, the number with CHE was significantly higher for HO-BSI group compared to CO-BSI and the control (p = 0.0458) For a threshold of 40%, the number with CHE was not significantly different between the BSI groups and the control (p = 0.1075)

Impoverishment status (poverty level) of patients with BSI and the ‘no BSI’ control who spent out-of-pocket (OOP) was determined at three thresholds as shown in Table 9: $1.5, $2 and $2.5. At the $1.5 threshold, 96.3% of patients with HO-BSI who spent OOP were impoverished, significantly higher (p = 0.000) than patients with CO-BSI (86.2%) and ‘no BSI’ control (82.1%). At the $2.0 threshold, 96.3% and 96.4% of patients with HO-BSI and ‘no BSI’ respectively who spent OOP were impoverished, significantly higher (p = 0.0000) than patients with CO-BSI (90.1%). At the $2.5 threshold, 96.3% and 98.2% of patients with HO-BSI and ‘no BSI’ respectively who spent OOP were impoverished, significantly higher (p = 0.0000) than patients with CO-BSI (93.1%).

Table 9.

Comparative statistics of impoverishment status (poverty level) of patients with BSI and no BSI control who spend out-of-pocket

Type of BSI	Number	$2.5 threshold		$2 threshold		$1.5 threshold
		% Impov’d (below pl)	% without Impov’d (above pl)	% Impov’d (below pl)	% without Impov’d (above pl)	% Impov’d (below pl)	% without Impov’d (above pl)
CO: BSI	116	108 (93.1)	8 (6.8)	105 (90.1)	11 (9.5)	100 (86.2)	16 (13.8)
HO: BSI	54	52 (96.3)	2 (3.7)	52 (96.3)	2 (3.7)	52 (96.3)	2 (3.7)
No BSI	56	55 (98.2)	1 (1.8)	54 (96.4)	2 (3.6)	46 (82.1)	10 (17.9)
Total	226	215 (95.2)	11 (5.8)	211 (93.4)	15 (6.6)	198 (87.6)	28 (12.4)

For threshold of 1 dollar, the number impoverished is significantly higher for HO-BSI compared to CO-BSI and the control (p = 0.0000) For threshold of 2 dollars, the number impoverished is significantly higher for HO-BSI and the control compared to CO-BSI group (p = 0.0000) For threshold of 2.5 dollars, the number impoverished is significantly higher for HO-BSI and the control compared to CO-BSI group (p = 0.0000) Note: Exchange rate of dollar to naira was based on the average exchange rate of 1531.73 for 2025

Discussion

The overall prevalence of pathogen-confirmed BSI in the study was 18.1%, which is within the range usually reported for many low-and-middle-income countries [27–32]. This pattern is consistent with previous studies reporting a higher burden of neonatal BSI, including findings from Nigeria by Oyekale et al. (23.8%) [27], Iregbu et al. [29]. (22.0%), Ogunkunle et al. (19.0%) [28], and Uzodimma et al. (35.0%) [30], as well as studies from Ethiopia by Sorsa et al. (29.3%) [33] and Egypt by El-Din et al. (40.7%) [34].

The observed prevalence of bacteria-confirmed BSI is influenced by several methodological and clinical factors, including the volume and number of blood samples collected, laboratory techniques employed for pathogen isolation, prior antibiotic exposure, and adherence to aseptic collection procedures, which may affect contamination rates [35]. In the present study, only a single set of blood cultures was obtained; however, adequate blood volumes (20 mL per collection) were ensured, and an automated blood culture system was utilized to enhance pathogen recovery. Despite these measures, prior antibiotic use before hospital presentation could not be reliably excluded, given the widespread practice of unregulated antibiotic use in our environment.

In this study, the prevalence of community-onset BSI (12.2%) exceeded that of healthcare-onset BSI (5.9%). This finding is consistent with the report by Naylor et al. [36], in which hospital-onset infections accounted for less than one-quarter of Escherichia coli bacteraemia cases. Collectively, these observations suggest that a substantial proportion of BSIs originate in the community rather than within healthcare settings. While strengthening infection prevention and control measures in hospitals remains essential, interventions limited to hospital settings alone are unlikely to substantially reduce the overall burden of BSI. Instead, greater emphasis should be placed on community-level strategies, including the early diagnosis and effective management of common infections such as urinary tract infections and pneumonia, which may otherwise progress to invasive bloodstream infections.

Bloodstream infections in this study were predominantly caused by Gram-negative bacteria (73.3%), compared to Gram-positive organisms (26.9%) across all age groups. However, Klebsiella spp. (22.1%) and Staphylococcus aureus (22.1%) emerged as the most frequently isolated individual pathogens. This distribution is consistent with previous reports from Nigeria by Oyekale et al. [27] and findings from other settings [37, 38], but contrasts with studies among neonates in Nigeria by Ogunkunle et al. [28], Ethiopia by Sorsa et al. [33], Egypt by El-Din [34]., and by other researchers in different settings [39–43] where Gram-positive organisms, particularly S. aureus and coagulase-negative staphylococci, predominated. The distribution of causative pathogens in BSI is known to vary widely across geographic regions and over time. Historically, Gram-positive bacteria were the leading causes of sepsis in the pre-antibiotic era, through the 1950s. This pattern shifted toward Gram-negative predominance following the widespread introduction of antibiotics between the 1960s and 1980s. By the 1980s, Gram-positive organisms, especially Staphylococcus spp., were again reported to account for more than half of sepsis cases [44, 45]. In contemporary hospital settings, however, the increased use of invasive procedures and medical devices may create selective pressure favouring Gram-negative pathogens, which could explain their predominance in the present study.

Antimicrobial susceptibility testing demonstrated an alarmingly high prevalence of multidrug resistance, affecting 95.7% of Gram-positive and 96.0% of Gram-negative isolates. Among Gram-positive pathogens (Staphylococcus aureus and Enterococcus spp.), resistance exceeded 50% across nearly all major antibiotic classes, including penicillins, third-generation cephalosporins, carbapenems, and fluoroquinolones, indicating extensive therapeutic compromise. Susceptibility was relatively preserved only for piperacillin–tazobactam, chloramphenicol, clindamycin, and linezolid, while universal susceptibility to vancomycin highlights its continued role as a critical last-line agent. This observation has been reported by other researchers in Nigeria [27] and in other parts of sub-Saharan Africa [46] as well. Gram-negative isolates exhibited a similarly severe resistance profile, with widespread resistance to first-line and commonly used agents, particularly β-lactams and cephalosporins, consistent with the likely contribution of extended-spectrum β-lactamase and carbapenemase-mediated mechanisms. Residual activity was largely confined to a limited number of agents, including colistin, tigecycline, amikacin, and carbapenems. underscoring a progressive shift towards reliance on last-resort therapies.

ESBL production was detected in 71.1% of Gram-negative isolates, putative carbapenem resistance in 59.2%, and methicillin resistance in Staphylococcus aureus (MRSA) in 71.1% of isolates. The high frequency of ESBL production plausibly underpins the marked resistance observed among Enterobacterales to third-generation cephalosporins. These estimates substantially exceed national surveillance data from Nigeria (2017–2022) [47], which reported rates of 19% for both ESBL-producing Enterobacterales and MRSA, and 4% for carbapenem-resistant Enterobacterales, indicating a concerning local intensification of antimicrobial resistance. The magnitude of carbapenem resistance observed further suggests the emergence and dissemination of carbapenemase-producing organisms, while the high prevalence of MRSA reflects sustained transmission of resistant Staphylococcus aureus strains, together compounding therapeutic constraints. Despite this, susceptibility to selected agents remained relatively preserved, with low resistance to clindamycin (12.9%) among Staphylococcus aureus and to colistin (16.0%) among Gram-negative isolates, consistent with observations from a comparable regional setting [46]. However, this residual activity is largely confined to a limited subset of agents, underscoring a progressively constrained therapeutic armamentarium. Overall, the resistance profile reflects sustained and likely intensifying antimicrobial selection pressure, with direct implications for empirical treatment efficacy, infection control practices, and the prioritisation of robust surveillance systems.

Hospital-onset BSI was associated with a markedly greater clinical and economic burden, with significantly higher in-hospital mortality, prolonged hospitalisation, and increased healthcare costs. In competing-risk analyses accounting for discharge alive, subdistribution hazard modelling demonstrated that hospital-onset BSI conferred an almost two-fold higher risk of in-hospital mortality compared with community-onset infection (sdHR 1.7). This excess risk likely reflects the cumulative impact of healthcare exposures, particularly in high-dependency settings such as intensive care units, where invasive devices, repeated instrumentation, and underlying disease severity synergistically increase both susceptibility to infection and the risk of adverse outcomes.

Notably, although BSI prevalence was highest among neonates, mortality was greater in children and adults, suggesting important differences in case mix and care processes. The relatively lower mortality observed in neonates may reflect the benefits of protocol-driven care, early recognition, and closer physiological monitoring in neonatal units. Collectively, these findings underscore the disproportionate impact of hospital-onset BSI on patient outcomes and health system resources, and highlight the need for targeted prevention strategies, optimised supportive care, and early risk stratification in hospitalised populations.

With discharge alive as a competing event, the Cog regression model with sub-distribution hazard ratio (_sdHR) showed that the risk of in-hospital mortality for MRSA BSI was almost 4 times (HR = 4.1) that for MSSA-BSI, and almost 7 times (HR = 6.8) for clindamycin-resistant compared to clindamycin-sensitive S. aureus BSI. In the study by Lambert et al. [48], risk of death (hazard ratio) associated with antimicrobial resistance (i.e., additional risk of death to that of the infection) was 1·2 (0·9–1·5) for bloodstream infections for a combination of four microorganisms (A. baumannii, P. aeruginosa, E. coli, and S. aureus), and was highest for S. aureus BSI [HR 1·6 (1·1–2·3)]. Our regression modelling with a counterfactual framework showed that an additional 7 deaths/100 BSI patients (Risk Difference of 0.071) were associated with resistant BSI compared to susceptible BSI, and an additional 10 deaths/100 HO-BSI (RD 0.10) were associated with resistant HO-BSI compared to susceptible HO-BSI. This indicates that antimicrobial resistance contributes to an extra burden (in-hospital mortality and extra length of hospital stay) in patients with bloodstream infections in our setting.

The observed lower mortality among patients with ESBL-producing and carbapenem-resistant infections compared with susceptible cases may be explained by several factors. The small number of isolates in these groups likely limited statistical power, resulting in unstable estimates and wide variability in outcome measures. In addition, improved clinical recognition and earlier initiation of appropriate antimicrobial therapy may have contributed to better outcomes in these patients. Differences in infection source and underlying disease severity may also have influenced mortality patterns, as some infections may have been less severe or more amenable to effective treatment. Furthermore, potential selection bias and heterogeneity in patient characteristics may have affected the observed associations. Therefore, these findings should be interpreted with caution.

In competing-risk analyses accounting for discharge alive, MRSA bloodstream infection (BSI) was associated with an approximately four-fold higher risk of in-hospital mortality compared with methicillin-susceptible Staphylococcus aureus (MSSA) BSI, while clindamycin-resistant S. aureus BSI was associated with an almost seven-fold increased risk relative to susceptible infections. These effect estimates are substantially higher than those reported in multicentre studies, where the excess mortality attributable to antimicrobial resistance has generally been more modest. For example, Lambert et al. reported a hazard ratio of 1.2 for mortality associated with antimicrobial resistance in bloodstream infections caused by a combination of Acinetobacter baumannii, Pseudomonas aeruginosa, Escherichia coli, and Staphylococcus aureus, suggesting a smaller incremental risk of death attributable to resistance compared with the markedly higher estimates observed in the present study. These differences may be explained by variations in study population, infection severity, healthcare setting, and antimicrobial stewardship practices, as well as potential residual confounding and smaller sample sizes in the present study, which may have contributed to wider effect estimates and possible overestimation of risk.

Using a counterfactual framework, we estimated that antimicrobial resistance accounted for an additional 7 deaths per 100 patients with BSI and 10 additional deaths per 100 patients with hospital-onset BSI. Together, these findings suggest that antimicrobial resistance is not only highly prevalent but also translates into a clinically meaningful excess risk of mortality and healthcare utilisation, underscoring its critical role in shaping patient outcomes in this setting.

In competing-risk analyses accounting for discharge alive, subdistribution hazard modelling showed that MRSA BSI was associated with a more than four-fold increased risk of in-hospital mortality compared with MSSA BSI (sdHR 4.1), while clindamycin-resistant S. aureus BSI was associated with an almost seven-fold increased risk compared with clindamycin-susceptible infections (sdHR 6.8). These estimates exceed those reported by Lambert et al., who observed a more modest excess mortality associated with antimicrobial resistance across major bloodstream pathogens, with the highest effect observed for S. aureus (HR 1.6 [1.1–2.3]).

Using a counterfactual modelling framework, we estimated that antimicrobial resistance was associated with an additional 7 deaths per 100 patients with BSI (risk difference 0.071), and 10 additional deaths per 100 patients with hospital-onset BSI (risk difference 0.10), compared with susceptible infections. Together, these findings indicate that antimicrobial resistance confers a measurable and clinically important excess mortality and hospital burden in patients with bloodstream infections in this setting.

In our study, healthcare costs were higher for patients with healthcare onset-BSI (median cost of 125,500; IQR 76,000 -184600) compared to those with community onset BSI (median cost of 70,800; IQR 51,600 − 114,950) and those without BSI (median cost of 75,000; IQR 56,750-112400), indicating a much higher economic burden for healthcare associated BSIs, although the wide interquartile range among patients with healthcare onset BSI reflects large variability, suggesting severe or complicated cases that require hospital drive cost. This higher cost in those with healthcare onset BSI may be due to prolonged hospitalisation, bed occupancy, procedures, medications and other direct and indirect costs incurred in the process of care delivery in the hospital. Furthermore, healthcare costs were higher for patients with Gram-negative BSIs compared to Gram-positive BSIs, and for all selected resistant BSIs (ESBL, carbapenem resistance, colistin resistance, MRSA and clindamycin resistance) compared to their susceptible counterparts. Estimates of healthcare costs of bacterial BSI and antimicrobial resistance have been well reported in many developed countries. Naylor et al.,³⁹ estimated the annual excess cost of over £14 million and over £500,000 for E. coli bacteraemia and ‘resistance to at least one tested antibiotic’ respectively. However, to the best of our knowledge, our study is the first to provide local data on healthcare costs of bacterial BSI and antimicrobial resistance in Nigeria.

Our study showed that all the patients with community onset BSI and those without BSI, and 96.4% of those with healthcare onset BSI, paid out-of-pocket (OOP), indicating near universal OOP practice among patients in this study. In Nigeria, the National Health Insurance Scheme (NHIS) enrolls less than 5% of the populace, and medical care, including antibiotics and diagnostics, is paid for through OOP expenses [49]. Our findings highlight the absence of effective financial risk protection and expose households to substantial economic shocks. Without systemic reforms to reduce OOP, BSI and antimicrobial-resistant infections will continue to generate significant financial hardship even among patients requiring hospital admissions.

Catastrophic health expenditure (CHE) was highly prevalent across the BSI categories and increased sharply at lower thresholds. At the 40% threshold, CHE affected 70.8% of all the patients irrespective of resistance status, with the highest burden among patients with healthcare onset-BSI (87.0%). At 30% and 20% thresholds, CHE prevalence increased sharply across the two BSI categories and also among patients without BSI, CHE prevalence was high, indicating weak financial risk protection among the population. However, patients with hospital-onset BSI consistently faced the highest CHE burden, reflecting greater treatment intensity required and additional costs. These findings demonstrate that BSI and AMR-related care frequently exceeds household capacity to pay, emphasizing the urgent need for stronger financial risk protection mechanisms among these patients.

Impoverishment following OOP payments for the BSI categories was widespread across all poverty thresholds, with the hospital-onset BSI category bearing the greatest burden, reflecting severe economic consequences of antibiotic therapy. At the $2.5/day threshold, over 95% of the patients were impoverished after OOP payments. These results show that healthcare expenditure for BSI and AMR do not only strains household finances but also actively drives poverty. This result portrays that strengthening financial protection is critical to preventing health-related economic consequences.

Conclusion

Our study showed that antibiotic-resistant bacterial BSIs when compared to susceptible bacterial BSIs, are associated with higher mortality, extra length of hospital stays and excess healthcare costs. There is therefore the need to enact policies that will reduce the burden of resistant BSI which should focus not only on improving infection prevention control and antimicrobial stewardship practices in the hospital settings but also in the community, where reduction in less serious community-onset infections such as urinary tract infections, pneumonia and other infections that can progress to sepsis and bloodstream infections, will significantly reduce the incidence of community-onset BSIs, which was high in this study.

Limitations of the study

Our study was limited by the fact that it was conducted in a single institution, which makes generalization of our findings difficult. The low number of patients with bacterial-confirmed BSI, which though falls within the range that has been reported in many low-and-middle-income countries, affected subgroup analysis, especially with respect to antimicrobial susceptibility of isolated bacterial pathogens. Furthermore, bacteria identified were limited to those commonly known agents of sepsis and BSI, while others, such as fungi, coagulase-negative staphylococci and atypical bacteria were not included. Finally, this study includes both neonates and adults in a combined analysis. These populations differ significantly in immune status, disease spectrum, and causative pathogens. Stratification by age group may have provided more precise estimates; however, this was limited by sample size constraints. Future studies should consider age-specific analyses to better define these differences.

In spite of these shortcomings, the prospective nature of our study enabled an accurate estimation of the incidence of BSI in the study area. By including all categories of patients (neonates, children and adults), direct comparison of the BSI characteristics between the study groups was possible. The screening criteria for BSI in each category of patients were stringent, which enabled enrolment of only highly probable BSI patients and thereby reduced wastage of resources. The use of carefully matched “no BSI control” cohort allowed accurate assessments of outcomes (mortality, length of stay and healthcare costs) with the BSI cohorts. Modelling of outcomes using different Cox regression models correctly adjusted for the time-dependency of exposure (for healthcare onset-BSI), time-varying co-founders and competing risks.

Authors’ contributions

IN, JI, DE, BA, AA, TO, RY, ST, KA, SM, MA, TO, DE, JI conceived and led the study. CO, JA, HE, DI, OJ, UN, VB, MA, IN, ST, UN and DE contributed to the study design and methodology. JA, CO, WO, VB, DU, SM, MO, DK, ADD, ID, OG, CM, EA, AA, MA, TO and CS participated in data acquisition and field coordination. ST, DI, UN, MA, OJ, TO, EA, AA, RY, IN, BA, CO, KA, AG, CS IN, HE and DE conducted data analysis and interpretation. IN, BA, WO, ST, DA, OJ, UN, SM, DU, HE, RY and CO drafted the initial manuscript. BA, CO, ST, KA, JA, HE, WO, DU, SM, AA, MO, DK, AG, ID, OG, CMO, EA, DU, Ak, TB, VB, CS, UN, JI and DE critically reviewed the manuscript for important intellectual content. All authors read and approved the final manuscript.

Funding

This work was supported by the UK Department of Health and Social Care (DHSC) through the Fleming Fund using UK aid. The funder had no role in study design, data collection, analysis, interpretation, or manuscript writing.

Data availability

The datasets generated and/or analysed during the current study are available from the corresponding author on reasonable request.

Declarations

Ethics approval and consent to participate

This study was conducted in compliance with the principles of the Helsinki Declaration. Ethical approval for this study was obtained from the Health Research Ethics Committee of the University of Port Harcourt Teaching Hospital (UPTH/ADM/90/S.11/VOL.XI/1822) before data collection. Written informed consent was obtained from all participants, and for participants younger than 16, consent to participate was obtained from their parents or legal guardians.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.