The Economics of Antibiotic Resistance: A Systematic Review and Meta-analysis Based on Global Research

Abstract

Background

Antibiotic resistance (ABR) is a growing global health threat; reliable evidence on its impact is crucial for prioritising public health interventions.

Objective

This study provides an updated, systematic review and meta-analysis to determine the true effect size of resistant infections on economic and clinical outcomes. It also evaluates methodologies used in ABR economic literature, offering recommendations for improving future research.

Methods

Following PRISMA guidelines, 11,252 articles published between 2000 and 2022 were reviewed from several databases. Studies were included if they reported the economic costs of ABR in humans and compared resistant with susceptible infections. Meta-analyses were conducted using random intercept models; standardised mean difference (SMD) was used for length of stay, and odds ratio (OR) for mortality. The Mantel-Haenszel method was applied to obtain pooled estimates.

Results

Results showed that 73% of the studies were conducted in high-income economies, the majority were performed at tertiary care settings (71%) and 67% employed only a hospital perspective. The available evidence indicated that the attributable cost of resistant infections ranged from EUR₂₀₂₂ − 21,629 to EUR₂₀₂₂ 74,452 per patient episode (with Pseudomonas spp. causing the highest costs). The majority of studies (93%) found that patients with ABR incurred higher costs than their susceptible counterparts (72% report statistically significantly higher costs). Results from meta-analysis indicated that, on average, the excess in hospital stay attributable to resistant infections was 8.72 days (95% confidence interval (CI) [6.42; 11.02], SMD = 0.91) and the odds of premature death were significantly higher in the resistance group, with a risk increase of 65% (OR 95% CI [1.44; 1.88]).

Conclusion

The findings of this study take the first steps in providing reliable evidence; they could be valuable to researchers, policymakers and clinicians involved in ABR control and health promotion across countries. Similarly, the reported estimates may prove useful for future modelling studies aimed at assessing the long-term economic impact of ABR.

Supplementary Information

The online version contains supplementary material available at 10.1007/s40258-025-01001-7.

Key Points for Decision Makers

At the patient level, most studies (93%) reported higher costs for resistant infections, with 72% showing statistically significant differences. No study found significantly higher costs for susceptible infections. At the national level, most studies reported notably higher costs from productivity losses due to premature deaths than from direct healthcare costs.

Meta-regression results showed that, on average, resistant infections were associated with an 8.72-day increase in hospital stay and a 65% higher risk of premature death compared to susceptible infections.

Our estimates of economic and health outcomes take the first steps in providing reliable evidence for professionals and policymakers. They also can support future modelling studies and guide more effective antibiotic resistance research.

Introduction

Globally, antibiotic resistance (ABR) is becoming a more significant health concern. It involves the natural process of bacteria adapting to the antibiotics treating them. This adaptation makes antibiotic treatments less effective [1]. Although ABR is a natural phenomenon, the prevalence of resistance is accelerated and intensified by the excessive utilisation of antibiotics [2, 3]. In this regard, the World Health Organization (WHO) declared this problem as one of the ten most significant threats to public health on a worldwide scale [3]. It is estimated that by 2050, the global burden of ABR could reach 1.91 million attributable deaths and 8.22 million associated deaths annually [4]. This trajectory suggests that, without additional and intensified action, the international target of a 10% reduction in AMR mortality—set under the 10–20–30 by 2030 initiative—will not be achieved [4, 5]. Economically, ABR may also reduce Gross Domestic Product (GDP) by 1.1% and it might cost more than USD 1 trillion globally per year beyond 2030, considering a low-impact scenario [6]. This figure is due to the higher consumption of health resources and increased morbidity and mortality experienced by ABR patients.

Most of the scientific literature demonstrates that ABR poses a threat to the medical practice as it potentially complicates infection management and causes higher mortality and healthcare costs [3, 7–9]. In this sense, those patients with resistant infections endure longer hospital stays and consume higher healthcare resources than those with susceptible strains or those with no infection [1, 10–12]. Hence, it is crucial to have an accurate figure and reliable evidence about the burden of ABR so that the health and economic consequences are better understood. In this way, urgent measures can be taken to promote public health. The main objective of this article is to present a comprehensive review of the global economic impact of ABR and to determine the true effect size of resistant infections on economic and clinical outcomes.

Seven systematic reviews on the economic impact of ABR have been carried out [13–19], of which only four incorporate meta-analysis. In general, these studies focus their analyses either on specific countries (and not on global output) or by restricting certain bacteria. Following the Preferred Reporting Items for Systematic Review and Meta-Analysis (PRISMA) guidelines, this article offers an updated systematic review on the economic burden of ABR, from a global research output, including all available bacteria and distinguishing between a large list of antibiotics and sites of infection. In addition, our study includes a significantly larger number of studies for analysis − 75 in total, compared to 29 in Poudel et al. [16], or 40 in Founou et al. [13]. This substantial increase in sample size provides a more comprehensive overview of the existing evidence in the systematic review and enhances the robustness and statistical power of the meta-analysis.

Methodology and Study Design

Search Strategy

This systematic review followed the PRISMA guidelines [20]. Two authors (SS and BC) independently proposed a list of keywords based on the topic, and the final list was agreed upon. Exclusions required at least two out of three votes from the reviewing authors (SS, BC and IL-M). The keyword search was based on search terms identified from previous relevant studies [15–19]. Specifically, two search strategies were agreed upon: (i) based on macroeconomic aspects related to the impact or economic burden of the problem, antibiotic resistance; and (ii) based on specific associated costs (medical, hospital and social). The search considered those studies whose title and/or abstract contained the keywords of either of the two strategies. Table 1 shows how the two strategies were applied in the search engines of the different databases.

Table 1.

Keywords and search strategy

Strategy	Keywords Used in the search engine
1	ABR OR AMR OR antibacterial resistan* OR antibiotic resistan* OR antimicrobial resistan* OR bacteria resistan* OR carbapenem resistan* OR methicillin resistan* OR microbial resistan* OR penicillin resistan* OR resistant infection OR vancomycin resistan* OR macrolide resistan* OR fluoroquinolone resistan* OR aminoglycoside resistan* OR carbapenem resistan* OR cephalosporin resistan* AND cost OR outcome OR impact OR evaluation OR burden OR loss AND economic OR financial OR social OR societal OR socioeconomic OR macroeconomic OR productivity
2	ABR OR AMR OR antibacterial resistan* OR antibiotic resistan* OR antimicrobial resistan* OR bacteria resistan* OR carbapenem resistan* OR methicillin resistan* OR microbial resistan* OR penicillin resistan* OR resistant infection OR vancomycin resistan* OR macrolide resistan* OR fluoroquinolone resistan* OR aminoglycoside resistan* OR carbapenem resistan* OR cephalosporin resistan* AND cost OR expense OR expenditure AND hospital OR medical OR direct OR indirect OR healthcare OR care OR incremental OR additional OR excess OR out-of-pocket

Source: authors’ original work

Eligibility Criteria and PRISMA Flowchart

The articles considered were original studies published in scientific (peer-reviewed) journals in English or Spanish between 2000 and 2022. The search for these articles was carried out in clinical (MEDLINE, CINAHL), economics and business (Business Source Premier) and multidisciplinary databases (Web of Science, SciELO and Scopus). The studies were selected by using the pre-specified eligibility criteria, as shown in Table 2.

Table 2.

Eligibility criteria and PICO

Exclusion criteria	Inclusion criteria	PICO
1. Studies not related to economic outcomes of ABR (e.g., not having information on healthcare, societal cost or treatment costs) or studies related to economic evaluations, but not reporting burden or costs 2. Theoretical studies based on ABR costs, but not reporting own estimations 3. Studies which report only health outcomes of ABR 4. Studies that report only molecular or biology outcomes of ABR 5. Studies related to ABR in animals or plants 6. Studies published in languages other than English or Spanish 7. Letters, notes, editorials and conference reports	1. Original studies reporting burden or economic costs of ABR (e.g., healthcare cost, societal cost…) or economic evaluations reporting relevant information (e.g., per patient costs for treatment due to ABR) 2. Studies comparing “resistant” with “susceptible”/ “sensitive” infections or another suitable comparator 3. Original case control, cohort, observational, randomised controlled trials and modelling studies 4. Studies related to ABR in humans 5. Studies published, in English or Spanish, in peer-reviewed journals	Population  Human   All sexes   All ages   Infected with resistant or susceptible bacteria Intervention  Not applicable
		Comparison  Antibiotic resistant infections vs. antibiotic susceptible infections or another suitable comparator
		Outcome  Economic outcomes such as burden or economic costs (e.g., healthcare, outpatient care, opportunity, treatment, length of stay, mortality…)

Source: authors’ original work

Once the search for each strategy (explained above) had been carried out in the databases, the references were exported to the Rayyan bibliographic manager [21]. A total of 19,048 articles were identified in this review. Of this total, 7783 were duplicates between databases and, therefore, eliminated. The articles that did not meet the objective and subject matter of this systematic review were excluded by reading the title and abstract. We discarded those that did not directly address antibiotic resistance (n = 6319), those that did not refer to humans (n = 1145), and those that did not evaluate economic aspects (n = 3412). Studies published in languages other than English or Spanish were also discarded despite initial screening (n = 3), as were books, chapters or other formats different from original articles (n = 174). These documents were discarded since they had not been peer reviewed.

For the remaining articles (n = 199), once their full texts had been read, the inclusion criteria in Table 2 were applied. This step left the final list of studies on which the analysis was based (n = 75). The PRISMA flow diagram showing the study selection process is provided in Fig. 1.

Fig. 1.

Preferred Reporting Items for Systematic Review and Meta-Analysis (PRISMA) 2020 flow diagram and selection process.

Source: authors’ original work

Data Extraction and Quality Assessment

Table 3 shows the main data gathered from the 75 studies selected. Cost data referring to local currencies and from before 2022 were adjusted using the World Bank GDP deflator [22], and converted into 2022 Euros using the European Central Bank exchange rate [23]. The total cost presented in all tables as a synthesis of the different studies is defined as the sum of bed charges, cost of medicines, and the cost of staff and diagnostic tests. This is what is now understood as ‘total hospital cost’. The ‘attributable cost’ is defined as the difference in healthcare cost between resistant and susceptible infections.1 When costs are referred to from a societal perspective, productivity loss is included in addition to the above costs. If the study presents the results for both resistant and susceptible infections, adjusted data were prioritised over unadjusted data, and overall mean estimates over individual means. In studies that did not report mean and standard deviation (SD) but provided the median and interquartile range (IQR) (especially in cost and length of stay (LOS) variables), these data were converted into mean and SD using techniques suggested by Luo et al. [24].

Table 3.

Metadata and microdata extracted from the included studies

Citation details	Title
	Authors
	Publication year
	Journal or institution
General information	Study period
	Study objectives
	Study country
	Income category of the country
	Study centre (single centre, multi-centre or national or multi-national)
	Study settings (primary, secondary or tertiary hospitals)
	Type of study (case control, cohort, observational, randomised control trial…)
	Study perspective (patients/payers, health care/ healthcare system, societal)
	Study population
	Sample size
Antibiotic resistance	Bacteria
	Antibiotic group
	Site of infection
Outcomes	Economic outcome indicators (e.g., cost of treatment), currency with cost year, mean or median total costs and/or attributable costs
	Clinical outcome indicators (e.g., mortality, length of stay, readmissions)
Study methodology and statistical analysis of cost data	Regression analysis, survival analysis, matching, multistate models, economic model, stepwise, sensitivity analysis, significance tests. . . Confidence intervals Confounding factors Addressing time-dependent bias and what methods of modelling were used (e.g., multistate model, state transition model, decision tree model)
Stated limitations
Study quality

Source: authors’ original work

The quality and risk of bias was independently revised by the authors using the Newcastle-Ottawa Quality Assessment Scale (NOS) [25]. As all the studies included in the meta-regression were non-randomised studies, such as cohort and case-control, the NOS scale was the preferred tool [16, 26, 27]. In order to compare the different studies, the number of stars achieved by every study with the NOS scale was converted into a score between 0 and 1. The number of achieved stars was divided by the possible maximum, as reported in Naylor et al. [15].

Regarding the study’s methodology and statistical analysis of cost data, several aspects were taken into account. As methodologies differ in modelling and between studies, related information was extracted in detail. Thus, information related to the matching methods, whether they had been considered confounding factors or not in the analysis, and what methods were used to reduce confounding effects (e.g., propensity score matching or Charlson comorbidity index) were also collected.

Meta-Regression Analysis

In addition to the systematic review, a meta-analysis was performed in order to assess the effect of the variable of interest (ABR) on the main outcome variables, LOS and mortality. We decided not to conduct a meta-analysis for healthcare costs due to the significant variability among studies in estimating costs, such as differences in cost categories and in health systems among countries. Consequently, we carried out a narrative synthesis following the guidance outlined in the Cochrane Handbook [28].

An exploratory analysis was carried out to assess how many publications could finally be used in the meta-analysis. The measures of effect selected to compare the resistance and sensitive groups is the standardised mean difference (SMD) for length of stay—following the thresholds established by Cohen [29] and Sawilowsky [30]—and odds ratio (OR) for the mortality variable.

To perform the meta-analysis and obtain the combined effect measure, random intercept models were selected. In this randomised model, it is assumed that the estimated effects may vary among the different studies, so that the differences between these effects are due to both sampling error and real variability among populations. This variability is equal to σ² + τ², where τ² is the heterogeneity parameter and quantifies the variance in the true effect measure. For the combination of the combined effect measure, the Mantel-Haenszel method [31] was applied. The estimation of the parameter τ² was carried out using restricted maximum likelihood (REML) [32]. For the estimation of confidence and prediction intervals, the Kenward-Roger method was used [33].

To estimate the individual effect measures and their sample variances, as well as the 95% confidence interval (CI) for the individual effect measures, it is necessary to have the means and standard deviations (SDs) of each group. Given that not all publications provide the same information about LOS, the following transformations had to be carried out when mean and SD were not available: (i) mean + 95% CI2; (ii) median + Q1 + Q3 (or IQR).3 In those publications that provided more than one measure of effect, an aggregation of the measures was done, taking into account that these measures may be autocorrelated.

A heterogeneity analysis, a sensitivity analysis, and an assessment of publication bias completed the analytical work. Meta-regression with possible moderator variables was performed in the mixed model environment [35] in order to incorporate publications with more than one moderator category, with restricted maximum likelihood estimation (REML). Once the different models had been obtained, the combined effect measure, heterogeneity measures, and prediction interval were re-estimated. For the significant models, the effect measures associated with the moderator were also obtained. A sub-group analysis was conducted using several moderators. For certain variables that included more than one level (e.g., type of bacteria or type of antibiotic), the selected meta-regression model was a multilevel model, in order to take this characteristic of the data into account. All the analyses were carried out using the free software R [36], and the metafor [37], meta [38] and dmetar [39] packages.

Results

Characteristics of the Studies

The main characteristics of the 75 studies included in this systematic review are presented in Table 4. Regarding the countries in this research, 55 out of 75 studies (73%) were conducted in high-income economies (HIEs), and the remaining ones were conducted in upper-middle income (UMIEs) and lower-middle income economies (LMIEs) (23% and 4%, respectively). It should be noted that there were no studies from low-income economies (LIEs). The countries with larger scientific production were the USA (n = 30, 40%), followed by China (n = 11, 15%) and Spain (n = 5, 7%). This classification based on countries’ incomes was defined by the World Bank [40].

Table 4.

Characteristics of the 75 studies included by type of bacteria

Author/s and publication year	Country, study centre/setting	Type of study and perspective	Exposure group	Control group	Resistance markers and specified bacteria	Site and origin of infection	Methods	Outcome of interest	Quality score (out of 1)
1. Staphylococcus aureus
Anderson et al. (2009) [41]	USA. Multi-centre: 1 tertiary care centre and 6 community hospitals	Case-control Payer perspective	MRSA (n = 150)	MSSA (n = 128)	Methicillin-resistant S. aureus	SSI HAI	Survival analysis	Hospital charges, LOS, mortality	0.44
Campbell et al. (2015) [42]	USA. National level	Cohort Payer perspective	MRSA (n = 119)	MSSA (n = 206)	Methicillin-resistant S. aureus	SSI HAI	Regression	Hospital charges, LOS, mortality	0.88
Engemann et al. (2003) [43]	USA. Multi-centre: 1 tertiary care and 1 community hospital	Cohort Payer perspective	MRSA (n = 121)	MSSA (n = 165)	Methicillin-resistant S. aureus	SSI HAI	Regression	Hospital costs, LOS, mortality	0.88
Parvizi et al. (2010) [44]	USA. Multi-centre: 2 academic institutions	Cohort Payer perspective	MRSA (n = 231)	MSSA (n = 160)	Methicillin-resistant S. aureus	BJI HAI	Significance test	Hospital costs, LOS	0.33
Barrero et al. (2014) [45]	Colombia. Multi-centre: 9 tertiary care centres	Cohort Payer perspective	MRSA (n = 102)	MSSA (n = 102)	Methicillin-resistant S. aureus	BSI HAI	Survival analysis	Hospital costs, LOS, mortality	0.77
Ben-David et al. (2009) [46]	USA. Single centre: 1 tertiary care hospital	Cohort Hospital perspective	MRSA (n = 53)	MSSA (n = 53)	Methicillin-resistant S. aureus	BSI HAI	Stepwise calculation	Hospital costs, LOS, mortality	0.88
Inagaki et al. (2019) [47]	USA. National level	Population-based Payer perspective	MRSA (n = 44,653)	MSSA (n = 47,435)	Methicillin-resistant S. aureus	BSI CAI	Regression	Hospital costs, LOS, mortality	0.77
Kim et al. (2014) [48]	South Korea. National level	Case-control Societal perspective	MRSA (n = 368)	MSSA (n = 249)	Methicillin-resistant S. aureus	BSI HAI	Survival analysis	Direct medical costs, societal costs, LOS, mortality	0.77
Lodise and McKinnon (2005) [49]	USA. Single centre: 1 tertiary care hospital	Cohort Hospital perspective	MRSA (n = 170)	MSSA (n = 183)	Methicillin-resistant S. aureus	BSI CAI and HAI	Survival analysis	Hospital costs, LOS, mortality	1
McHugh and Riley (2004) [50]	USA. Single centre: 1 tertiary care hospital	Case-control Hospital perspective	MRSA (n = 20)	MSSA (n = 40)	Methicillin-resistant S. aureus	BSI CAI and HAI	Regression	Hospital costs, LOS, mortality	0.67
Park et al. (2011) [51]	South Korea. Single centre: 1 tertiary care hospital	Case-control Hospital perspective	MRSA (n = 145)	MSSA (n = 121)	Methicillin-resistant S. aureus	BSI CAI and HAI	Matching	Hospital costs, LOS, mortality	0.55
Rubio-Terrés et al. (2010) [52]	Spain. Multi-centre: 27 tertiary care hospitals	Cohort Hospital perspective	MRSA (n = 121)	MSSA (n = 245)	Methicillin-resistant S. aureus	BSI CAI and HAI	Significance test	Direct healthcare costs, LOS, mortality	1
Shoji et al. (2022) [53]	Japan. Multi-centre: 16 tertiary care centres	Cohort Hospital perspective	MRSA (n = 23)	MSSA (n = 50)	Methicillin-resistant S. aureus	BSI HAI	Significance test	Hospital costs, LOS, mortality	1
Thampi et al. (2015) [54]	Canada. Multi-centre: 4 tertiary care centres	Cohort Hospital perspective	MRSA (n = 58)	MSSA (n = 377)	Methicillin-resistant S. aureus	BSI CAI	Matching	Direct healthcare costs, LOS, mortality	0.88
Tsuzuki et al. (2021) [55]	Japan. Single centre: 1 tertiary care hospital	Cohort Payer perspective	MRSA (n = 63)	MSSA (n = 84)	Methicillin-resistant S. aureus	BSI CAI	Matching	Hospital costs, LOS, mortality	0.77
Ott et al. (2010) [56]	Germany. Single centre: 1 tertiary care hospital	Case-control Payer perspective	MRSA (n = 41)	MSSA (n = 41)	Methicillin-resistant S. aureus	RTI HAI	Significance test	Hospital costs, LOS, mortality	0.33
Shorr et al. (2010) [57]	USA. Single centre: 1 tertiary care hospital	Cohort Hospital perspective	MRSA (n = 87)	MSSA (n = 55)	Methicillin-resistant S. aureus	RTI HAI	Significance test	Hospital charges, LOS, mortality	0.88
Shorr et al. (2006) [58]	USA. Multi-centre: 59 hospitals (16 teaching and 43 nonteaching)	Cohort Hospital perspective	MRSA (n = 59)	MSSA (n = 95)	Methicillin-resistant S. aureus	RTI HAI	Regression	Hospital charges, LOS, mortality	0.77
Taneja et al. (2010) [59]	USA. Single centre: 1 tertiary care hospital	Cohort Payer perspective	MRSA (n = 55)	MSSA (n = 73)	Methicillin-resistant S. aureus	RTI CAI	Significance test	Hospital costs, LOS, mortality	0.88
Capitano et al. (2003) [60]	USA. Single centre: 1 long-term care facility (LTCF)	Cohort LTCF perspective	MRSA (n = 41)	MSSA (n = 49)	Methicillin-resistant S. aureus	UTI, SSTI, RTI, BSI HAI	Significance test	Hospital costs, mortality	0.88
Filice et al. (2010) [61]	USA. Community level	Cohort Payer perspective	MRSA (n = 335)	MSSA (n = 390)	Methicillin-resistant S. aureus	BJI, UTI, SSTI, RTI, BSI, SSI, IE HAI	Regression	Hospital costs, hospital utilization, LOS, mortality	0.88
Klein et al. (2019) [62]	USA. National level	Cohort Hospital perspective	MRSA (n = 358,140)	MSSA (n = 257,930)	Methicillin-resistant S. aureus	BSI, RTI and “Unspecified infections” CAI	Matching	Hospital costs, LOS, mortality	0.77
Kopp et al. (2004) [63]	USA. Single centre: 1 academic medical centre	Case-control Hospital perspective	MRSA (n = 36)	MSSA (n = 36)	Methicillin-resistant S. aureus	BSI, RTI, SSTI, UTI, SKI. CAI	Significance test	Hospital costs, LOS, mortality	0.67
Zhen et al. (2020) [64]	China. Multi-centre: 4 tertiary care hospitals	Cohort Payer perspective	MRSA (n = 1,335)	MSSA (n = 1,397)	Methicillin-resistant S. aureus	Several (n.s.) HAI	Matching	Hospital costs, LOS, mortality	0.88
2. Pseudomonas aeruginosa
Yang et al. (2021) [65]	China. Single centre: 1 tertiary care hospital	Cohort. Societal perspective	CRPA (n = 65)	CSPA (n = 155)	Carbapenem-resistant P. aeruginosa	BSI CAI and HAI	Modelling/economic evaluation	Direct costs, indirect costs, LOS, mortality	0.77
Chen et al. (2019) [66]	China. Single centre: 1 tertiary care hospital	Cohort Hospital perspective	CRPA (n = 270)	CSPA (n = 270)	Carbapenem-resistant P. aeruginosa	BSI, RTI, SSTI, UTI CAI	Matching	Hospital costs, LOS, mortality	1
Eagye et al. (2009) [67]	USA. Single centre: 1 tertiary care hospital	Case-control Hospital perspective	MRPA (n = 58)	MSPA (n = 125)	Meropenem-resistant P. aeruginosa	BSI, RTI, UTI HAI	Regression	Hospital costs, LOS, mortality	0.67
Judd et al. (2016) [68]	USA. Single centre: 1 tertiary care hospital	Cohort Hospital perspective	MRPA (n = 32)	MSPA (n = 350)	Meropenem-resistant P. aeruginosa	RTI, SSTI, BSI HAI	Regression	Hospital costs, LOS, mortality	0.55
Lautenbach et al. (2006) [69]	USA. Single centre: 1 tertiary care hospital	Ecological, cohort and case control Hospital perspective	IRPA (n = 142)	ISPA (n = 737)	Imipenem-resistant P. aeruginosa	BSI, UTI, RTI, IA, SSTI HAI	Regression	Hospital costs, LOS, mortality	0.77
Morales et al. (2012) [70]	Spain. Single centre: 1 tertiary care hospital	Cohort Hospital perspective	RPA (n = 119)	NRPA (n = 149)	Resistant P. aeruginosa to carbapenems, quinolones, tobramycin and gentamicin	RTI, SSTI, IA HAI	Regression	Hospital costs, LOS, mortality	1
3. Escherichia coli
Naylor et al. (2019) [71]	UK. National level	Cohort Hospital perspective	REC (n = 2,661)	NREC (n = 8,402)	Resistant E. coli to ciprofloxacin, third generation cephalosporins, gentamicin, piperacillin/tazobactam or carbapenems	BSI CAI	Multistate models	Hospital costs, LOS, mortality	0.77
Alam et al. (2009) [72]	UK. Multi-centre: 10 GP practices	Cohort Healthcare perspective	REC (n = 495)	NREC (n = 476)	Resistant E. coli to ampicillin or trimethoprim	UTI CAI	Regression	Direct costs	0.88
Iskandar et al. (2021) [73]	Lebanon. Multi-centre: 10 tertiary hospitals	Cohort Payer perspective	REC (n = 250)	NREC (n = 217)	Resistant E. coli to carbapenems, extended-spectrum cephalosporins and fluoroquinolones	UTI CAI and HAI	Regression	Hospital costs, LOS, mortality	1
Noskin et al. (2001) [74]	USA. Single centre: 1 tertiary care hospital	Cohort Payer perspective	REC (n = 42)	NREC (n = 47)	Resistant E. coli to ampicillin, cotrimoxazole, fluoroquinolones, levofloxacin or ciprofloxacin	UTI CAI	Significance test	Hospital costs	–
Meng et al. (2017) [75]	China. Single centre: 1 tertiary care hospital	Case-control Hospital perspective	CREC (n = 49)	CSEC (n = 98)	Carbapenem-resistant E. coli	RTI, UTI, SSI, BSI, IA HAI	Regression	Hospital costs, mortality	0.67
4. Enterobacter
Cosgrove et al. (2002) [76]	USA. Single centre: 1 tertiary care hospital	Cohort Hospital perspective	3GCREN (n = 46)	3GCSEN (n = 113)	Third-generation cephalosporin-resistant Enterobacter	UTI, SSTI, RTI, BSI, UTI HAI	Regression	Hospital costs, LOS, mortality	0.88
5. Enterococcus
Butler et al. (2010) [77]	USA. Single centre: 1 tertiary care hospital	Cohort Payer perspective	VRE (n = 94)	VSE (n = 182)	Vancomycin-resistant Enterococcus	BSI. CAI	Matching	Hospital costs, LOS	0.88
Cheah et al. (2013) [78]	Australia. Single centre: 1 tertiary care hospital	Cohort Hospital perspective	VRE (n = 116)	VRE (n = 116)	Vancomycin-resistant Enterococcus	BSI HAI	Regression	Hospital costs, LOS, mortality	1
Pelz et al. (2002) [79]	USA. Single centre: 1 tertiary care hospital	Cohort Hospital perspective	VRE (n = 12)	VSE (n = 22)	Vancomycin-resistant Enterococcus	BSI, UTI, SSTI n.s.	Regression	Hospital costs, LOS, mortality	-
Puchter et al. (2018) [80]	Germany. Single centre: 1 tertiary care hospital	Case-control Hospital perspective	VRE (n = 42)	VSE (n = 42)	Vancomycin-resistant Enterococcus	SSI, BSI, IA HAI	Regression	Hospital costs, LOS, mortality	0.55
Webb et al. (2001) [81]	USA. Single centre: 1 tertiary care hospital	Case-control Hospital perspective	VRE (n = 262)	VSE (n = 157)	Vancomycin-resistant Enterococcus	UTI, SSTI, BSI HAI	Regression	Hospital costs, LOS, mortality	0.55
6. Streptococcus pneumoniae
Reynolds et al. (2014) [82]	USA. National level	Survey data Hospital perspective	Resistant (n = 108,750)	Susceptible (n = 353,543)	Resistant S. pneumoniae to penicillin, erythromycin or fluoroquinolones	RTI CAI	Modelling	Hospital costs	-
Quach et al. (2002) [83]	Canada. Multi-centre: 2 tertiary care hospitals	Case-control Hospital perspective	PRSP (n = 36)	PSSP (n = 108)	Penicillin-resistant S. pneumoniae	BSI, RTI, IM, SSTI, BJI CAI	Significance test	Hospital costs, LOS, mortality	0.55
7. Acinetobacter
Lemos et al. (2014) [84]	Colombia. Multi-centre: 4 tertiary care hospitals	Cohort Hospital perspective	CRAB (n = 104)	CSAB (n = 61)	Carbapenem-resistant A. baumanii	RTI, BSI, SSI, UTI, SSTI, IA HAI	Regression	Hospital costs, LOS, mortality	0.88
Zhen et al. (2017) [85]	China. Single centre: 1 tertiary care hospital	Cohort Hospital perspective	CRAB (n = 2,126)	CSAB (n = 854)	Carbapenem-resistant A. baumanii	BSI, UTI, RTI HAI	Regression	Hospital costs, LOS	0.88
8. Klebsiella pneumoniae
Huang et al. (2018) [86]	China. Single centre: 1 tertiary care hospital	Cohort Hospital perspective	CRKP (n = 267)	CSKP (n = 1,328)	Carbapenem-resistant K. pneumoniae	RTI, BSI, UTI, IA HAI	Matching	Hospital costs, LOS, mortality	0.67
9. Salmonella enterica
Broughton et al. (2010) [87]	China. Single centre: 1 tertiary care hospital	Cohort. Hospital perspective	QRSE (n = 162)	QSSE (n = 163)	Quinolone-resistant Salmonella enterica	IA CAI	Regression	Hospital costs, LOS, mortality	0.88
10. Several bacteria
Kraker et al. (2011) [88]	31 countries. National level	Observational study (Panel survey) Hospital perspective	MRSA (n =  27,711) G3CREC (n = 15,183)	MSSA (n = 80,723) G3SEC (n = 148,292)	Methicillin-resistant S. aureus and third-generation cephalosporin-resistant E. coli	BSI	Modelling	Hospital costs, LOS, mortality	–
Chandy et al. (2014) [89]	India. Single centre: 1 tertiary care hospital	Cohort Payer perspective	Resistant (n = 133)	Susceptible (n = 87)	Resistant E. coli, K. pneumoniae, Enterobacter, S. aureus, S. pneumoniae and Enterococcus to Piperacillin-Tazobactam, cephalosporins, fluoroquinolones, aminoglycosides and carbapenems	BSI CAI	Significance test	Hospital costs, LOS, mortality	0.88
Riu et al. (2016) [90]	Spain. Single centre: 1 tertiary care hospital	Case-control Hospital perspective	Resistant (n = 167)	Susceptible (n = 404)	E. coli and K. pneumoniae resistant to carbapenems, aminoglycosides, and first-, second-, third-, and fourth-generation cephalosporins; P. aeruginosa resistant to carbapenems, aminoglycosides, monobactams, cephalosporins, quinolones; S. aureus resistant to penicillin and glycopeptides	BSI HAI	Matching	Hospital costs, LOS, mortality	0.55
Song et al. (2022) [91]	South Korea. Multi-centre: 10 secondary or tertiary hospitals	Case-control Societal perspective	Resistant (n = 486)	Susceptible (n = 486)	VRE or CRE resistant S. aureus, Enterococcus, A. baumannii, P. aeruginosa, and Enterobacteriaceae	BSI HAI	Modelling/economic evaluation	Hospital costs, LOS, mortality	0.55
Stewardson et al. (2016) [92]	Europe. Multi-centre: 9 tertiary and 1 secondary hospitals	Cohort Hospital perspective	Resistant (n = 523)	Susceptible (n = 2,985)	Resistant third-generation cephalosporin Enterobacteriaceae and MRSA	BSI CAI and HAI	Multistate model	Hospital costs, LOS, mortality	0.67
Zhu et al. (2021) [93]	China. Single centre: 1 tertiary care hospital	Cohort Hospital perspective	Resistant (n = 112)	Susceptible (n = 112)	Resistant E. coli and K. pneumoniae to carbapenems	BSI HAI	Matching	Hospital costs, LOS, mortality	0.88
Evans et al. (2007) [94]	USA. Single centre: 1 tertiary care hospital	Cohort Hospital perspective	Resistant (n = 137)	Susceptible (n = 467)	Resistant E. coli, Klebsiella spp., Enterobacter spp. and Pseudomonas spp. to aminoglycosides, cephalosporins, carbapenems and fluoroquinolones	SSI HAI	Regression	Hospital costs, LOS, mortality	0.88
Zilberberg et al. (2019) [95]	USA. National level	Cohort Hospital perspective	Resistant (n = 1,059)	Susceptible (n = 7,910)	Resistant Klebsiella spp., P. aeruginosa, A. baumannii and Enterobacteriaceae to carbapenems	RTI HAI	Regression	Hospital costs, LOS, mortality	1
Vargas-Alzate et al. (2019) [96]	Colombia. Single centre: 1 tertiary care hospital	Cohort Hospital perspective	Resistant (n = 86)	Susceptible (n = 55)	Resistant K. pneumoniae, Enterobacter cloacae and P. aeruginosa to beta-lactamases, carbapenems and cephalosporins.	UTI HAI	Regression	Hospital costs, LOS, mortality	0.88
Tabak et al. (2019) [97]	USA. Multi-centre: 78 acute care hospitals	Cohort Payer perspective	Resistant (n = 273)	Susceptible (n = 1,365)	Resistant P. aeruginosa, A. baumannii and Enterobacteriaceae to carbapenems	UTI CAI and HAI	Matching	Hospital costs, LOS, mortality	0.88
Cheong et al. (2022) [98]	South Korea. Single centre: 1 tertiary care hospital	Cohort Hospital perspective	Resistant (n = 87)	Susceptible (n = 154)	Resistant Enterobacteriales (E. coli, Klebsiella spp., Proteus spp., Enterobacter spp., or Citrobacter spp.) to ciprofloxacin	UTI CAI	Regression	Hospital costs, LOS, mortality	1
Iskandar et al. (2021) [99]	Lebanon. Multi-centre: 10 tertiary acute-care centres	Cohort Payer perspective	Resistant (n = 508)	Susceptible (n = 781)	Enterobacteriaceae resistant to carbapenems, fluoroquinolones, aminoglycosides, and third-generation cephalosporins; A. baumanii resistant to carbapenems; P. aeruginosa resistant to carbapenems; E. faecalis and E. faecium resistant to vancomycin; S. aureus resistant to oxacillin; S. pneumoniae resistant to penicillins; and Streptococcus B resistant to clindamycin	UTI, SSTI, RTI, BSI, IA CAI and HAI	Regression	Hospital costs, LOS, mortality	1
Lee et al. (2021) [100]	Australia. National level	Cohort Hospital perspective	Resistant (n = 1,874)	Susceptible (n = 18,516)	3GCRKP, 3GCREC, Ceftazidime-resistant P. aeruginosa, MRSA and VRE	BSI, UTI, RTI HAI	Multistate survival	Hospital costs, LOS, mortality	0.88
López-Montesinos et al. (2020) [101]	Spain. Single centre: 1 tertiary care hospital	Case-control Hospital perspective	Resistant (n = 97)	Susceptible (n = 97)	Resistant E. coli, K. pneumoniae, Proteus spp., P. aeruginosa, S. aureus, H. influenzae, Streptococci viridans, S. pneumoniae, E. faecalis, Morganella morganii, Providencia spp., Enterobacter cloacae and S. epidermidis	UTI, SSTI, RTI, BSI, IA CAI and HAI	Regression	Hospital costs, LOS, mortality	0.88
Neidell et al. (2012) [102]	USA. Multi-centre: 1 community hospital, 1 paediatric hospital and 2 tertiary care hospitals	Cohort Hospital perspective	Resistant (n = 1,170)	Susceptible (n = 2,228)	K. pneumoniae resistant to imipenem/meropenem, P. aeruginosa resistant to levofloxacin, S. aureus resistant to oxacillin, VRE, and A. baumanii resistant to ampicillin	UTI, RTI, BSI CAI and HAI	Matching	Hospital charges, LOS, mortality	0.77
Shrestha et al. (2018) [103]	USA and Thailand. National level	Observational. Societal perspective	Resistant (n = n.s.)	Susceptible (n = n.s.)	S. aureus resistant to oxacillin, E. coli and K. pneumoniae resistant to 3rd- generation cephalosporin, and P. aeruginosa and A. baumanii resistant to carbapenems	Not specified	Modelling/economic evaluation	Direct and indirect costs	–
Wozniak et al. (2019) [104]	Australia. National level	Population-based model. Hospital perspective	Resistant (n = n.s.)	Susceptible (n = n.s.)	Ceftriaxone-resistant E. coli and K. pneumoniae, ceftazidime-resistant P. aeruginosa, VRE and MRSA	BSI, UTI, RTI CAI and HAI	Modelling/economic evaluation	Direct costs, LOS, mortality	–
Wozniak et al. (2022) [105]	Australia. National level	Population-based model. Societal perspective	Resistant (n = n.s.)	Susceptible (n = n.s.)	Ceftriaxone-resistant E. coli and K. pneumoniae, ceftazidime-resistant P. aeruginosa, VRE and MRSA	BSI, UTI, RTI CAI and HAI	Modelling/economic evaluation	Direct medical costs, societal costs, LOS, mortality.	–
Touat et al. (2021) [106]	France. National level	Observational. Hospital perspective	Resistant (n = 11,680)	Susceptible (n = 61,564)	Resistant E. coli, K. pneumoniae and S. aureus	UTI, BJI, IA, SSTI, RTI, IE CAI	Regression	Hospital costs, LOS	–
Zhen et al. (2021) [18]	China. National level	Cohort Hospital perspective	Resistant (n = 5,459)	Susceptible (n = 3,163)	Resistant S. aureus, E. faecalis, E. faecium, E. coli, K. pneumonia, P. aeruginosa, and A. baumannii	Several HAI	Matching	Hospital costs, LOS, mortality	0.88
Imai et al. (2022) [107]	Japan. Multi-centre: 55 hospitals	Cohort Hospital perspective	Resistant (n = 86)	Susceptible (n = 9,431)	Resistant Acinetobacter spp., Citrobacter freundii, Enterobacter cloacae, E. coli, Klebsiella spp., Proteus spp., and P. aeruginosa to carbapenems	RTI, UTI, IA, BSI CAI	Regression	Hospital costs, LOS, mortality	0.88
Mauldin et al. (2010) [108]	USA. Single centre: 1 tertiary care hospital	Cohort Hospital perspective	Resistant (n = 193)	Susceptible (n = 276)	Resistant Acinetobacter spp., E. coli, Klebsiella spp., Enterobacter spp. and P. aeruginosa to fluoroquinolones, carbapenems or extended-spectrum cephalosporins	BSI, RTI, UTI, SSI HAI	Regression	Hospital costs, LOS	0.77
Merrick et al. (2021) [109]	UK. Single centre: 1 tertiary care hospital	Cohort Hospital perspective	Resistant (n = 221)	Susceptible (n = 221)	Carbapenem-resistant Pseudomonas spp., Enterobacter spp., Klebsiella spp., E. coli, Acinetobacter spp., Serratia spp., Hafnia spp. and Elizabethkingia spp.	RTI, UTI, SSTI, BSI, BJI, IA HAI	Significance test	Hospital costs, LOS, mortality	0.77
Vargas-Alzate et al. (2018) [110]	Colombia. Single centre: 1 tertiary care hospital	Cohort Hospital perspective	Resistant (n = 48)	Susceptible (n = 170)	Resistant K. pneumoniae, Enterobacter cloacae and P. aeruginosa to carbapenems	RTI, BSI, SSI, IA HAI	Regression	Hospital costs, LOS, mortality	0.77
Zhen et al. (2020) [111]	China. Multi-centre: 4 tertiary care hospitals	Cohort Payer perspective	Resistant (n = 3,740)	Susceptible (n = ,8282)	Resistant K. pneumoniae, P. aeruginosa and A. baumannii to carbapenems	n.s. n.s.	Matching	Hospital costs, LOS, mortality	0.88
Zhen et al. (2020) [112]	China. Multi-centre: 4 tertiary care hospitals	Cohort Hospital perspective	Resistant (n = 3,735)	Susceptible (n = 6,073)	Resistant E. coli to ampicillin or trimethoprim	BSI, UTI, IA n.s.	Matching	Hospital costs, LOS, mortality	0.88
Lashari et al. (2022) [113]	Indonesia. Single centre: 1 tertiary referral hospital	Case-control Payer perspective	Resistant (n = 270)	Susceptible (n = 540)	Resistant E. coli, P. aeruginosa, A. baumannii and K. pneumoniae to carbapenems	UTI, BSI, IA, IE, BJI HAI	Significance test	Hospital costs, LOS	0.55
Kengkla et al. (2022) [114]	Thailand. Single centre: 1 tertiary referral hospital	Cohort Hospital perspective	Resistant (n = 627)	Susceptible (n = 3,882)	Carbapenem-resistant K. pneumoniae, E. coli, Klebsiella oxytoca, Proteus mirabilis and Proteus spp.	UTI, RTI, BSI, IA n.s.	Regression	Hospital costs, LOS, mortality	–

Source: authors’ original work

SSI surgical site infection, BSI bacteraemia, UTI urinary tract infection, SSTI skin and soft tissue infection, RTI respiratory tract infection, BJI bones and joints infection, IE Eedocarditis, IA intra-abdominal infection, LOS length of stay, SKI skeletal infection, IM meningitis, G3CREC third-generation cephalosporin-resistant E. coli, G3SEC third-generation cephalosporin-susceptible E. coli, 3GCREN third-generation cephalosporin-resistant Enterobacter, 3GCSEN third-generation cephalosporin-susceptible Enterobacter, CRKP carbapenem-resistant K. pneumoniae, CSKP carbapenem-susceptible K. pneumoniae, CRPA carbapenem-resistant P. aeruginosa, CSPA carbapenem-susceptible P. aeruginosa, MRPA meropenem-resistant P. aeruginosa, MSPA meropenem-susceptible P. aeruginosa, IRPA imipenem-resistant P. aeruginosa, ISPA imipenem-susceptible P. aeruginosa, RPA resistant P. aeruginosa, NRPA non-resistant P. aeruginosa, VRE vancomycin-resistant Enterococcus, VSE vancomycin-susceptible Enterococcus, REC resistant E. coli, NREC non-resistant E. coli, CREC carbapenem-resistant E. coli, CSEC carbapenem-susceptible E. coli, PRSP penicillin-resistant S. pneumoniae, PSSP penicillin-susceptible S. pneumoniae, CRAB carbapenem-resistant A. baumanii, CSAB carbapenem-susceptible A. baumanii, CAI community-associated infection, QRSE quinolone-resistant Salmonella enterica, QSSE quinolone-susceptible Salmonella enterica, HAI healthcare-associated infection, n.s. not specified

The studies without a quality score are those that were not included in the meta-analysis, either because they only reported cost outcomes or, in cases where they did report data on LOS and/or mortality, they lacked the necessary statistical details (e.g., SDs, IQRs) required for the analysis

In terms of the studies’ settings, the majority were conducted in a single centre (n = 39, 52%). Out of 75 studies, 71% (n = 53) were conducted in tertiary-care settings.4 The most frequent type of study was cohort (n = 53, 70%) and the most reported perspective was the hospital perspective (n = 50, 67%). In terms of the bacteria analysed, the species studied most were Staphylococcus aureus (19.3%), Klebsiella spp. (14.2%) and Pseudomonas spp. and Escherichia coli (13.7% each one) (see Figs. 2 and 3 for more details). The most commonly reported site of infection was the bloodstream (31.4%). Out of the total of studies included, 51% (n = 38) focused on hospital-acquired or nosocomial infections and 24% (n = 18) on community-acquired infections, and 19% (n = 14) examined both. The remaining studies did not clearly distinguish the origin of infection. Finally, regarding the methodology used in the studies included, regression was the most employed technique (n = 32, 43%) in order to estimate LOS and mortality outcomes due to ABR.

Fig. 2.

Type of bacteria analysed in the studies reviewed.

Source: authors’ original work. The category ‘several bacteria’ refers to studies that include more than one bacterium of interest

Fig. 3.

Resistant bacteria analysed in the reviewed studies (%).

Source: authors’ original work

Reported Bacteria, Sites of Infection and Antibiotic Groups

Of the 75 selected studies, 28 focused on the joint evaluation of several bacteria (see Fig. 2). The second most frequent type of research, with 24 published studies, considered only one bacterium, Staphylococcus aureus. Studies specifically on Pseudomonas aeruginosa and Escherichia coli represented six and five, respectively.

Of the total number of body sites analysed, the bloodstream was the most commonly studied area (31%). The second most analysed body site was the respiratory tract (19%), followed by the urinary tract (19%). The least analysed infection sites were skin and soft tissue (9%), intra-abdominal (9%), surgical site infection (6%), bones and joints (4%), infective endocarditis (2%) and meningitis and the skeleton (both 0.5%). Regarding the antibiotic groups, methicillin resistance was the resistance marker under greatest scrutiny, representing 22% of the total resistance markers, followed by carbapenems (21%) and cephalosporins (12%). The least analysed antibiotics were quinolones, vancomycin, penicillin (9% each), aminoglycosides (4%), clindamycin, monobactams, glycopeptides, beta-lactamases, trimethoprim, cotrimoxazole and macrolides (1% each).

Quality of the Studies Included

On the basis of standardised scoring-across tools (explained in Sect. 2.3), the maximum quality score that a study can achieve is 1, and the minimum 0, where a ≥ 0.76 score means low risk of bias, a score between 0.51 and 0.75 means medium risk, and less than 0.50 means high risk [16]. The studies included in this review ranged from 0.33 to 1 in the NOS scale, with a median quality score of 0.88 and a IQR of 0.19. Out of all the studies analysed, 49 studies had a low risk of bias (reporting quality scores ≥ 0.76), 13 had a medium risk (scores between 0.51 and 0.75), and three had high risk of bias. If we analyse the quality scores obtained by type of study, we find that cohort studies had a median quality of 0.88 (vs. 0.55 of case-control) and an IQR of 0.11. A noteworthy fact is that ABR researchers have taken uncertainties and risk of bias seriously while estimating the economic burden of ABR, as the risk of bias scores have decreased in recent years, especially since 2016 (see Table 4 for detailed scores).

Economic Burden Attributable to Antibiotic Resistance

From 75 studies reviewed, the attributable cost (adjusted mean, deflated and applied exchange rate for 2022) of resistant infections ranged from EUR -21,6295 (p = 0.510, higher for RTI MSSA infections compared to MRSA infections) to EUR 74,452 (p = 0.001, higher for resistant infections in multiple sites caused by Pseudomonas spp.) per patient episode. A total of 70 out of 75 studies (93%) reported higher attributable costs for resistant infections compared to susceptible infections, although there were 54 studies that clearly reported significantly higher costs for resistant strains (72%). However, five studies reported higher costs for susceptible infections compared to resistant infections. Nevertheless, none of the five studies reported significantly higher healthcare costs for susceptible strains. The higher costs for susceptible infections compared to resistant ones could be attributed to several factors. For instance, some results were based on diagnostic billing codes [62], which may introduce bias in the reporting and billing of MRSA infections. Additionally, there may have been challenges in accurately estimating true costs using data originally collected for other purposes [71, 85, 97].

Regarding the study perspective, 70% of studies reported economic costs from the hospital perspective, where the attributable costs for resistant infections (compared to susceptible ones) range from EUR − 18,629 to EUR 74,452 (costs significantly higher for resistant strains in 63% of the studies) per patient episode. A total of 24% of studies reported costs from the payer’s perspective, ranging from EUR − 21,629 to EUR 55,639 (costs significantly higher for resistant strains in 88%) per episode. Finally, from the societal perspective (6%), economic costs ranged from EUR 1,111 to EUR 13,227 per case (costs significantly higher for resistant strains in 20%).

Based on healthcare settings, at tertiary care settings (n = 53), the attributable costs ranged from EUR − 21,629 to EUR 74,452 per episode, being the attributable costs in 43 studies significantly higher for resistant infections (95% CI, p < 0.10). Regarding the mixed settings (secondary, tertiary and/or paediatric) (n = 3), attributable costs range from EUR − 18,916 to EUR 37,123 per patient.

Concerning the study design, the attributable costs of resistant infections ranged from EUR − 21,629 (p = 0.510) to EUR 73,197 (p = 0.0001) per patient, in cohort studies; from EUR 371 (p < 0.001) to EUR 74,452 (p = 0.001) in case-control studies and from EUR 236 (p-value not reported) to EUR 1096 per case (p-value not reported) in observational studies based on population.

The economic cost and variation range attributable to ABR in the reviewed studies are presented by bacteria of interest in Table 5. The bacterium with a higher variation range was Pseudomonas spp. (EUR − 919, p = not reported, to EUR 74,452, p = 0.001, per patient). This same bacterium was also the costliest of all bacteria analysed. Specifically, MRPA in multiple sites was the costliest type of infection. The bacterium that reported always higher costs for the resistant group was Klebsiella spp., Acinetobacter spp., Enterobacter spp. and Salmonella enterica. Of the studies that reported significantly higher costs for resistant strains, Salmonella enterica (100% of studies) and Pseudomonas spp. (83% of studies) reached the highest concentration. National level cost and segregated cost for resistant and susceptible infections (by bacterium) are also presented in the Tables S1−S10 of the Online Supplementary Material.

Table 5.

Economic costs and variation range attributable to antibiotic resistance (ABR) by type of bacteria

Bacteria	Number of studies	Variation range of attributable cost (EUR2022)				% of studies reporting higher costs for resistant infections	% of studies reporting significantly higher costs for resistant infections
		Minimum (p-value)	Type of infection	Maximum (p-value)	Type of infection
S. aureus	27	− 21,629 (p = 0.510)	MRSA RTI	55,639 (p < 0.001)	MRSA SSI	92% (n = 25)	48% (n = 13)
Pseudomonas spp.	12	− 919 (p not reported)	Ceftazidime-resistant P. aeruginosa, RTI	74,452 (p = 0.001)	MRPA, Multiple sites	91% (n = 11)	83% (n = 10)
E. coli	8	− 1,818 (p not reported)	CREC, BSI	2,442 (p = 0.05)	CREC, Multiple sites	75% (n = 6)	50% (n = 4)
Enterococcus spp.	8	− 4,664 (p not reported)	VRE, BSI	58,444 (p = 0.002)	VRE, BSI	87% (n = 7)	63% (n = 5)
Klebsiella spp.	5	4,037 (p not reported)	3GCRKP, UTI	15,606 (p > 0.01)	Imipenem/meropenem-resistant K. pneumoniae, Multiple sites	100% (n = 5)	40% (n = 2)
Acinetobacter spp.	4	6,339 (p < 0.001)	CRAB, multiple sites	11,331 (p < 0.000)	CRAB, Multiple sites	100% (n = 4)	75% (n = 3)
Enterobacter spp.	2	3,894 (p not reported)	3GCREN, BSI	58,734 (p < 0.001)	3GCREN, Multiple sites	100% (n = 2)	50% (n = 1)
S. pneumoniae	1	722 (p = 0.18)	PRSP, multiple sites	–	–	0% (n = 0)	0% (n = 0)
Salmonella enterica	1	509 (p < 0.05)	QRSE, IA	–	–	100% (n = 1)	100% (n = 1)
Several bacteria	19	4,375 (p not reported)	BSI resistant infection	73,197 (p = 0.0001)	SSI resistant infection	95% (n = 18)	79% (n = 15)

Source: authors’ original work

Estimating Costs at National Level

A total of six studies reported costs of ABR at a national scale (see Table 6). For several cost categories, the extent of the ABR problem can be observed on a national scale. Considering a societal perspective (by adding healthcare costs and productivity losses due to premature deaths), the attributable costs of ABR ranged from EUR 4,323,397 (due to resistant UTI caused by Klebsiella spp. in Australia) to EUR 2.96 billion per year (due to resistant infections caused by several bacteria in multiple sites in the USA). This large variability may be due to the different complexities of infections considered in the studies and the country size, as well as differences in the costs of the countries’ healthcare systems.6 If we analyse the attributable costs of premature deaths exclusively, it can be observed that in most studies these costs derived from productivity losses greatly exceed the direct healthcare costs. In relative terms, the costs of premature deaths can account for up to 90% of the total cost, depending on the severity and type of infection.

Table 6.

Extrapolated costs attributable to antibiotic resistance (ABR) at a national scale

Author/s	Year of recruitment	Reported cost	Original currency (country)	Attributable excess costs a (EUR2022)	Attributable excess costs (adjusted per 100,000 population, International $ PPP, 2022)b
Staphylococcus aureus
BSI
Kraker et al. (2011) [88]	2007	Total cost	EUR2007 (31 countries)	55,294,580	–
Kim et al. (2014) [48]	2011	Total cost	USD2011 (South Korea)	55,252,234	172,870.67
		Healthcare costs		33,407,043	104,522.43
		Premature deaths cost		20,140,496	63,014.66
Wozniak et al. (2019) [104]	2014	Healthcare costs	AUD2014 (Australia)	4,267,667	28,399.61
Wozniak et al. (2022) [105]	2020	Total cost	AUD2020 (Australia)	84,647,941	515,642.91
		Healthcare costs		16,929,588	103,128.58
		Premature deaths cost		67,718,353	412,514.33
RTI
Wozniak et al. (2019) [104]	2014	Healthcare costs	AUD2014 (Australia)	1,163,909	7,745.34
UTI
Wozniak et al. (2022) [105]	2020	Total cost	AUD2020 (Australia)	17,282,287	105,277.08
		Healthcare costs		1,763,498	10,742.55
		Premature deaths cost		15,518,789	94,534.53
Pseudomonas spp.
BSI
Wozniak et al. (2022) [105]	2020	Total cost	AUD2020 (Australia)	18,552,006	113,011.73
		Healthcare costs		1,763,498	10,742.56
		Premature deaths costs		16,788,508	102,269.17
UTI
Wozniak et al. (2022) [105]	2020	Total cost	AUD2020 (Australia)	30,967,038	188,639.36
		Healthcare costs		10,722,072	65,314.76
		Premature deaths costs		20,244,966	123,324.59
RTI
Wozniak et al. (2019) [104]	2014	Healthcare costs	AUD2014 (Australia)	931,127	6,196.27
Wozniak et al. (2022) [105]	2020	Total cost	AUD2020 (Australia)	134,590,226	819,872.23
		Healthcare costs		6,912,915	42,110.83
		Premature deaths costs		127,677,311	777,761.39
Escherichia coli
BSI
Kraker et al. (2011) [88]	2007	Excess cost	EUR2007 (31 countries)	22,585,000	–
Wozniak et al. (2019) [104]	2014	Healthcare costs	AUD2014 (Australia)	4,500,448	29,948.67
Wozniak et al. (2022) [105]	2020	Total cost	AUD2020 (Australia)	20,527,124	125,043.39
		Healthcare costs		1,622,418	9,883.15
		Premature deaths costs		18,904,706	115,160.24
Wozniak et al. (2022) [105]	2020	Total cost	AUD2020 (Australia)	6,983,455	42,540.53
		Healthcare costs		6,983,455	42,540.53
		Premature deaths costs		0	0
Enterococcus spp.
BSI
Wozniak et al. (2019) [104]	2014	Healthcare costs	AUD2014 (Australia)	1,086,315	7,228.99
Wozniak et al. (2022) [105]	2020	Total cost	AUD2020 (Australia)	25,747,082	156,841.38
		Healthcare costs		1,410,799	8,594.04
		Premature deaths costs		24,336,283	148,247.33
UTI
Wozniak et al. (2022) [105]	2020	Total cost	AUD2020 (Australia)	18,199,307	110,863.22
		Healthcare costs		7,124,535	43,399.94
		Premature deaths costs		11,074,772	67,463.27
Streptococcus pneumoniae
RTI
Reynolds et al. (2014) [82]	2004	Total cost	USD2012 (USA)	257,049,806	120,256.48
		Direct medical costs		100,298,467	46,922.97
Klebsiella spp.
BSI
Wozniak et al. (2019) [104]	2014	Healthcare costs	AUD2014 (Australia)	1,008,721	6,712.63
Wozniak et al. (2022) [105]	2020	Total cost	AUD2020 (Australia)	8,041,553	48,986.06
		Healthcare costs		1,551,878	9,453.44
		Premature deaths costs		6,489,675	39,532.62
UTI
Wozniak et al. (2022) [105]	2020	Total cost	AUD2020 (Australia)	4,323,397	26,336.48
		Healthcare costs		4,323,397	26,336.48
		Premature deaths costs		0	0
Several bacteria
Multiple sites
Wozniak et al. (2022) [105]	2020	Total cost	AUD2020 (Australia)	359,753,750	2,191,482.39
		Healthcare costs		50,788,764	309,385.74
		Premature deaths costs		308,964,986	1,882,096.65
Shrestha et al. (2018) [103]	2016	Total costs	USD2016 (USA and Thailand)	529,389,624 (Thailand) and 2.96 billion (USA)	1,556,442.35 and 1,255,082.34

Source: authors’ original work. (^a) The excess cost refers to the higher differential cost of resistant versus susceptible infections. (^b) To facilitate comparisons across countries with different population sizes and cost structures, we have reported the attributable excess costs per 100,000 population, adjusted to international dollars using 2022 PPP conversion factors [116]. This standardization allows for more consistent cross-country comparison by accounting for both differences in purchasing power and national population size

A cross-country comparison in Table 6 shows that the highest attributable cost from a societal perspective is reported in Australia (International $ 2,191,482.39) for resistant infections across multiple sites caused by several bacteria. The difference with the USA and Thailand may reflect real differences in incidence, health expenditure, or the methodologies used in the respective studies. Nevertheless, all three values exceed International $ 1 million per 100,000 population, underscoring the substantial economic impact of antibiotic resistance when considered in aggregate across multiple infection sites.

Meta-analysis of the Impact of Antibiotic Resistance on Length of Stay and Mortality

Effects on Length of Stay

From the 75 studies reviewed, 47 provided data about the length of stay and suitable information (median, IQR, mean and/or SD), making it possible to estimate the mean LOS for resistant and susceptible infections. Thus, the data were included for meta-analysis.

The SMD is significantly different from 0 (0.91; 95% CI 0.67; 1.15). Thus, it can be inferred that the number of hospitalisation days is significantly higher in patients with ABR (8.72 extra days 95% CI 6.42; 11.02). Figure 4 presents the forest plot showing the SMD.

Fig. 4.

Forest plot with the estimate of the combined standardised mean difference (SMD), length of stay (LOS).

Source: authors’ original work. R is for the resistant group and S is for the susceptible group

Several variables (type of bacterium, type of Gram, site of infection, origin of infection and antibiotic groups) were incorporated into different meta-regression models to evaluate whether any of them were related to the differences in the number of hospitalisation days between the two groups (resistant and susceptible). The results of the moderator tests (H0: moderator does not change prevalence) are presented in Table S11 in the Online Supplementary Material. In this sense, significant differences were found in the following moderators: bacteria, site of infection, gram type and antibiotic agent.

Inside this sub-group analysis, we found that infections caused by Klebsiella pneumoniae (relative to other bacteria), infections manifested in several sites (with respect single body sites), Gram-negative infections (with respect Gram-positive) and resistance to carbapenems (with respect other antibiotics) had higher individual SMDs and, therefore, longer hospital stays. Specifically, Gram-negative resistant infections caused around one extra day of hospitalisation compared to Gram-positive resistant infections (mean difference: 8.57 and 7.97 days, respectively). With regard to age and sex, given that there is separate information for the resistant and susceptible groups, a preliminary analysis was done to see if there were significant differences regarding these variables. As this was not the case, the mean of both groups was incorporated in the meta-regression model (Fig. S1 in the Online Supplementary Material).

Table S12 in the Online Supplementary Material presents the meta-regression results. For the four variables that are incorporated in multilevel models, the aggregate forest plot is presented as a function of the levels for each of the variables (see Figs. S2−S5 in the Online Supplementary Material).

Mortality

From the 75 reviewed studies, 52 provided data about mortality and suitable information that made it possible to estimate the odds ratio (OR) for resistant and susceptible infections.

Regarding the results of measure of effect, the OR is notably different from 1 (1.65 (95% CI [1.44; 1.88])), so it can be inferred that the odds of death are significantly higher in the resistance group (risk increased by 65%), compared to the susceptible group. Figure 5 shows the forest plot with the measures of effect.

Fig. 5.

Forest plot with the estimate of the combined odds ratio (OR), mortality.

Source: authors’ original work

Analysing the results of the meta-regression, significant differences were found only in the variable related to origin of infection (CAI or HAI). With this subgroup analysis, we can see that hospital-acquired infections (HAIs) caused higher mortality than if the infections came from the community (CAIs) (OR 1.77 vs. OR 1.34, respectively). For these variables, the aggregate forest plot is presented as a function of the levels for each of the variables (see Fig. S6 in the Online Supplementary Material). Significant differences were not found for age and gender (Fig. S7 in the Online Supplementary Material), and the mean of both groups had been incorporated in order to include the resulting variable in a meta-regression model.

Heterogeneity, Outliers, Publication Bias and Sensitivity Analysis

Once the outliers had been identified7 in both the LOS and mortality analyses, they were excluded from the model. The aim was to observe whether or not the heterogeneity parameters presented variations. We found that, both in LOS and mortality, once the outliers were excluded, the heterogeneity was significantly reduced (see Figs. S8 and S9 in the Online Supplementary Material). The reduction in heterogeneity after excluding outliers implies greater consistency among the remaining studies, which in turn increases the reliability and interpretability of the pooled estimates in both the LOS and mortality analyses [117]. In addition, several methods were used to assess publication bias; the conclusion was reached that the presence of publication bias could not be detected. The two Peters tests indicated the absence of asymmetry in the funnel plot, a fact related to the absence of publication bias (see Fig. S10 in the Online Supplementary Material). In the radial plot of Fig. S11 in the Online Supplementary Material (Galbraith plot), all values fell within the confidence interval demonstrating no publication bias. Finally, the Rosenberg method also seemed to rule out the presence of publication bias, given the large number of non-significant studies that would be necessary to modify the conclusions of the meta-analysis—LOS: Observed Significance Level of 0, Target Significance Level of 0.05 and Fail safe N: 547566; Mortality: Observed Significance Level of 1.53e−56, Target Significance Level of 0.05 and Fail safe N: 3347.

Discussion

This systematic review and meta-analysis offer a thorough and current evaluation of how resistant infections affect hospital stay, mortality and economic costs. Compared to previous reviews, which often focused on individual countries, selected bacterial pathogens, or limited types of infections, our study provides a broader, comprehensive and more robust analysis. By including 75 studies—almost double or more than previous reviews such as Poudel et al. [16] and Founou et al. [13]—and covering a global perspective across all bacteria and antibiotic groups, our review offers a complete picture of the economic burden of antibiotic resistance. A total of 75 studies were identified as eligible for inclusion in this analysis, and results showed that the majority of research was conducted in tertiary care settings. Only a very limited number of studies were conducted in primary- and secondary-care settings (28%). The most studied bacteria were Staphylococcus aureus (19%), Klebsiella spp. (14%) and Pseudomonas spp. and Escherichia coli (13% each). Studies on the economic burden most commonly adopted a hospital or healthcare provider perspective, instead of a payer or societal one. There was significant variation in the methods used to estimate the burden and outcomes of ABR, particularly regarding healthcare costs. Regression was the most widely used method to estimate LOS and mortality outcomes attributable to ABR.

The costs associated with resistant infections were notably higher than those for susceptible infections (93% of included studies). Overall, 72% of the studies that reported costs show significantly higher attributable costs for resistant infections. This review clearly demonstrates that the healthcare costs associated with resistant infections are substantially greater than those from susceptible infections. Furthermore, these conclusions about the studies’ characteristics are in line with previous meta-analyses published in the literature [15, 16, 18]. In contrast to previous literature reviews, our article presents the results of studies that estimate the economic burden of ABR from a societal perspective, including productivity losses.

If we analyse the costs attributable to ABR, prior studies stated that the attributable cost for ABR ranged from EUR − 1869 to EUR 22,894 per patient episode [16].8 As our results incorporate a greater range of studies, a higher variability of costs was found. Our results ranged from EUR − 21,629 (p = 0.510, higher for RTI MSSA infections compared to MRSA infections) to EUR 74,452 per patient episode. In HIE settings, there are higher variations in attributable costs from resistant infections. There are also higher attributable costs (this is a result also found by other authors [15, 16, 19]). In contrast, in UMIE settings such as China, Thailand or Lebanon, the attributable costs were found to be lower, and the cost range narrower, compared to HIEs. These lower costs in UMIEs could be attributed to less advanced treatment facilities and lower resource costs per unit, but may also be influenced by exchange rate differences [16].

This wide range of differences in the economic and health outcomes across the studies may be due to several factors, including the data sources used, the design and methodology of each study, whether adjustments were made for confounding factors or time-dependent variables, the study setting, or the intrinsic characteristics of the health system in a country. One must also bear in mind that the differences may be explained by the varying complexity of infections that are evaluated in the articles. Based on the quality assessment of the studies included in this review, the quality of recent research on economic burden was found to be relatively higher than that of previously published studies. In this sense, the overall median quality in our study was 0.88 (IQR = 0.19) compared to 0.79 (IQR = 0.12) from the study of Poudel et al. [16]. This fact supports the idea that the quality of the studies has improved over the last few years. In this sense, ABR researchers have taken uncertainties and the risk of bias seriously while estimating the economic burden of ABR. The risk of bias scores has decreased in recent years, especially since 2016.

Similar to healthcare costs, mean excess LOS and mortality are found to be higher in resistant infections than in susceptible ones. With respect to the length of hospital stay, our meta-analysis revealed that, on average, infections caused by resistant bacteria result in 8.72 extra days of hospitalisation (SMD 0.91 (95% CI [0.67; 1.15])). These results based on ABR causing longer hospitalisations are also consistent with previous literature. Matsumoto et al. [14] revealed a similar result for 20,662 patients with resistant infections in Japan (SMD 1.27 (95% CI [1.23; 1.31])); and Poudel et al. [16] found that resistant strains cause also higher days of hospitalisation (SMD 0.387 (95% CI [0.198;0.576]). Then, our estimates are at the midpoint of what has been published so far in the literature. Regarding mortality, higher rates were observed in patients with resistant infections than those with susceptible strains isolates. The results of the meta-analysis showed that the odds of death were significantly higher in the resistant group (OR 1.65 (95% CI [1.44; 1.88])), with an increased risk of 65% of dying. These findings are also consistent with previous studies where authors found similar figures, from OR 1.84 [16] to OR 2.25 [14]. The pooled mortality odds ratio from previously published meta-analyses conducted in developing countries was higher than the results of our review. Founou et al. [13] found an OR of 2.83 in their study on developing countries. This difference could be due to the fact that healthcare facilities in developing countries are less advanced compared to their developed counterparts and this disparity leads to higher mortality rates [13].

Limitations

This study is not without limitations. We obtained a very limited sample of ABR studies based on lower-middle income economies (4% of the total). Therefore, the findings of the review may not be representative of all types of economies and health systems. It is also worth noting that an intrinsic limitation of the literature can also be present our study; not all possible moderators that impact on the outcome variables of the study may be considered. A further limitation is the inability to include costs in the meta-analysis, due to significant variability among studies in estimating costs, as well as differences in cost categories and healthcare systems among countries [16, 115].

From this systematic review of the literature, some strengths and weaknesses of the published studies should be highlighted for future research. First, it would be beneficial to clearly report how the control groups were identified and selected, as well as how cases and controls were matched. Further information is also needed on how confounding factors were considered. Secondly, regarding cohort studies, it would be extremely helpful to obtain clear patient follow-up reports (e.g., duration of follow-up, as well as percentage of lost observations and their main characteristics). Finally, as sensitivity analyses are required for modelling studies, alternative versions of models are needed to reduce methodological uncertainty and demonstrate robustness. Heterogeneity must also be dealt with; this work must be documented in a report (i.e., running the models separately for different sub-groups). Finally, future studies should integrate the societal perspective in order to estimate further costs, which have remained underexplored (and may exceed all other costs) [118–120]. Of particular relevance is the estimation of the intangible costs associated with a lower quality of life due to the physical and emotional suffering experienced by those affected by ABR. This review of the literature highlights a noteworthy gap in the inclusion of intangible costs.

Conclusions

This study aims to present an updated and systematic review on the economic burden of antibiotic resistance (ABR). This review advances the existing literature by addressing key limitations of earlier studies, offering a global and complete synthesis of the economic burden of antibiotic resistance. The substantially larger number of included studies and the broader scope enhance the reliability and generalizability of our findings.

An extensive body of published literature based on the economics of ABR was reviewed. Moreover, we carried out a meta-analysis in order to determine the true effect size of resistant infections on two relevant clinical outcomes: mortality and length of stay (LOS). The quality of the selected studies was assessed, and recommendations for improving the quality of future research on the economic burden of antibiotic resistance were provided. Despite variations in the findings, recent studies indicate that the health and economic impact of ABR is substantial and affects every country across the globe. Based on the evidence from the studies included, the costs attributable to resistant infections were generally high compared to those for susceptible infections, with a few exceptions. Higher attributable costs and variations in the costs were found in HIE. Similar to attributable costs, excess LOS and mortality were higher in those patients with resistant strains.

Now that an analysis has been made of published studies, and with the aim of minimising methodological limitations, we recommend that future researchers design a prospective study. These researchers should account for more potential confounding factors, and conduct the research across multiple centres, incorporating diverse perspectives and considering the inclusion of intangible costs. It is important to ensure an adequate sample size, include various healthcare settings and, where feasible, apply a multistate modelling approach to assess long-term impacts and mitigate bias caused by time-dependent variables.

Our findings make it possible to collect reliable evidence that could be valuable to researchers, policymakers, clinicians and professionals involved in ABR control and the promotion of health. Beyond synthesizing the clinical and economic consequences of antibiotic resistance, this study offers valuable evidence to support modelling analyses and policy planning. Our meta-analyses provide relative effect estimates across multiple infection sites, origin of infection, antibiotic groups and bacterial pathogens, which are essential inputs for economic evaluations and simulation models assessing the cost-effectiveness of interventions aimed at reducing antibiotic resistance. This addresses a frequently cited gap in the literature, as highlighted by Painter et al. [121], where a lack of robust, comparative parameters has limited the development of reliable models. By offering stratified results by infection source and resistance profile, our study contributes empirically grounded data that can help improve the accuracy of future health economic models and inform the prioritization of antibiotic stewardship strategies.

Supplementary Information

Below is the link to the electronic supplementary material.

Declarations

Funding

Open Access funding provided thanks to the CRUE-CSIC agreement with Springer Nature. This research has been developed in the framework of grant PID2021-127898OB-I00 (WAntRed), funded by MICIU/AEI/10.13039/501100011033 and “ERDF: A way of making Europe”. This research was also supported by an FPU grant from the Spanish Ministry of Universities.

Conflicts of Interest

Not applicable.

Data Availability

The datasets generated and/or analysed during the current study are not publicly available due to all citation data was managed in an online software, but are available from the corresponding author on reasonable request.

Ethics Approval

Not applicable.

Consent to Participate

Not applicable.

Consent for Publication (from Patients/Participants)

Not applicable.

Code Availability

Not applicable.

Authors’ Contributions

SS and BC conceptualised the design of the study. SS and IL-M acquired the funding. SS and BC were responsible for the conception of the research question and search strategy. SS was responsible of collecting, storing and analysing data, screening of studies and writing the manuscript. BC contributed to the screening of studies. BC and IL-M provided overall supervision. All authors read and approved the final manuscript.