Article Text
PDF
Abstract
Objective The important effect modifiers of high disease severity on the relationship between the different volumes of early fluid resuscitation and prognosis in septic patients are unknown. Thus, this study was designed to assess whether the efficacy of different volumes in the early fluid resuscitation treatment of sepsis is affected by disease severity.
Design Retrospective cohort study.
Setting Adult intensive care unit (ICU) patients with sepsis from 2001 to 2012 in the MIMIC-III database.
Interventions The intravenous fluid volume within 6 hours after the sepsis diagnosis serves as the primary exposure. The patients were divided into the standard (≥ 30 mL/kg) and restrict (<30 mL/kg) groups. Disease severity was defined by the sequential organ failure assessment (SOFA) score at ICU admission. Propensity score matching analysis was performed to ensure the robustness of our results.
Primary and secondary outcome measures The primary endpoint of this study was 28-day mortality. Days without needing mechanical ventilation or vasopressor administration within 28-day of ICU admission serving as the secondary endpoint.
Results In total, 5154 consecutive individuals were identified in data analysis, 776 patients had a primary end-point event, 386 (49.68%) in the restrict group and 387 (49.81%) in the standard group. Compared with the restrict group, the standard group had higher 28-day mortality (adjusted HR, 1.32; 95% CI 1.03 to 1.70; p=0.03) in the subgroup with a sequential organ failure assessment (SOFA) score ≥10. By contrast, the risk of mortality reduction was modest in the subgroup with an SOFA score <10 (adjusted HR, 0.85; 95% CI 0.70 to 1.03; p=0.10). The effect of the interaction between the SOFA score and fluid resuscitation strategies on the 28-day mortality was significant (p=0.0035).
Conclusions High disease severity modifies the relationship between the volume of fluid resuscitation and mortality in patients with sepsis in the ICU; future studies investigating this interaction are warranted.
- Adult intensive & critical care
- INTENSIVE & CRITICAL CARE
- PRIMARY CARE
Data availability statement
No data are available.
This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See: http://creativecommons.org/licenses/by-nc/4.0/.
Statistics from Altmetric.com
Strengths and limitations of this study
A large and sufficient number of intensive care unit cases with sepsis were collected from the MIMIC-III database.
The use of a well-characterised data set, case section using an objective definition, reporting according to Strengthening the Reporting of Observational Studies in Epidemiology guidelines, large sample size and sensitivity analysis using propensity-matched cohort.
The efficacy of different volumes in the early fluid resuscitation treatment of sepsis is affected by disease severity.
Analysis of the effect of crystalloids on early resuscitation of sepsis only, without going into the analysis of colloids and other blood products.
The results of this study are descriptive, and we cannot determine causality through this research design.
Introduction
Similar to all infectious diseases, the earlier the treatment of sepsis, the better the outcome. Early goal-directed therapy (EGDT) is a protocolised approach to treat sepsis. In 2001, Rivers et al reported that EGDT could significantly reduce patient mortality in early haemodynamic recovery from severe sepsis and septic shock.1 Subsequent nonrandomised controlled studies have shown that EGDT helps improve patient survival.2–4 Thus, EGDT was included in the treatment guidelines for sepsis.5 However, an increasing number of observational studies have reported potential adverse effects, such as hypertension, peripheral oedema, pulmonary oedema,6 respiratory failure7 and increased heart demand related to excessive fluid in patients with sepsis.
From 2014 to 2015, three large multicentre trials were published to evaluate the results of EGDTs. These trials—the Protocolized Care for Early Septic Shock (ProCESS),8 Australasian Resuscitation in Sepsis Evaluation9 and Protocolized Management in Sepsis10—while reporting all-time low sepsis mortality, question the continued need for all of the elements of EGDT or the need for protocolised care for patients with severe and septic shock. One of the possible reasons that these studies failed to determine that a uniform result is that sepsis is a heterogeneous disease,1 9 11 originating from different infectious sites, organism types, genetic backgrounds and host coexistences.
The ProCESS trial patients with the lowest biomarker quartile showed a benefit from EGDT. However, those with the highest biomarker quartile seem to be at risk of harm from EGDT.12 Kalil et al13 demonstrated that the baseline severity of patients with sepsis seems to affect its therapeutic effect. For example, severity was judged using the Simplified Acute Physiology Score II (SAPS II) and Acute Physiology and Chronic Health Evaluation II scores of patients who entered the ICU. The results illustrated that the benefits of EGDT are significantly reduced when these scores are high, suggesting that further study is needed to explore whether the efficacy of EGDT depends on baseline disease severity.
The current study aimed to analyse whether disease severity affects the association between fluid resuscitation using more than 30 mL/kg volume and adverse outcomes compared with that using less than 30 mL/kg volume.
Methods
Data source
We non-selectively and consecutively collected data for all the participants from the Medical Information Mart for Intensive Care (MIMIC-III),14 a large US-based clinical databases with more than 38 605 ICU patients admitted to the Beth Israel Deaconess Medical Center (Massachusetts) between 2001 and 2012. No informed consent was required because this study was a retrospective cohort study. The project was approved by the Institutional Review Boards of Beth Israel Deaconess Medical Center (Boston, Massachusetts) and the Massachusetts Institute of Technology (Cambridge, Massachusetts). Requirement for individual patient consent was waived because the project did not impact clinical care and all protected health information was deidentified. After online training at the National Institutes of Health (no. 28862843) was completed, access to the database was gained and data were documented. The Strengthening the Reporting of Observational Studies in Epidemiology statement was followed in the design of this study.15
Patient and public involvement
No patients or public were involved in the design or conduct of this study.
Data selection
The present single-centre, retrospective, longitudinal cohort study of septic patients was conducted according to the previously published Angus method.16 Details regarding the specific Angus diagnostic criteria can be found in the additional file 1: online supplemental document1. All the patients were aged ≥15 years, and only data about the initial ICU admission were analysed. Patients with incomplete data concerning the follow-up time and resuscitation volume records for the first 6 hours of the sepsis diagnosis were excluded. Patients who had been administered ≥30 mL/kg of intravenous fluid during the first 6 hours of the sepsis diagnosis for resuscitation were categorised as the standard group, while the remainder was included in the restrict group. After screening by the inclusion and exclusion criteria, 5154 participants were enrolled in the study (figure 1). Because the fluids managed by the emergency medical service before hospital arrival are not recorded in the MIMICIII database, this information was excluded. Only crystalloids administered as boluses were included in the calculation for the volume of resuscitation within the first 6 hours.
Supplemental material
Inclusion/exclusion criteria. MIMIC, Medical Information Mart for Intensive Care.
The specific crystalloids are classified in online supplemental table 1. The onset of our sepsis diagnosis is the point in time when a patient with organ failure in the ICU has a suspected infection and antibiotics are administered as the onset of sepsis. The chart of the timing of the onset of the sepsis diagnosis was added to the online supplemental figure 1.
The data on the following information were extracted: age on entering ICU; gender; weight; day 1 SAPS II score17; The highest SOFA score on the first 24 hours18; Elixhauser comorbidity score19; mechanical ventilation; inotropic and vasopressor agent administration; sedative administration within 24 hours of ICU admission and certain laboratory tests and comorbidities (online supplemental table 2).
Primary and secondary outcomes
Death within 28 days following ICU admission (28-day mortality) served as the primary study endpoint and days without needing mechanical ventilation or vasopressor administration within 28 days of ICU admission served as the secondary endpoint.
Missing data
The missing values in this study were treated as follows: missing variables >50% were converted into missing and non-missing levels to use the presence or absence of several deletions as independent variables. Central venous pressure (CVP) values and laboratory brain natriuretic peptide(BNP), troponin and creatine kinase tests were used as covariates and binary variables.17 Details regarding the missing values are found in online supplemental figure 2.
Statistical analysis
In this study, the presentation of continuous variables was divided according to their distribution status: (1) normal distribution (mean±SD) and (2) skewed distribution (medium (IQR). Non-normally distributed continuous variables were compared using the Wilcoxon signed-rank test. χ2 test or Fisher’s exact test were employed to compare the differences in the categorical variables.
To assess the impact of disease severity on the fluid resuscitation approach–day mortality relationship, we converted the SOFA scores to categorical variables: SOFA ≥10 and SOFA <10 (according to smooth curve fitting; figure 2). We also converted the SOFA scores to other categorical variables: ≤3, 4–6 and 7–22 (according to tertiles). Additionally, the SAPS II scores were converted to categorical variables: 5–18, 19–22 and 23–42 (according to tertiles).
Interaction between different fluid resuscitation approaches and SOFA score on 28-day mortality. A nonlinear association between SOFA score and 28-mortality was found in a generalised additive model (GAM). The solid red line represents the fitting spine in the restrict group. Blue bands represent the fitting spine in the standard group. All adjusted for age; weight; first care unit; ethnicity; Elixhauser score; Vasopressor use; CVP; weight loss; metastatic cancer; congestive heart failure. CVP, central venous pressure; SOFA, sequential organ failure assessment.
The HRs and 95% CIs for the risk of the 28-day mortality associated with different volumes of fluid resuscitation and each subgroup, as defined by the SOFA and SAPS II score levels, were estimated using Cox proportional hazards models with adjustment for pertinent variables and their interactions were tested. The covariates considered in the model included age, weight, first care unit, ethnicity, SOFA score, Elixhauser score, vasopressor use (first 24 hours), CVP (test), weight loss, metastatic cancer, congestive heart failure and site of infection. These covariates were chosen according to their relationship with the outcomes of interest or >10% effect estimate changes.19 20 The non-linearity of continuous confounding and outcome variables was evaluated via smooth curve fitting (penalised spline method).21 Online supplemental figures 3–6 show the smooth curve fitting of the SOFA score, SAPS II score, age and body weight with the outcomes of interest.
Additionally, to minimise the potential bias of treatment allocation and confounding, patients with similar baseline characteristics were identified using propensity score methods to eliminate the variability between patients in the two study groups. Associations between the different fluid resuscitation approaches and mortality were estimated by multivariable Cox regression models using propensity score matching.22 The primary analysis used matching in propensity score analyses. In the postmatched cohort, log-rank tests were used for survival analyses to assess whether different fluid resuscitation approaches affect the 28-day and 90-day mortality, and Kaplan-Meier curves were used for evaluation.
In the propensity score matching analysis, propensity score matching relies on conditional probabilities of a given exposure considering a particular group of baseline covariates.22 23 A non-parsimonious multivariable logistic regression model was used to estimate these scores.24 The different fluid resuscitation approaches were used as dependent variables, while all the baseline characteristics outlined in table 1 were used as covariates.25 Given that SAPS II and SOFA measured the same clinical characteristics, SOFA was employed for propensity score calculation using a greedy-matching algorithm (1:1 matching, no replacement), with a calliper width=0.01 of the SD of the propensity score logit. Prematching and postmatching and standardised difference values for the baseline covariates were assessed to gauge the postmatching balance.26 Standardised differences <0.1 for a given covariate revealed a negligible imbalance. The covariate balance is shown in online supplemental figure 7, which indicates the satisfactory balancing of these variables following matching.
Baseline characteristics before and after propensity score matching
We conducted sensitivity analyses in which the 90-day mortality was also the outcome. To further confirm that our findings are robust to potential confounders, we performed stratified analyses by subgroups defined by major covariables known to affect death risk, including age, gender, ventilation, renal failure, coagulopathy and hypertension. The R statistical package (http://www.r-project.org; The R Foundation; V.3.6.1) was employed for data analysis. A p value<0.05 was considered statistically significant.
Results
Study participants and baseline characteristics
After the admission data of 38 605, MIMIC III database adults were reviewed, 17 420 cases of hospitalised sepsis were identified using the Angus method. After the patients’ first ICU admissions were included, those with missing data for the follow-up time and fluid resuscitation volume for the first 6 hours, as well as those who died within 6 hours of entering the ICU, were excluded from the analysis. The final cohort comprised 5154 patients (figure 1). The histogram reveals the patient distribution based on the administered fluid volume (online supplemental figure 8).
Patients with respiratory and urinary infections received restrict fluid strategy, while gastrointestinal and other infections received a standard fluid strategy. The standard group had significantly higher severity scores on admission than the restrict group (SAPS II score: 24.12 (± 5.53) vs 20.93 (± 4.96); SOFA score: 8.13 (± 3.97) vs 5.9 (± 3.49)). The patients in the standard group received more mechanical ventilation (70.5%) and vasopressor treatment (63.4%) than those in the restrict group (55.3% and 34.2%, respectively) during the first 24 hours of their ICU stay. Over a follow-up of 28 days, 2096 patients (40.7%) had a primary end point outcome, 1709 (39.05%) from the restrict group and 387 (49.81%) from the standard group.
Effect modification of the disease severity on the association between mortality and different early fluid resuscitation approaches
In this study, the non-linear relationship between the SOFA score and 28-day mortality in different fluid resuscitation approaches was analysed (figure 2). Smooth curve fitting and the generalised additive model showed the conspicuous interaction of different fluid resuscitation approaches in the relationship between the SOFA score and 28-day mortality after adjusting for confounders. The two lines intersect at the SOFA score of 9–10, indicating that patients with a SOFA score ≥10 did not benefit from fluid resuscitation with the guideline-recommended volume (≥30 mL/kg).
Table 2 further quantifies the effect modification of disease severity on the association between the 28-day mortality and different fluid resuscitation approaches in the prematched cohort. A total of 2096 patients died within 28 days after ICU admission, 1709 (39.05%) from the restrict group and 387 (49.81%) from the standard group. After the adjustment for age, weight, first care unit, ethnicity, SOFA score, Elixhauser score, vasopressor use, CVP (test), weight loss, metastatic cancer and congestive heart failure, the unequal fluid resuscitation approach was not independently related to the 28-day mortality (adjusted OR, 1.12; 95% CI 0.99 to 1.27; p=0.07). The standard group with the high SOFA score (≥10) demonstrated a relative risk of 28-day mortality of 38% in the adjusted model (HR, 1.38; 95% CI 1.13 to 1.68; p=0.001) compared with the restrict group. By contrast, the risk reduction in the low SOFA score (<10) subgroup was not significant (HR, 0.95; 95% CI 0.81 to 1.13; p=0.57). The effect of the interaction between the SOFA score and different fluid resuscitation approaches on the 28-day mortality was significant (p=0.003). Standard treatment increased the risk of 28-day mortality by 33% (HR, 1.33; 95% CI 1.13 to 1.55; p=0.0004) among patients with SAPS II Tertile 3 compared with the restrict treatment. The effect of the interaction between the SAPS II (Tertile) subgroup and different volumes in fluid resuscitation on the 28-day mortality was significant (p=0.0008).
Modification of high disease severity on the association between early fluid resuscitation and 28-day mortality in the prematched cohort
Similar analyses were also conducted for the postmatched cohort, as shown in table 3. In propensity score analyses with matching, the HR was 1.32 (95% CI 1.03 to 1.70; p=0.03) in the high SOFA score (≥10) subgroup of the standard group compared with that in the restrict group. By contrast, the risk reduction in the low SOFA (<10) subgroup was modest (HR, 0.85; 0.70 to 1.03; p=0.10). The effect of the interaction between the SOFA score and different fluid resuscitation approaches on the 28-day mortality was significant (p=0.0035).
Modification of high disease severity on the association between early fluid resuscitation and 28-day mortality in the postmatched cohort
Kaplan-Meier curve analysis (figure 3) illustrated comparable 28-day and 90-day mortalities between groups in the total propensity score-matched cohort and high SOFA score (≥10) subgroup. The 28-day and 90-day mortalities were not significantly different (p=0.75 and p=0.43 based on log-rank tests, respectively) in the whole propensity score-matched cohorts. However, restricted fluid resuscitation was related to reduced 28-day and 90-day mortalities (p=0.012 and p=0.015 based on log-rank tests, respectively) in the high SOFA score (≥10) subgroup.
Kaplan-Meier survival curves. (A) Kaplan-Meier survival curves showing that 28-day mortality was not significantly different between the two groups in the whole propensity score-matched cohort. (B) Kaplan-Meier survival curves showing that 90-day mortality was not significantly different between the two groups in the whole propensity score-matched cohort. (C) Kaplan-Meier survival curves showing that 28-day mortality was significantly different between the two groups in the subgroup of high SOFA score (more than 10) of the propensity score-matched cohort. (D) Kaplan-Meier survival curves showing that 90-day mortality was significantly different between the two groups in the subgroup of high SOFA score (more than 10) of the propensity score-matched cohort. SOFA, sequential organ failure assessment.
We also analysed the interactions among the disease severity, different groups and 90-day mortality. Regardless of the prematched and postmatched cohorts, the conspicuous modification relationship between different groups and mortality by the SOFA score was found (online supplemental tables 3, 4 of additional file 1).
Secondary outcome study with propensity score matching
Several secondary outcomes were studied to investigate the potential factors for the adverse effects of standard fluid resuscitation in patients with high disease severity. Multiple secondary outcome differences were detected between groups. First, the standard group exhibited significantly longer vasopressor utilisation than the restrict group (days without vasopressors within 28 days=7 vs 9; p=0.014). Second, mechanical ventilation duration was significantly prolonged in the standard group relative to the restrict group (days without ventilation within 28 days=6 vs 7; p=0.019). The use of dobutamine was comparable between the groups (table 4).
Secondary outcome studies with propensity score matching
Sensitivity analysis
To improve the stability of our results, all subgroups analyses were revealed in online supplemental table 5 (see additional file 1). As shown in figure 4, stratified subgroup analyses were conducted based on variables including age, gender, ventilation, renal failure, coagulopathy and hypertension. Figure 4 indicates that the standard resuscitation volume may reduce the risk of 28-day mortality among the low SOFA score (<10) patients irrespective of subgroup, with HRs ranging from 0.59 to 0.94. By contrast, the efficacy of the standard resuscitation volume was considerably attenuated in patients with a high SOFA score (≥10), with HRs ranging from 1.00 to 2.10, a finding that is consistent with previous outcomes. In order to understand whether the harmful effects of ‘standard’ treatment are due to those patients receiving the highest amount of fluid. We, therefore, reanalysed the population of those receiving less than 50ml/kg of fluid. The results remained stable regardless of whether the outcome was 28-day mortality in the online supplemental table 6. For Surgical-ICU patients admitted to the ICU from the operating room, we reperformed the cox regression model for the effect of different fluid management on 28-day mortality. As the regression results in the online supplemental table 7, the results remained stable in SICU.
Subgroup analyses of the association between different fluid resuscitation and 28-day mortality. The multivariate model adjusted for age, weight, first care unit, ethnicity, Elixhauser score, vasopressor use; CVP, weight loss, metastatic cancer and congestive heart failure. None of the stratified variables, including sex, age, weight or ethnicity, significantly modified the effect on the association between fluid resuscitation methods and 28-day mortality; the p values of all interaction tests were >0.1. CVP, central venous pressure; SOFA, sequential organ failure assessment.
Discussion
In this single-centre retrospective cohort study, the difference in the 28-day mortality between patients who received fluid resuscitation per SSC guidelines and those who received fluid resuscitation with lower than the recommended volume among septic patients was not detected. Remarkable differences in the association between fluid resuscitation approaches and 28-day mortality were observed among the subgroups, as defined by the SOFA score. An increase in the 28-day death risk of 32% for patients who received the standard volume of fluid resuscitation was observed in the high SOFA (≥10) subgroup (HR, 1.32; 95% CI 1.03 to 1.70; p=0.03). By contrast, standard fluid resuscitation treatment did not significantly affect the 28-day mortality of the low SOFA (<10) subgroup. The possible cause was that after patients with a high SOFA score (≥ 10) were excluded, those that remained provided a sample with insufficient statistical power to resolve differences in the relationship between mortality and different resuscitation volumes (type II error). However, clear trends remained visible despite the loss of statistical significance.
Being aware of the heavy burden of sepsis, the Medical Insurance and Medicaid Service Center implemented the Severe Sepsis and Sepsis Shock Early Management Kit (SEP-1) in 2015.12 27–29 At the same time, people are increasingly concerned about the possibility of unintended consequences of sepsis authorisation and the culture of ‘treating before asking questions’ that they encourage. Some are concerned that some aspects of the SEP-1 reporting requirement mandates may paradoxically harm some patients by pressuring clinicians to provide aggressive, rapid, rigid and reflexive care that is not suitable for all patients.30–34
EGDT is a protocolised approach to treat sepsis. EGDT is an ‘all-or-nothing’ measure that obliges clinicians to decide to administer 30 cc/kg of crystals to patients with sepsis in early fluid resuscitation. (1) With the development of medical science, a growing body of large clinical trials has been conducted to test whether EGDT fluid strategy and larger volumes of intravenous fluid could reduce mortality but have yielded conflicting results.30 31 33 35 The high heterogeneity of sepsis may explain the different test results to differ, such as the severity of the disease in each patient.
Kalil et al13 suggested that the baseline severity of patients with sepsis seems to affect the therapeutic effect of EGDT. For example, the severity was judged by the SOFA and SAPS II scores of patients who entered the ICU; among 20 000 participants, the benefit of EGDT was significantly decreased in those with high scores. Their conclusions were consistent with the findings of the present study. The relationship between the baseline severity of patients with sepsis and the efficacy of EGDT-based fluid replacement therapy may be real. This type of treatment based on varying degrees of disease severity could be attributed to the heterogeneity of the treatment effects. Heterogeneity indicates that, under different disease severities, the same therapy may have two different outcomes (ie, potentially harmful in those with severe disease and helpful in those with less severe disease).36
The effect of the qualitative interaction between the fluid resuscitation approach and SOFA score on the 28-day mortality may be caused by the following factors: (1) similar to the U-shape effect of the association between the 90-day all-cause mortality and platelet-to-lymphocyte ratio, the fixed bundle protocol approach may worsen the condition of the patients with high disease severity although the fluid resuscitation recommended by the SSC guidelines may be beneficial for those with low disease severity36; (2) EGDT possibly comprises harmful and beneficial effects. Disease severity determines the direction of the effect. For example, there are benefits to increase fluid intake, while using inotropes to target central venous oxygen saturation (ScvO2) is harmful. The proportion of harmful or beneficial effects applied to each patient differs. Additionally, the number of harmful effects each patient could withstand varies. Either of these two hypotheses may explain the finding that patients with the more severe disease become worse under standard fluid resuscitation.36
Fluid resuscitation has long been a cornerstone of the treatment of infectious shock, despite the lack of an evidence-based basis for its prognostic impact. A growing body of literature suggests that early and rapid fluid rehydration may be harmful. Potential mechanisms of harm include cardiovascular failure associated with vasodilation, cardiotoxicity, impaired endothelial glycocalyx and inflammatory response. In addition, fluid overload due to fluid infusion may also contribute to harm.37
Once tissue hypoperfusion due to sepsis is identified, rapid rehydration therapy is recommended and resuscitation goals should include CVP, blood pressure, urine output and ScvO2. Improvements in these surrogate indicators are thought to predict improved tissue perfusion and an improved prognosis. However, although early signs of cardiovascular improvement can occur after rapid rehydration therapy, they will lead to worsening cardiovascular function and poor prognosis.38
An RCT is currently underway evaluating restrictive and liberal fluid protocols during active resuscitation (Crystalloid Liberal or Vasopressors Early Resuscitation in Sepsis (CLOVERS); NCT03434028) that may better inform discussions and guide clinical practice.39
The strengths of the present study included that we used clinical data to discuss the impact of fluid resuscitation based on EGDT mandatory guidance on the outcome of sepsis patients, according to a high-quality database and rigorous validation of all the data. Non-linearity was also explored and further addressed. Multivariate Cox proportional hazard models and Propensity Score Matching(PSM) analysis were applied to ensure robust results. We also conducted a series of sensitivity analyses to ensure the stability of the results.
The limitations of this study should also be acknowledged. First, one key limitation of this study is its retrospective nature despite the adjustments made for age, weight, first care unit, ethnicity, SOFA score, Elixhauser score, vasopressor use (first 24 hours), CVP (test), weight loss, metastatic cancer, congestive heart failure and the biases that were not fully accounted for. Data not documented in the medical record and biases could not be fully accounted for in the analyses. The results of this study are descriptive, and we cannot determine causality through this research design. It is also difficult to determine why patients in the restrict group did not receive the recommended amount of fluid resuscitation—whether their underlying comorbidity puts them at risk of fluid overload, whether the patient appears to be clinically mild or whether there are other plausible reasons for diagnosis. However, with the analytic approaches used in the examination of our observational cohort, we have tried to minimise possible confounding in various ways. We had not collected and adjusted the amount of fluid before the use of vasopressors, which may affect our conclusions. Notably, if the amount of fluid given was adjusted prior to the introduction of the vascular compressor, then the potential generated by standard fluid resuscitation would be biased towards ineffectiveness, leading to an underestimation of the association between standard fluid resuscitation and 28-day mortality.
Second, other limitations of our research include the lack of data on certain variables and potential inaccuracies in the MIMIC III database health records. Third, these findings were based on the data from a single quaternary academic medical centre, potentially constraining our ability to generalise these results to other sepsis patients in small, community-based hospital ICUs. Fourth, another limitation involved the dates of patient recruitment for this study. While the cohort data were relatively old (2002–2012), the fluid volume received by the patients during the early fluid resuscitation phase was very similar to that used in the most recent studies. Fifth, the lack of analysis of the specific time and volume of fluid resuscitation make it impossible to support the causal inference. The results of this study are descriptive and we cannot establish causality with such a study design. It was hard to determine why patients in the restrict group did not receive the recommended fluid resuscitation—whether their underlying complications put them at risk of fluid overload, or whether the patients seem to be less sick or had other reasonable diagnoses. Sixth, an important limitation is that only crystalloid fluids were considered when measuring the volume of fluids. Seventh, the SOFA score was evaluated on the first day of admission, so the therapeutic effect might modify this score. So we can not further discuss the causal relationship between SOFA score and early fluid resuscitation. The lasted limitation that is difficult to address with this data set is that Surviving Sepsis guidelines generally recommended the use of the near full EGDT bundle created by Rivers et al for the entire enrolment period of this database, thus many practices that are no longer standard of care were included in the care of the patients evaluated in the cohort. Our study was conducted based on the original recommendations of the EGDT study, did not plan the study based on the most recent sepsis bundle and does not reflect current guideline-based care. MIMIC-III database time point is 2001–2012, a temporal limitation of our chosen population; because of the recent release of MIMIC-IV, we will next explore the impact of new guidelines on early fluid resuscitation on sepsis prognosis through new populations.
Conclusion
In septic patients with a SOFA score greater than 10, compared with receiving restrict fluid therapy, receiving standard fluid therapy is associated with great 28-day mortality. SOFA evaluated on the first day of admission may modify the association between the volume of fluid resuscitation and mortality in patients with sepsis. Due to the nature of such retrospective studies, these results should be interpreted with caution. However, these data provide supportive evidence for the efficacy of different volumes in the early fluid resuscitation treatment of sepsis. Future RCTs prospectively investigating explore the efficacy of resuscitation on septic patients are warranted.
Data availability statement
No data are available.
Ethics statements
Patient consent for publication
Ethics approval
Not applicable.
References
Supplementary materials
Supplementary Data
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Footnotes
Contributors RZ designed the study, LL collected and analysed data, and XJ and WL contributed to writing this manuscript. All authors have read and approved the final manuscript.RZ is the initials of the author acting as guarantor.
Funding The authors have not declared a specific grant for this research from any funding agency in the public, commercial or not-for-profit sectors.
Competing interests None declared.
Patient and public involvement Patients and/or the public were not involved in the design, or conduct, or reporting, or dissemination plans of this research.
Provenance and peer review Not commissioned; externally peer reviewed.
Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.