Attributable mortality due to nosocomial sepsis in Brazilian hospitals: a case–control study

Background Nosocomial sepsis is a major healthcare issue, but there are few data on estimates of its attributable mortality. We aimed to estimate attributable mortality fraction (AF) due to nosocomial sepsis. Methods Matched 1:1 case–control study in 37 hospitals in Brazil. Hospitalized patients in participating hospitals were included. Cases were hospital non-survivors and controls were hospital survivors, which were matched by admission type and date of discharge. Exposure was defined as occurrence of nosocomial sepsis, defined as antibiotic prescription plus presence of organ dysfunction attributed to sepsis without an alternative reason for organ failure; alternative definitions were explored. Main outcome measurement was nosocomial sepsis-attributable fractions, estimated using inversed-weight probabilities methods using generalized mixed model considering time-dependency of sepsis occurrence. Results 3588 patients from 37 hospitals were included. Mean age was 63 years and 48.8% were female at birth. 470 sepsis episodes occurred in 388 patients (311 in cases and 77 in control group), with pneumonia being the most common source of infection (44.3%). Average AF for sepsis mortality was 0.076 (95% CI 0.068–0.084) for medical admissions; 0.043 (95% CI 0.032–0.055) for elective surgical admissions; and 0.036 (95% CI 0.017–0.055) for emergency surgeries. In a time-dependent analysis, AF for sepsis rose linearly for medical admissions, reaching close to 0.12 on day 28; AF plateaued earlier for other admission types (0.04 for elective surgery and 0.07 for urgent surgery). Alternative sepsis definitions yield different estimates. Conclusion The impact of nosocomial sepsis on outcome is more pronounced in medical admissions and tends to increase over time. The results, however, are sensitive to sepsis definitions. Supplementary Information The online version contains supplementary material available at 10.1186/s13613-023-01123-y.


Title IMPACTO-MAPA Study Identifying the Mortality Attributable to Sepsis in Hospitalized Patients in Brazil A study of the Platform of projects in support of the National Plan of Action for the Prevention and Control of Antimicrobial Resistance -IMPACTO MR Program Coordinating Center
Institute of Research of the Hospital do Coração (IP-HCor) Rua Abílio Soares, 250 -Paraíso CEP: 04005-000 São Paulo, SP -Brazil Phone: ++55 11 3053 6611 Ext: 8210 Fax: ++55 11 3886 4695

Study Design
Observational case-control study Primary objective To measure the mortality attributable to sepsis in hospitalized patients in Brazil.

1.
To evaluate the length of hospital stay in patients with or without sepsis 2.
To evaluate the mortality attributable to other hospitalization complications (infarction, cerebrovascular accident, pulmonary thromboembolism) 3.
To evaluate the completeness quality of the medical record data 4.
To evaluate the quality and consistency of the filling of the declaration of hospital deaths Eligibility of cases and controls Each hospital will provide clinical, demographic, and laboratory data for the last 50 deaths at the institution. The controls will be the data of the patient who was discharged alive with the closest date and time to the date of death of the case. Among the most frequent complications, we can highlight the occurrence of nosocomial infection. In developing countries, nosocomial infections are extremely frequent and affect more than 15 out of every 100 hospitalized patients [1]. In some more severe situations, the infection causes the dysfunction of one or more organs, a situation called sepsis and which has high morbidity and mortality [2]. Despite its frequency, the mortality attributable to sepsis, that is, the increase in mortality resulting from the occurrence of nosocomial sepsis in a patient is unknown.
There is a great amount of data on the prevalence and incidence of sepsis and the mortality of patients with sepsis, including in Brazil [3]. We also have some data on the mortality attributable to sepsis among patients with sepsis, that is, the total number of deaths that can be attributed directly to sepsis rather than other concomitant problems. For example, in patients admitted to the ICU, the mortality fraction attributable to sepsis is approximately 15% in developed countries [4]. Case-control study to be performed using medical record analysis.

Sample size
Sample calculation was performed using the Fleiss, Tytun, and Ury method to estimate the power of a two-tailed test for proportion difference using the Hmisc package of R software (version 3.5.0). Sample size was estimated for a power above 90% for an odds ratio of at least 1.4 assuming an incidence of nosocomial sepsis of 15% in the controls. Thus, 2,700 patients would be needed. We chose to round the number to 3,000 so as to ensure a margin of safety. This total of patients will be included in a group of 30 Brazilian hospitals that fulfill the following characteristics: 1. At least 100 active beds 2. Presence of intensive care unit and emergency room 3. Availability to collect data for the study Each hospital will provide clinical, demographic, and laboratory data for the last 50 deaths at the institution. The controls will be the data of the patient who was discharged alive with the closest date and time to the date of death of the case. Fifty controls per hospital will be used, amounting to 100 patients per hospital. The consecutive pairing with the closest living discharge will ensure the minimization of any temporal impacts.

Data collection
Data collection will be performed by each hospital using an electronic data collection system designed specifically for the study. The filling in will be done by each center based on medical record data, according to local availability, respecting the maximum term of two years. All data collected will be anonymous, in order to preserve the identity of the participants and to ensure confidentiality. The data to be collected are: Laboratory and physiological variables will be collected from the admission and during the first three days of the patient's hospitalization.
When an antimicrobial treatment is added, started, or changed and/or some type of microbiological examination (culture) is performed, we will again collect physiological and laboratory data for an additional period of three days before and after the date of the occurrence to follow the occurrence of organ dysfunction associated with suspected infection (sepsis). In these situations, we will also note: 1. The cause of sepsis 2. The prescribed antibiotic and its suitability (i.e. whether the antibiotic prescribed at the initial time was effective against the pathogen that eventually proved to be the cause of the event) 3. Whether or not the infection was associated with a surgical procedure or invasive device (prosthesis infection, catheter-related bloodstream infection, etc.).
We will also note the occurrence of other clinical complications, such as: 1. Pulmonary thromboembolism 2. Myocardial ischemia/Myocardial infarction

Cerebrovascular accident (ischemic or hemorrhagic)
For the cases (that is, for hospital deaths), we will collect the information present in the death certificate (immediate cause and its sequence, as well as contributing causes).
Definition of the exposure factor: The exposure of interest is the nosocomial sepsis, that is, nosocomial infection that occurs with the occurrence of organ dysfunction. As organ dysfunction, we will consider the presence of at least one of the following factors when the therapeutic antimicrobial treatment is started or changed in hospitalized patients:

Statistical analysis
The traditional univariate analysis comparing the groups (case and control) related to the variables described above will be performed using ttest or Mann-Whitney U test for continuous variables (as appropriate depending on the occurrence of normality) and Chi square test for comparison of proportions.
The main analysis (attributable mortality) will be done using logistic regression with predictors of event (hospital death) as predictors. Predictors will be chosen based on their clinical relevance, which will include: We plan to carry out a marginal structural model (MSM) analysis that modulates the daily sequential data of the patient with outcomes. Through this approach, we can use the data collected on the days before and after the infection for more robust causality inference. The MSM design will be based on the behavior of the collected physiological and laboratory variables.
Some sensitivity analyses will be performed. We plan to analyze the mortality attributable to sepsis according to specific infectious foci (bloodstream infection and nosocomial pneumonia) and according to the presence of infections by multidrug-resistant bacteria. We will also evaluate whether the use of appropriate antibiotics modulates the association between nosocomial sepsis and outcome.
Additional Objectives: The study will provide regional data on the occurrence of complications in hospitalized patients; thus, a better targeting of resources for prevention may be carried out in the future. We plan to analyze the quality of the data included in patient records. The review of the medical records of 3,000 hospitalized patients will also allow us to point out the quality of the data completeness, eventual failures in the filing of death certificates, among others. This may also assist in developing campaigns to improve the clinical records of hospitalized patients.

Data management
All data will be included in a cloud-based electronic system developed specifically for the study. Measures to ensure the filling and quality of the data will be used, including: 1. Face-to-face and/or distance training for all researchers

Study Schedule
The total duration of the study is three years, with the following schedule: Year X X X X X X Approval of centers Initial submission of the protocol to the coordinating center X X X X X X Follow-up of protocol submissions and approvals in the centers X X X X X X Submission of partial and final reports to CEP X X X X X X Payment of fee to CEP X X X X X X Management of centers (inclusion of patients) Selection of centers X X Development and printing of support material X Training of centers X X X X X X Stimulation of patient inclusion X X X X X X Assistance to researchers in conducting the protocol X X X X X X Monitoring X X X X X X Payment of researchers X X X X X X Data Management and analysis of results CRF elaboration (printed and electronic) X X X System maintenance X X X X X X Cleaning of data X X X X X Closing of the database X X Data analysis X Publication Submission of results for publication/congress X

Dissemination of Results
The Steering Committee of the IMPACTO MR-MAPA Study undertakes to publish its results, whatever they may be. Because it will be a large-scale randomized collaborative study, we aim to refer the main publications to high impact journals. The study data will be made available immediately to the Ministry of Health.
We intend to give access to the study database to other researchers.
In the first two years after the publication of the main study, researchers will focus on sub-studies proposed by collaborative researchers. At this stage, the database will be kept under the custody of the study coordinators, and access will be allowed to third parties only upon express authorization of the Steering Committee of the IMPACTO MR-MAPA study after evaluation of the proposal accompanied by a statistical analysis plan.

Project Risks
We believe that the IMPACTO MR-MAPA study presents a low risk.
We will perform a retrospective analysis using medical records in 30 Brazilian hospitals. Data collection will be done anonymously through a computerized central system. All analyses will be performed without exposing the patients' names or individualized information of the participating hospitals. The data collection system will be designed in a way that minimizes the input of incorrect information. The data inserted will be checked by the data management of the study.

Ethical considerations
This will be a case-control study, using a retrospective analysis of medical record data, with anonymous data collection. In this way, we anticipate that the risk for patients and institutions is minimal. Because of the very nature of the study, we will request exemption from obtaining the informed consent from the centers. There will be no contact or interaction with the included patients.
PREvalence Assessment Database, SPREAD): an observational study. To measure attributable mortality from nosocomial sepsis in hospitalized patients in Brazil. secondary objectives

1.
Assess length of hospital stay in patients with or without sepsis 2.
Evaluate the quality of completion of medical records 4.
Assess the quality and consistency of filling in the hospital death certificate

Eligibility of cases and controls
Each hospital will provide clinical, demographic and laboratory data for the last 50 deaths that occurred in the institution. Data from the patient who was discharged alive with a date and time closer to the date of death of the case will be used as controls. Among the most frequent complications, the occurrence of nosocomial infection stands out. In developing countries, nosocomial infections are extremely frequent, reaching more than 15 in every 100 hospitalized patients [1]. In some more severe situations, the infectious condition generates dysfunction of one or more organs, a situation called sepsis and which has high morbidity and mortality [2]. Despite its frequency, the attributable mortality of sepsis, that is, the increase in mortality that results from the occurrence of nosocomial sepsis in a patient, is unknown.
We have a lot of information on the prevalence and incidence of sepsis, and lethality in patients with sepsis, including good data from Brazil [3]. We also have some data on mortality attributable to sepsis among patients with sepsis, that is, of the total number of deaths that can be directly attribute to sepsis rather than other concomitant problems. For example, in patients admitted to the ICU, the mortality fraction attributable to sepsis is approximately 15% in developed countries [4]

Study design
Case-control study to be carried out through analysis of medical records.

Sample size
The sample size calculation was performed using the Fleiss , Tytun and Ury method to estimate the power of a two-tailed test for proportion difference using the Hmisc package of the R software (version 3.5.0). The sample size was estimated for a power above 90% to detect an odds ratio of at least 1.4 assuming an incidence of nosocomial sepsis in controls of 15%. Two thousand seven hundred patients would be needed. We have chosen to round the number to 3,000 in order to ensure a safety margin.

Eligibility criteria for institutions
This total of patients will be included in a group of 30 Brazilian hospitals that meet the following characteristics: 1. At least 100 active beds 2. Presence of intensive care unit and /ouemergency room 3. Availability to collect data for the study Eligibility criteria for patients Inclusion criteria Each hospital will provide clinical, demographic and laboratory data for the last 50 deaths that occurred in the institution. Data from the patient who was discharged alive with a date and time closer to the date of death of the case will be used as controls. Fifty controls will be used per hospital, totaling 100 patients per hospital. Consecutive matching with the closest live high will ensure that any temporal impacts are minimized.
In addition to the general criteria mentioned, the specific inclusion criteria are: • Patients over 18 years of age; • Hospital stay for more than 24 hours;

Exclusion criteria
Discharge cases due to transfer, evasion or home care should be excluded..

Data collection
Data collection will be performed by each hospital using an electronic data collection system designed specifically for the study. The filling will be done by each center based on data from medical records, according to local availability, respecting the maximum period of completion of two years. All data collected will be anonymous, in order to preserve the identity of the participants and guarantee confidentiality. The data to be collected are: Laboratory and physiological variables will be collected from admission and during the first three days of the patient's hospitalization. In the event of addition, initiation or change of antimicrobial regimen and/or collection of some type of microbiological examination (culture), we will perform physiological and laboratory data collection again for an additional period of three days before and after the date of your occurrence. in order to monitor the occurrence of organic dysfunction associated with a suspected infectious condition (sepsis). In these situations, we will also note: 1. The causative focus of sepsis 2. The antibiotic prescribed and its suitability (i.e., whether the antibiotic prescribed at the initial time was effective against the pathogen that eventually proved to be the cause of the event ) 3. Whether or not the infection was associated with a surgical procedure or invasive device (prosthesis infection, catheter-related bloodstream infection, etc. ).
We will also note the occurrence of other clinical complications, such as: 1. Pulmonary thromboembolism 2. Myocardial ischemia/myocardial infarction 3. Stroke (ischemic or hemorrhagic) For cases (that is, for hospital deaths), we will collect the information contained in the death certificate (immediate cause and its sequence, as well as contributing causes).
Definition of the exposure factor: The exposure of interest is nosocomial sepsis, that is, a nosocomial infection that courses with the occurrence of organic dysfunction. As organic dysfunction, we will consider the presence of at least one of the factors below when starting or changing the therapeutic antimicrobial regimen in a hospitalized patient: We will also present the marginal effect of having one sepsis episode on the whole period, with an estimate of average attributable fraction obtained from the model. All analyses will be performed using the R software Sensitivity analysis: Different definitions of sepsis will be explored; the same analysis method described above will be used for different definitions as sensitivity analyses. An internal consensus was created within the steering committee on probability degrees of sepsis according to combinations of use of antibiotics, organ failure, results of cultures, and occurrence of clinically relevant events. Two sensitivity analyses will be made, one estimating the attributable fraction considering definitive, very probable, and probable sepsis, and a second sensitivity analysis considering only definitive and very probable sepsis.
Data management

Dissemination of Results
The Steering Committee of the IMPACTO MR-MAPA Study is committed to publishing its results, whatever they may be. As this is a largescale, collaborative, randomized study, we aim to refer the main publications to high-impact journals. Study data will be immediately made available to the Ministry of Health.
We intend to open access to the study database to other researchers. In the first two years after the publication of the main manuscript, the investigators will dedicate themselves to carrying out analyzes for substudies proposed by investigators of the collaborative group. At this stage, the database will be kept under the custody of the study coordinators, and its access will be allowed to third parties only with the express authorization of the Steering Committee of the IMPACTO MR-MAPA study after evaluation of the proposal accompanied by a statistical analysis plan.

Project Risks
We believe that the IMPACTO MR-MAPA study presents a low risk.
We will perform a retrospective analysis using medical records in 30 Brazilian hospitals. Data collection will be done anonymously through a computerized central system. All analyses will be performed without exposing the patients' names or individualized information of the participating hospitals. The data collection system will be designed in a way that minimizes the input of incorrect information. The data inserted will be checked by the data management of the study.

Ethical considerations
This is a case-control study, through a retrospective analysis of medical record data, with anonymous data collection. Thus, we anticipate that the risk to patients and institutions is minimal. Due to the very nature of the study, we will request exemption from obtaining the free and informed consent form from the centers. There will be no contact or interaction with the patients included.