The use of antibiotics has been considered an important driver of bacterial resistance since they were introduced to modern medicine.1 Given that these drugs are necessary for human and animal care as well as to ensure food safety, their use must meet best practices of consumption and be treated as a common good. Bacterial infections due to resistant microorganisms, like other infectious diseases, have been demonstrated to be associated with socioeconomic status, access and quality of healthcare.2 Furthermore, the burden of bacterial resistance is different in different areas of the globe.3 In this sense, understanding the determinants of bacterial resistance in a given area is important to support public policies leading to more efficient control of its spread.

The World Health Organization as well as Infectious Diseases Societies have proposed that governments coordinate actions to minimize and limit the spread of antimicrobial resistance.4,5 In Brazil, a law restricting over-the-counter sales of antibiotics was passed by the federal government in 2010, which had a variable impact on sales in different geographic areas of the country.6 Another study, investigating socioeconomic determinants of antibiotic use in the state of São Paulo, Brazil, found an association between higher consumption of antimicrobial drugs and markers of social inclusion such as higher human development index, density of private health establishments, and lower illiteracy.7 To date, no studies were published investigating the impact of this Brazilian law on the occurrence of multidrug resistant organisms (MDRO) either in the community or in the healthcare setting.

In this study we explored the impact of overall antibiotic use, i.e. the sum of what is consumed at the community and in hospitals, on the incidence of infections caused by MDRO within intensive care units in the same area and adjusted for socioeconomic indexes, quality and access to healthcare.

The aim of this study was to evaluate the impact of overall antimicrobial use (community plus hospitals) on the incidence of bloodstream infections at intensive care units adjusted by socioeconomic factors, quality of healthcare, and access to the healthcare system.

Despite efforts of medical societies and governments, excessive prescription of antimicrobials is still frequent and contributes to expose patients to unnecessary adverse effects, increasing costs and the burden of bacterial resistance.1,13 This study aimed to explore the association between the burden of total antimicrobial use, mixing community and hospital use, and its impact on bacterial resistance at the hospital level adjusted for socioeconomic indices, access and quality of the healthcare system. Although we found an association between antimicrobial use and resistance, it was not consistent for all combinations of antimicrobial groups and pathogens.

As individual antimicrobial compounds were highly correlated, we used a statistical procedure to cluster them in four groups. This strategy aimed to reduce the mass of data on the use of antibiotics, with a minimum possible loss of information. We examined antibiotic use during the previous year in relation to our primary outcome i.e. incidence of bacteremia caused by MDRO in the intensive care unit. Whether or not a one-year gap is enough to demonstrate such association remains a matter of debate.1

Antimicrobials of group 1 were associated with MRSA and carbapenem-resistant Acinetobacter spp. interestingly, most group 1 antibiotics had broad spectrum and are predominantly for hospital use only. Broad spectrum hospital use of antimicrobials and incidence of MDRO infection and/or colonization is an association demonstrated in other ecologic studies.14,15 For all other groups, the associations were inversely proportional. Moreover, no association was observed for third generation cephalosporin-resistant E. coli and Klebsiella spp. This lack of consistent association between overall use of antibiotics and bacterial resistance in ICUs, led us to hypothesize that community and hospital environments function as two different compartments for selective pressure. Concurring to this idea, selective pressure exerted at community level seems to affect bacterial resistance at this level as already demonstrated in other ecologic studies.16 On the other hand, selective pressure exerted within hospitals would lead to a greater resistance burden in this scenario14 with interhospital spread of resistant strains as demonstrated by David et al.17 Based on our findings, community use of antimicrobials does not seem to have a direct effect on bacterial resistance in critical patients in ICUs, even when adjusted by socioeconomic indices and access and quality of healthcare.

Socioeconomic covariates were used for adjustment in this study. The hierarchical model chosen to explain the relationship of the multiple explanatory factors and the outcome requires a conceptual framework for epidemiological analysis that assumes that distal factors can affect the outcome directly or indirectly mediated by other factors (medial and/or proximal).18 In this logic, the initial model tests distal factors only. A second model tests the medial factor (properly) adjusted for the distal factor. A third model tests the proximal factor (properly) adjusted for the distal and medial factors. In our study, socioeconomic factors were classified as distal factors because, at least in theory, health indexes (medial factors) are not real confounders, rather they are partly determined by them. The same was assumed for the use of antimicrobials in relation to the distal and medial factors.

Interestingly, our findings demonstrated consistently higher incidences of most MDROs among public hospitals and lower incidences among catholic hospitals for all studied pathogens. For public hospitals, we hypothesize cross transmission having an incremental impact on the burden of resistance. Most of public hospitals have overcrowded emergency departments in which identifying patients harboring MDRO and controlling cross-transmission is a huge challenge.19 The lower incidence of MDRO infections among catholic hospitals are more difficult to explain. Although these hospitals merge care delivery to both private and public sectors, in the state of São Paulo these hospitals are located mostly in small towns. Among the 57 catholic hospitals in the state of São Paulo, 27 (47%) are the only hospital in the city in which they are located (data from Sao Paulo Health Care Department). Most of them are overcrowded as well have financial constraints as recently exposed by the lay media in Brazil.20 In this way, we would expect to see similar incidences of BSI caused by MDRO in catholic and public hospitals. One possible explanation for their low incidence of MDRO might be poor laboratory diagnostic performance among catholic hospitals due to shortage of financial resources. Costa et al.21 found an association between laboratories serving hospitals located outside state capitals in Brazil and low laboratory quality. Further investigation is needed to clarify this finding.

Fifty-three percent of ICUs had 10 or less beds. We believe that this distribution reflects the organization of the Brazilian healthcare system. Tertiary health care is concentrated in large and medium-sized cities. Thus, most municipalities have few ICU beds available. Moreover, 146 out of 645 municipalities host all ICUs in the State of São Paulo. In addition, this indicates inequality in access, especially to tertiary care, of the Brazilian population. Although the State of São Paulo State is the wealthiest and the number of ICU beds per inhabitant is above the national rate, there is a huge difference in access for patients from the private and public sectors, namel 29 versus 14.9/100.000 inhabitants.22 The relationship between the number of ICU beds and the incidence of infections caused by multi-drug resistant organisms is not established in the literature. Even though, Correa et al. found that multi-drug resistance can be a problem even in small hospitals.23

Human development index (HDI) is the socioeconomic indicator most positively correlated (i.e. as HDI increases, the incidence of BSI caused by some MDRO also increases, e.g. CR Acinetobacter sp. and P. aeruginosa, 3rd generation cephalosporin-resistant E. coli and vancomycin-resistant Enterococcus sp.). HDI is a composite of life expectancy at birth plus average income plus educational level and taken in a wider perspective this indicator is associated to equity within a society. Thus, instead of causality it might function as a marker of risky areas for resistance occurrence. Interestingly, MRSA was not associated with higher HDI. This finding is in line with the report by Auguet et al.2 that analyzed a potential association between MRSA carriage and markers of social and material deprivation. The indicators of quality and access of healthcare as well as markers of complexity were not correlated to MDRO incidences. It is important to highlight that socioeconomic covariates were used for adjustment in this study and their contribution as independent factors require testing in further studies.

This study has limitations. First, there is always a possibility of ecologic fallacy in such a design. Thus, we cannot assume that associations raised in aggregated data analyses necessarily occur at the individual level. Concerning the impact of antibiotic use on bacterial resistance, it is difficult to establish an ideal timeframe to explore the gap between antibiotic exposure and laboratory identified resistance. A one-year time lag between use and incidence of BSI caused by MDRO was analyzed. Although other studies were based on 1–2-year time lag, the best approach is not clear.16

Moreover, we assumed that sales in the region were equal to antibiotic use. Patients’ compliance with prescriptions, especially among outpatients, was not evaluated and is indeed hard to achieve since a good surrogate for antibiotic use in large population is lacking. Patient transfers between different hospitals are not formally established on a reference and counter-reference basis in the region. Thus, adjustment for the area-level hospital attendance in order to model potential transmission among facilities was not possible. Furthermore, we have not addressed the antimicrobial stewardship initiatives and policies for preventing HAIs and the spread of MDRO within the studied hospitals. These issues may have had some impact on the outcome, but we could not include these data to adjust for overall antimicrobial use.

Finally, as we did not characterize genotypically the isolates assumptions about transmission between facilities or about a pathway from community to healthcare system were not possible.

We examined a potential association between overall antibiotic use in the state of São Paulo and bacterial resistance within the intensive care setting after adjustment for socioeconomic covariates. The observed association was poor and not consistent for all combinations of antimicrobial groups and pathogens. No association was observed for third generation cephalosporin-resistant K. pneumoniae and E. coli.

Selective pressure exerted at the community level seemed not to affect the incidences of MDRO infection observed in intensive care setting.

The author declares no conflicts of interest.

Ícaro Boszczowski: Conceptualization, Data curation, Formal analysis, Funding acquisition, Investigation, Methodology, Project administration, Resources, Supervision, Validation. Francisco Chiaravalloti Neto: Conceptualization, Data curation, Formal analysis, Funding acquisition, Methodology, Resources, Software, Supervision, Validation. Marta Blangiardo: Formal analysis, Methodology. Oswaldo Santos Baquero: Formal analysis, Methodology. Geraldine Madalosso: Conceptualization, Resources. Denise Brandão de Assis: Conceptualization, Resources. Thais Olitta: Data curation, Resources. Anna S. Levin: Conceptualization, Formal analysis, Funding acquisition, Methodology, Project administration, Resources, Supervision, Validation.