(C) PLOS One
This story was originally published by PLOS One and is unaltered.
. . . . . . . . . .



The impact of inpatient bloodstream infections caused by antibiotic-resistant bacteria in low- and middle-income countries: A systematic review and meta-analysis [1]

['Kasim Allel', 'Department Of Disease Control', 'Faculty Of Infectious', 'Tropical Diseases', 'London School Of Hygiene', 'Tropical Medicine', 'London', 'United Kingdom', 'Antimicrobial Resistance Centre', 'Institute For Global Health']

Date: 2023-06

We provide an overview of the impact ARB BSIs in limited resource settings derived from the existing literature. Drug resistance was associated with a substantial disease and economic burden in LMICs. Although, our results show wide heterogeneity between WHO regions, income groups, and pathogen–drug combinations. Overall, there is a paucity of BSI data from LMICs, which hinders implementation of country-specific policies and tracking of health progress.

We conducted a systematic review by searching 4 medical databases (PubMed, SCIELO, Scopus, and WHO’s Global Index Medicus; initial search n = 13,012 from their inception to August 1, 2022). We only included quantitative studies. Our final sample consisted of n = 109 articles, excluding studies from high-income countries, without our outcomes of interest, or without a clear source of bloodstream infection. Crude mortality, ICU admission, and LOS were meta-analysed using the inverse variance heterogeneity model for the general and subgroup analyses including bacterial Gram type, family, and resistance type. For economic costs, direct medical costs per bed-day were sourced from WHO-CHOICE. Mortality costs were estimated based on productivity loss from years of potential life lost due to premature mortality. All costs were in 2020 USD. We assessed studies’ quality and risk of publication bias using the MASTER framework. Multivariable meta-regressions were employed for the mortality and ICU admission outcomes only. Most included studies showed a significant increase in crude mortality (odds ratio (OR) 1.58, 95% CI [1.35 to 1.80], p < 0.001), total LOS (standardised mean difference “SMD” 0.49, 95% CI [0.20 to 0.78], p < 0.001), and ICU admission (OR 1.96, 95% CI [1.56 to 2.47], p < 0.001) for ARB versus ASB BSIs. Studies analysing Enterobacteriaceae, Acinetobacter baumanii, and Staphylococcus aureus in upper-middle-income countries from the African and Western Pacific regions showed the highest excess mortality, LOS, and ICU admission for ARB versus ASB BSIs per patient. Multivariable meta-regressions indicated that patients with resistant Acinetobacter baumanii BSIs had higher mortality odds when comparing ARB versus ASB BSI patients (OR 1.67, 95% CI [1.18 to 2.36], p 0.004). Excess direct medical costs were estimated at $12,442 (95% CI [$6,693 to $18,191]) for ARB versus ASB BSI per patient, with an average cost of $41,103 (95% CI [$30,931 to $51,274]) due to premature mortality. Limitations included the poor quality of some of the reviewed studies regarding the high risk of selective sampling or failure to adequately account for relevant confounders.

BSIs are one of the most lethal infections, having an estimated overall crude mortality of 15% to 30% [ 4 , 6 ]. BSIs are intrinsically more deadly as pathogens can spread quickly via blood, producing multiple infections and leading to organ damage and dysfunction. Extensive literature has examined the excess burden of ARB BSIs in specific locations [ 7 – 13 ]. For example, compared to their sensitive counterparts, carbapenem-resistant Klebsiella spp. [ 12 ] and methicillin-resistant Staphylococcus aureus (MRSA) [ 11 ] BSIs are associated with 9.08 (95% CI [1.17 to 70.51]) and 2.23 (95% CI [1.14 to 4.37]) times greater mortality, respectively. Higher admission to the intensive care units (ICUs), (OR 8.57; 95% CI [3.99 to 18.38]), greater length of hospital stay (LOS), (4.89 additional days; 95% CI [0.56 to 11.52]) and sizeable hospital costs ($23,318, 95% CI [$858 to $57,090]) have been linked to vancomycin-resistant versus -sensitive Enterococci BSIs [ 13 ]. Studies conducted in high-income countries contribute disproportionately to these estimates [ 14 – 16 ]; data from LMICs are scant. This comprises a critical gap in our understanding of the impact of drug-resistant BSI in countries with higher underlying health risks (e.g., cancer, neutropenia and haematological malignancies, pneumonia, and diabetes) [ 17 ].

Antibiotic-resistant bacteria (ARB) constitute a global health priority, particularly where resistance proportion is highest in low- and middle-income countries (LMICs) [ 1 ]. Resource-limited hospital infrastructure, poor health system capacity, and inadequate sanitation and hygiene infrastructure partly explain the spread and impact of ARB in LMICs [ 1 , 2 ]. Ameliorating health inequities is hampered by the feedback caused by ARB infections resulting in increased morbidity and mortality, more complicated treatments due to the use of reserved antibiotics, and prolonged hospitalisations, all of which exacerbate costs to countries’ health systems and society [ 1 , 3 ]. Recent figures from the World Health Organization (WHO) Global Antimicrobial Resistance and Surveillance System (GLASS) report show that the proportion of Escherichia coli bloodstream infections (BSIs) caused by third-generation cephalosporins resistant E. coli was more than triple in LMICs compared to high-income countries, (58.3% and 17.53%, respectively) [ 4 ]. A similar trend was observed for other WHO critical- and high-priority BSI pathogens, including Klebsiella pneumoniae and Staphylococcus aureus [ 4 , 5 ].

Excess economic costs per patient (i.e., costs associated with ARB BSI minus costs associated with ASB BSI) were computed only for excess length of stay, separated by ICU and non-ICU wards. Hospital-day costs included all the inpatient hospitality costs per patient stay for primary and secondary level and teaching hospitals and were calculated based on WHO-CHOICE costs [ 23 ]. ICU costs were calculated per patient stay for tertiary/teaching hospitals and were retrieved from the literature for countries with available information [ 24 – 36 ], or by using an approximation ratio between hospital and ICU costs [ 37 – 39 ]. Direct medical costs comprised hospital-day and ICU admission costs per patient, adjusted to their respective patients’ LOS in the hospitalised or ICU services. We also calculated excess productivity losses per patient associated with premature mortality from ARB BSIs (compared to ASB BSIs) using the life expectancy at death and human capital approaches [ 40 ]. Excess productivity losses associated with premature mortality costs were computed by multiplying the years of life lost, based on the reference standard life expectancy at the average age of death [ 41 ] from ARB BSI (i.e., costs associated with ARB BSI minus costs associated with ASB BSI), using the study-weighted average age for all patients over all studies, without age-weights and a 5% time discount [ 42 ]. All costs were expressed in 2020 USDs, adjusting for inflation using US GDP implicit price deflators. Due to a lack of data, we excluded direct and indirect nonmedical costs (e.g., travel). Cost computations and methods are detailed in S1 Text , section 4.

As a secondary analysis, we used univariate and multivariable meta-regressions to explore the main determinants of mortality and ICU admission (LOS was not included because of a small sample size). We included the bacterial family and resistance profile, demographics, and underlying health condition variables in the univariate regression. Variables were transformed to odds between ARB and ASB groups. We evaluated the associations with the original and fully imputed observations. Multiple imputations were performed using fully completed data as factors and with 1,000 repetitions following a multivariable normal regression design. Variables associated with our outcomes in the univariate analysis with p < 0.05 using non-imputed data were included in the fully imputed multivariable model.

Firstly, we employed population-weighted descriptive statistics of the health and demographic characteristics collated by studies’ patients having ARB and ASB BSIs to contrast both groups and check whether mean differences across patient features existed. Secondly, the overall estimates for excess mortality, ICU admission, and LOS associated with resistant strains compared to their sensitive counterparts were meta-analysed using the inverse variance heterogeneity model [ 22 ]. The heterogeneity was calculated using the I 2 statistics; I 2 values were classified as high (>75%), moderate (50% to 75%), and low (<50%) heterogeneity. All results were computed using odds ratios (ORs) for mortality and ICU admission rates, and the standardised mean difference (SMD) for LOS. We estimated ORs based on studies’ crude numbers or unadjusted ORs provided. Forest plots and meta-analyses were computed by outcome and subgroups of variables, including bacterial family, Gram type, reported resistance type, most common antibiotic-resistant microbial strains, World Bank income group, and WHO region. P-values (p) were reported using a two-tailed t test (p < 0.05) for the ORs for mortality and ICU admissions and LOS’s standardised mean difference. We also analysed and compared, whenever reported, the unadjusted and confounder-adjusted ORs, for studies reporting univariate and multivariable regression analyses.

We used a unified framework to evaluate the methodological quality of analytic study designs (MASTER scale) [ 21 ]. This framework comprises 36 questions classified into 7 domains concerning equal recruitment, retention, implementation, prognosis, ascertainment, sufficient analysis, and temporal precedence. Each question was scored independently by 2 reviewers as 1 if the study complied with the domain or 0 if it did not. Therefore, a higher score indicates higher study quality. Two independent reviewers performed a risk of bias assessment. Conflicts were addressed through scholarly discussion.

We extracted the following outcomes by case/control group: mortality (crude 30-day mortality, whenever available, or overall crude mortality if timing was not reported), LOS (average total days and standard deviation), and ICU admission (patients admitted). We also collected data on demographics and underlying conditions: average age, previous surgery and hospitalisation, community- or hospital-acquired BSI, any underlying condition (diabetes, hypertension, cardiovascular or heart diseases, solid tumour or malignancy, liver or kidney disease, pulmonary/respiratory diseases, and any hematologic disease), and BSI source (urinary tract, intravenous or catheter, pulmonary, and intrabdominal or gastrointestinal). Pitt bacteraemia score, APACHE II, and CHARLSON scores were collected if presented. We compared ARB and ASB groups by comparing variables’ proportion or mean using McNemar’s χ 2 or T-tests for binary and continuous data, respectively. Additionally, we classified the studies by World Bank income level, WHO region, WHO Global Priority Pathogens List, bacterium family and antibiotic class, pathogen strain, and bacterium Gram type. We used Microsoft Excel 2022 to compile and extract included articles’ data. We used double data extraction reviewing, and inconsistencies (14% disagreement) were resolved through scholarly discussion.

We extracted data including authors, publication year, country, study setting, population characteristics, bacterium type, resistance type, and sample sizes (for cases and control groups). We classified pathogen resistance based on the specific pathogen-resistance profiles evaluated in each study (e.g., cephalosporin-resistant Acinetobacter baumanii). For completeness, we also collated data on ESBL+ and non-ESBL (ESBL-) groups for gram-negative pathogens. For the analysis, the case group comprised infections with resistant strains (ARB), whereas the control group comprised sensitive-strain infections (ASB). Selected studies were organised using unique identifiers (e.g., 1, 2, 3), and sub-studies within the primary articles were classified using consecutive numbers separated by a dot (e.g., 1.1, 1.2, 1.3) if they presented bacterium- or resistance type-specific information ( S1 Data ).

To avoid our study hinging only on published articles’ results, we systematically reviewed the grey literature and other current literature reviews analysing similar topics. Four referees resolved any disagreement presented at any stage of study selection through scholarly discussion. Two native Spanish speakers fluent in Portuguese and English, a native English speaker, and a native Chinese speaker fluent in English conducted the screening and consecutive data extraction. Papers written in any other language were translated to English using Google Translate PDF (<1% of the included articles). We used the Rayyan free online tool ( https://rayyan.ai/ ) to screen, select, and decide which articles were included. Double article screening for eligibility was employed, and discrepancies were resolved via scholarly dialogue.

Studies were analysed only when the number of patients was reported. We only included the adult population (average ≥18 years of age) because (i) the number of studies focusing on children was limited (n = 4) after looking at the provisional results; and (ii) children’s inherent behaviour and exposure level differ from adults [ 3 ]. Only data on WHO-priority pathogens were retained [ 20 ]. The Results section (PRISMA chart) and Table A in S1 Text present the complete list of search criteria used.

We selected articles according to a step-guided protocol. First, articles were excluded if carried out in high-income countries; these were defined according to the 2021 World Bank classification list (i.e., gross national income “GNI” per capita > $12,696) [ 19 ]. Second, studies were only included if BSIs were presented based on laboratory-confirmed positive blood cultures. Either primary or secondary BSIs were included. Articles that analysed patients with different culture types (e.g., blood, urine, wound, nasal) were removed unless BSI episodes were clearly detailed. Third, articles were included if the ASB and ARB groups were identified among adult patients presenting BSIs in the hospital. Fourth, participants with chronic or severe diseases (e.g., HIV, cancer) were removed unless they were present in the ARB and ASB groups (e.g., studies were withdrawn if HIV–positive patients having ARB BSIs were compared with HIV–negative patients having ASB BSIs). Finally, studies were removed if they did not present our selected outcomes (i.e., mortality, ICU admission, LOS, or costs). Experimental and observational articles were included. We removed correspondence letters or opinions, short reports without data analysis, literature reviews, and single-case studies.

We searched the literature for studies examining the burden of ARB BSIs compared with ASB BSIs among inpatients from LMICs. PubMed, SCIELO, Scopus, and WHO’s Global Index Medicus (Latin American and Caribbean Health Sciences Literature “LILACs” and African Index Medicus “AIM”) were searched without restrictions to language or year of publication using a family of keywords related to antibiotic/drug-resistance, bloodstream infections/bacteraemia, and burden measures among inpatients. We searched articles published through August 1, 2022. The complete list of terms, abbreviations, and Boolean connectors used by search engine can be found in the Supporting information ( S1 Text , section 1).

When applying the leave-one-out method to our meta-analyses, we observed that after assessing the effect of every single study on the overall estimates, the numbers presented a relative variation with respect to overall estimates ranging between −2% and 4% for mortality (OR 95% CI [1.57 to 1.58]), −8% and 4% for ICU admission (OR 95% CI [1.95 to 1.97]), and −10% and 4% for LOS (SMD 95% CI [0.48 to 0.50]) ( S1 Text , section 6). These results suggest a moderate influence of our studies in the overall estimates if relative variations are compared, especially for ICU admission and LOS.

General mortality estimates from studies in China were not different from studies conducted elsewhere. However, we found larger disaggregated estimates for subgroup meta-analyses, such as Enterobacteriaceae, Moraxellaceae, Pseudomonaceae, and Staphylococcaceae species (8%, 25%, 26%, and 20%, respectively) compared to the average mortality estimates reported in Table 2 for the same subgroups. General LOS SMD was 16% higher among countries other than China, compared to the estimates reported in Table 2 , specifically driven by Moraxellaceae and Staphylococcaceae species. Finally, the odds for excess ICU admission were 25% greater in China, with respect to average ICU admission found in all included studies, driven by 27% elevated odds among patients having BSIs caused by gram-negative bacteria. Full results in Tables U and V in S1 Text .

Using the MASTER scale for methodological assessment, we calculated, on average, 25.1, 23.7, and 23.6 points (out of 36) for the mortality, ICU admission, and length of hospital stay outcomes, respectively ( Table 4 ). Our scores reflect that few studies addressed key confounders (e.g., using statistical methods to control for other correlated risk factors) to account for different prognoses and equal ascertainment (especially for participants, analysts, and caregivers’ blindness towards evaluation; <2% of included studies). Only 37%, 11%, and 13% of the studies incorporated statistical techniques (e.g., regression analyses, stratification, matching, among others) for an equal prognosis for the mortality, ICU admission, and LOS outcomes, respectively ( Table 4 , equal prognosis scores). Most studies achieved equal retention (e.g., low missing data and null attrition) and sufficient analyses safeguards (e.g., absence of numerical contradictions and data dredging), regardless of the outcome analysed. Full results are found in S1 Text sections 8 and 9 and S1 Data , Master Scale spreadsheet.

Fig 4 displays the direct medical and productivity loss due to premature mortality costs per patient by country (panel B). Direct medical costs (i.e., hospital bed-day costs and bed-day ICU costs per day multiplied by the average hospital and ICU respective LOS) were estimated at $12,442 (95% CI [$6,693 to $18,191]). The average total excess costs for a patient with ARB compared to ASB BSI, comprising direct medical and years of potential life lost, were $53,545 (95% CI [$39,838 to $67,251]). Excess costs for ICU adjusted to ICU’s length of stay were 14 times higher compared with hospital-bed LOS-adjusted among patients with ARB BSIs. Lower middle-income countries had the lowest economic burdens per patient; however, we found substantial between-country differences.

We estimated an average excess of productivity loss (indirect costs associated with ARB BSI for an average patient) from years of potential life lost due to premature mortality of $41,102 (95% CI = $30,931 to $51,274) for all bacteria combined (Table L in S1 Text ). Romania presented the highest excess producitivity lossess attributed to years of potential life-lost costs per patient, while Ethiopia had the lowest ($86,217 and $6,070, respectively). Mortality costs due to premature mortality using the life expectancy approach had an observed average of $132,560 per patient (95% CI [$99,753 to $165,363]) among all sampled countries (Table L in S1 Text ).

Tables W and X in S1 Text show the results of the univariate and multivariable meta-regressions for mortality and ICU admission, respectively. Among the variables selected from the univariate analyses, our multivariable meta-regression showed that patients with resistant Moraxellaceae BSIs and hypertension had higher mortality odds when ARB versus ASB BSI patients were compared (OR 1.67, 95% CI [1.18 to 2.36], p 0.004; OR 1.13, 95% CI [1.00 to 1.28], p 0.035, respectively). Yet, countries from the Southeast Asia WHO region displayed lower mortality odds (OR 0.62, 95% CI [0.46 to 0.85], p 0.004). For the ICU admission multivariable meta-regression, we found a weak negative association between BSIs originating as a secondary infection from the urinary tract and the odds of mortality between patients having ARB and ASB BSIs (OR 0.72, 95% CI [0.51 to 1.02], p 0.06).

The crude SMD for LOS was 0.49 (95% CI [0.20 to 0.78], p < 0.001; Table 2 , section III). In other words, the curve representing the distribution of LOS times was shifted to the right by 0.49 standard deviations for the ARB BSIs group (i.e., LOS is approximately 7 days longer for the ARB group; derived from multiplying SMD by LOS’s standard deviation among all patients [0.49*13.91]). The SMD was higher for resistant pathogens classified as WHO high-priority pathogens (or gram-positive, SMD 0.71, 95% CI [0.03 to 1.39], p 0.04) compared with WHO critical priority pathogens (or gram-negative, SMD 0.37, 95% CI [0.17 to 0.57], p 0.13). Studies reporting MRSA accounted for the greatest excess LOS estimated (SMD 0.82; Table 3 ), compared to methicillin-sensitive S. aureus. The highest excess LOS was observed in studies from Turkey (SMD 1.29). Studies from Europe (SMD 1.29) and Brazil (SMD 0.43) contributed substantially to the greater LOS in ARB BSI patients (Table G in S1 Text ).

Overall, the crude odds of ICU admission were 1.96 times higher for ARB compared to ASB BSIs (95% CI [1.56 to 2.47], p < 0.001) ( Table 2 , section II). Patients with WHO critical priority pathogens resistant to antibiotics were twice as likely to be admitted to ICU (OR 2.02, 95% CI [1.62 to 2.52], p < 0.001), with the highest observed ratio for gram-negative BSIs caused by antibiotic-resistant Enterobacteriaceae (OR 2.59, 95% CI [1.95 to 3.45], p < 0.001). Carbapenem-resistant Enterobacteriaceae in general (OR 2.66, 95% CI [1.98 to 3.57], p < 0.001), and specifically Escherichia coli (OR 3.88, 95% CI [2.74 to 5.49], p < 0.001), accounted for the highest figures ( Table 3 ). Among gram-positive bacteria, Methicillin-resistant Staphylococcus aureus had an OR of 1.91 for ICU admission rate (95% CI [0.86 to 4.25], p 0.11), and vancomycin-resistant Enterococcus faecium/faecalis had an OR of 1.48 (95% CI [0.87 to 2.54], p 0.15) ( Table 3 ). The Western Pacific region had the highest increase in ICU odds (OR 2.42, 95% CI [1.88 to 3.12], p < 0.001), followed by the Americas (OR 1.77, 95% CI [1.08 to 2.89], p 0.02), whereas the Southeast Asia region had the lowest odds of ICU admission of ARB BSIs compared to ASB BSIs (Table G in S1 Text ).

The crude OR for mortality of ARB versus ASB BSIs was 1.58 (95% CI [1.35 to 1.80], p < 0.001); we obtained similar values for gram-negative or WHO critical priority pathogens (OR 1.59, 95% CI [1.34 to 1.83], p < 0.001) ( Table 2 , section I). The highest OR of crude mortality for resistant pathogens was for carbapenem-resistant Enterobacteriaceae (OR 1.97, 95% CI [1.37 to 2.56], p < 0.001) ( Table 3 ). The impact seemed to be lower among gram-positive bacteria, with an OR of 1.51 (95% CI [0.76 to 2.26], p 0.13) for MRSA and an OR of 1.31 (95% CI [1.01 to 1.60], p 0.02) for vancomycin-resistant Enterococcus species. Compared to ASB BSIs, ARB BSIs in upper-middle-income countries (OR 1.64, 95% CI [1.36 to 1.92], p < 0.001) from Europe and Western Pacific WHO regions (OR 1.79, 95% CI [1.49 to 2.11], p < 0.001, and OR 1.66, 95% CI [1.18 to 2.14], p < 0.001, respectively) had the highest excess mortality (Table G in S1 Text ). Among priority pathogens defined by the WHO, crude excess mortality from carbapenem-resistant K. pneumoniae was substantially higher than for other pathogens (OR 1.79, 95% CI [1.15 to 2.43], p 0.002; Table 3 ), compared to sensitive counterparts. Among studies reporting both adjusted and unadjusted ORs for mortality (N = 12), we found 1.35 and 1.57 times higher unadjusted and adjusted mortality figures, respectively, for patients having BSIs caused by ARB versus ASB (Fig AJ in S1 Text ). We found lower mortality estimates among studies reporting adjusted ORs for gram-negative ARB BSIs (OR = 1.88), specifically for Enterobacteriaceae and Moraxellaceae species (OR 1.91 and OR 1.73, respectively), compared to the same unadjusted estimates (OR 2.95 and OR 3.28, respectively) (Figs AK and AL in S1 Text ). However, and surprisingly for the most part, adjusted ORs for mortality among ARB versus ASB BSI patients reflected greater odds compared to unadjusted ORs. This is explained by a single, highly influential study [ 45 ] among unadjusted estimates displaying a smaller OR (although confidence intervals overlap between unadjusted and adjusted ORs, and study’s weight is lower among adjusted estimates).

(A) Number of included studies by bacterial family (B) Number of included studies by antimicrobial susceptibility of interest (C) Number of included studies by bacterial Gram-type (D) Number of included studies by WHO priority pathogen list. Enterobacteriaceae included Escherichia coli and Klebsiella pneumoniae. Enterococcus spp. stands for Enterococcus species pluralis (multiple species), which included Enterococcus faecalis and faecium. The multiple categories stand for either multiple bacteria or antibiotics analysed throughout our selected studies, which were not reported disaggregated by bacterial family, biological strain, gram type, or WHO priority pathogen list. † Studies could include more than 1 subcategory per biological feature (i.e., a study might report Enterobacteriaceae and Pseudomonadaceae species separately in their analyses, or altogether, in which case it was classified as “Multiple,” meaning no clear distinction between subcategories). Categories might not be exclusive per study. WHO, World Health Organization.

Of the 109 articles, 100 (91.7%; 100/109) studies reported the impacts of ARB BSIs on mortality, 42 on hospital LOS, but only 18 displayed the average LOS with its standard deviation (16.5%; 18/109) and 52 (47.7%; 52/109) reported on ICU admission ( Table 1 ). Studies were primarily conducted in China (44.9%; 49/109, N = 12,092 patients), Brazil (11.9%; 13/109, N = 1,559 patients), and Turkey (8.3%; 9/109, N = 2,190 patients) ( Fig 2 ). Most studies collected data from the Western Pacific region according to the WHO classification (46.8%; 51/109) and 88% (96/109) were from upper-middle-income countries ( S1 Text , section 2). The majority of the studies reported on gram-negative bacteria, mainly Enterobacteriaceae (41.3%; 45/109), Moraxellaceae or Acinetobacter baumanii (15.6%; 17/109), and Pseudomonas aeruginosa (11.9%, 13/109) ( Fig 3 ). The main gram-positive pathogens reported were Staphylococcus aureus (19.3%; 21/109) and Enterococcus spp. (7.3%; 8/109); 75.2% (82/109) of the pathogens reported were classified as a critical priority following the WHO criteria ( Fig 3 ). β-lactam antibiotics were among the most tested antibiotic class within the studies (67.9%; 74/109), 71.6% (53/74) of which were carbapenems or cephalosporins ( Fig 3 ). The total number of patients and most prevalent features per country’s studies are reported in Table E in S1 Text . Table F in S1 Text presents the weighted unadjusted differences for sociodemographic and health variables among ARB and ASB groups. We found no statistically significant difference between ARB and ASB groups for most of these variables (χ 2 test p > 0.05). S1 Text section 2 describes the distribution of our studies by WHO region, World Bank income group, year, and outcomes densities per ARB/ASB group.

Discussion

Antibiotic resistance imposes substantial morbidity, mortality, and societal costs in LMICs [153]. Bloodstream infections with ARB are among the most lethal, imposing a large disease burden. Examining all available data for hospitalised patients in LMICs, we found that ARB BSIs with WHO critical- and high-priority pathogens were associated with increased mortality (OR 1.58, 95% CI [1.35 to 1.80]), overall length of stay (SMD 0.49, 95% CI [0.20 to 0.78]), and ICU admission (OR 1.96, 95% CI [1.56 to 2.47]).

Our findings on mortality are consistent with the recent estimates by the Global Burden of Disease study [154]. The largest mortality impact was associated with resistant A. baumannii and Enterobacteriaceae. Both bacteria featured in the global top 5 contributors to resistance-associated and -attributable deaths in 2019 [154]. Between a quarter and half of the patients with ARB BSIs caused by Enterobacteriaceae, A. baumannii or P. aureginosa die, corroborating findings from different country settings for Enterobacteriaceae [8,67], P. aeruginosa [155], and large university hospitals in Israel and the US for A. baumanii [156,157].

Our results suggest that patients who acquired ARB BSIs during their hospital stay had an overall hospital stay that is about a week longer than patients that acquired ASB BSIs. However, in our study, we could not distinguish between excess length of stay before or after BSI, and as such this is likely an overestimation. Depending on the pathogen, resistant infections have previously been shown to increase LOS typically by 2.0 to 12.7 days [158]. Longer hospital stay, especially before BSI onset, is a primary risk factor for acquiring a resistant infection due to the cumulative risk of hospital transmission of ARBs [158,159]. We found that MRSA had the greatest impact on LOS (extending stay by 14 days relative to sensitive S. aureus). Others have also shown considerably increased LOS as a result of MRSA compared with sensitive S. aureus: Tsuzuki and colleagues [160] showed an excess overall LOS and LOS after BSI onset of 20 and 7 days, respectively; similarly, Graffunder and colleagues [161] showed MRSA patients presented an overall LOS of 3 weeks longer. Resistant infections are more difficult to treat and increase the rate of ICU admissions. Our analysis showed that resistant Enterobacteriaceae infections more than doubled the odds of ICU admission. This finding is comparable with the 2.69 higher odds of ICU admission previously shown among patients with carbapenem-resistant K. pneumoniae BSIs [162]. Our exploratory analysis for studies performed in China and LMICs other than China exhibited divergent results. We found that China’s patients with antibiotic-resistant gram-negative BSIs (A. baumanii, Enterobacteriaceae, and P. aeruginosa) displayed higher excess mortality, ICU admission, and LOS, compared to the other LMICs with reported data. Large increases in antibiotic consumption and resistance levels over the last 20 years and the rapid development or acquisition of drug resistance among gram-negative pathogens might explain the greater excess mortality and morbidity for ARB BSIs in China [1,163,164]. Correspondingly, inappropriate administration of empirical treatments and low testing rates could increase the burden outcomes for patients with ARB BSIs in these settings [165].

Despite being fundamental to resource allocation for healthcare provision, we found very little data on excess costs associated with ARB BSIs among the reviewed studies. One study conducted in Thailand, reported excess costs associated with hospital-acquired carbapenem-resistant A. baumannii of $5,682 [61]. A study conducted in Colombia, reported excess hospitalisation costs associated with MRSA BSI of $10,212, compared to sensitive S. aureus [53]. We estimated costs associated with mortality, LOS, and ICU admissions from the provider and societal perspective following the WHO-CHOICE standards and human capital approach. We found that the average hospital-related 2020 USD excess costs were $12,442 (95% CI [$6,693 to $18,190]) per ARB BSI patient, compared to ASB, ranging between Ethiopia, with the lowest figures, to Mexico, with the highest. These differences are partly explained by the countries’ disparate economies (Pearson correlation = 0.27 between GDP and hospital costs). Several LMIC-setting studies detailing excess costs of resistant infections were excluded from our review because they did not meet specific inclusion criteria. Cost estimates from these studies include 1 from Turkey in which excess hospital stay and treatment costs were $10,002 [166]. Our estimate for Turkey of $10,403 is similar; however, our estimates did not include therapy/treatment costs. Our estimate for China ($12,516) was higher than a previous study including BSI treatment costs for carbapenem-resistant K. pneumoniae ($10,763) [167]. The average excess total costs comprising direct medical costs and years of potential life lost associated with premature mortality were $53,545 (95% CI [$39,838 to $67,251]) per patient with ARB BSI. WHO [168] recently reported that 58.3% of 22,371 isolates were identified as ARB E. coli, while 33.3% of 23,031 isolates were ARB S. aureus in LMICs, indicating the high relevance of these costs.

This study has limitations. First, the most important limitation is consistent with conclusions from the Global Burden of Diseases study [154]: there is a sparsity of data on ARB from LMICs. Only 18 of the 137 (13%) LMICs published any AMR outcome study. Consistent antibiotic resistance surveillance puts demands on clinical bacteriology, quality control, and data linkage between culture test results and clinical outcomes, which is beyond the capabilities of many LMICs. Applying the leave-one-out method to our meta-analyses (S1 Text, section 6) showed a minor-to-moderate influence of individual studies likely due to the heterogeneity in clinical settings, indicating that our model’s results are robust (assuming countries’ missing information and selection biases are heterogeneously distributed). Future efforts to improve coverage should prioritise WHO’s Africa region, where data were remarkably absent, with no estimates for resistance-associated LOS or ICU admissions. Our results indicate that the studies from the Western Pacific and European areas show the highest excess mortality from ARB BSIs. Studies from Africa show among the lowest but this region has limited data and substantial uncertainty; it is essential to improve epidemiological surveillance of ARB BSIs in this region in particular [169]. Second, some articles were of low quality or reported limited data. Studies often failed to account for confounding factors; hence our analyses relied upon crude estimates. ARB surveillance networks vary in blood culture sampling, potentially overestimating the number of severe cases if selective sampling among patients fulfilling the case definition is present. Third, we did not estimate the total relative harm of ARB BSIs relative to where such infections were prevented (compared to non-infected patients) [170], primarily because of the limited number of studies [171]. While we accounted for some key risk factors when comparing antibiotic-sensitive and antibiotic-resistant groups in the metaregression, others were unavailable. We could not match comparison groups by factors known to impact patients’ underlying health conditions, such as illness severity, prolonged previous hospital stays, or the use of invasive devices. The reported LOS does not distinguish between total LOS and LOS following BSI infection, thus risking reverse causality [172]. This ecological study was designed to identify associations; consequently, our results should be interpreted cautiously. Also, we adjusted WHO-CHOICE country estimates using US GPD implicit price deflators, which may not necessarily reflect price changes in some LMICs, particularly for non-tradable cost components of healthcare. Finally, we may have overestimated the true effect size of the association between ARB BSIs and mortality as indicated by the exploratory analysis of studies’ adjusted—compared to unadjusted—ORs reporting both estimates, specifically among gram-negative species.

Here, we described an updated evaluation of the health impact and excess economic costs of resistant BSIs in low-resourced settings. Our results highlight regions where improved surveillance, expanding microbiology laboratory capacity, and data collection systems are most needed and where the current evidence indicates WHO critical and high-priority drug-resistant pathogens exert the greatest toll on morbidity and mortality.

[END]
---
[1] Url: https://journals.plos.org/plosmedicine/article?id=10.1371/journal.pmed.1004199

Published and (C) by PLOS One
Content appears here under this condition or license: Creative Commons - Attribution BY 4.0.

via Magical.Fish Gopher News Feeds:
gopher://magical.fish/1/feeds/news/plosone/

You are viewing proxied material from magical.fish. The copyright of proxied material belongs to its original authors. Any comments or complaints in relation to proxied material should be directed to the original authors of the content concerned. Please see the disclaimer for more details.