Characteristics and outcomes of culture-negative versus culture-positive with fungus in sepsis patients: a retrospective analysis of the MIMIC-III database

BACKGROUND: We compared the characteristics of culture-positive and culture-negative with fungi in septic patients to determine whether fungi culture status is associated with mortality and the relationship between antifungal therapy and sepsis patient mortality. METHODS: The study was based on the Medical Information Mart for Intensive Care (MIMIC) III database, we included all intensive care unit (ICU) admissions between 2001 and 2012 with sepsis, which met the Martin’s criteria. The primary outcome was hospital mortality. Secondary outcomes included the usage of antifungal drugs, duration of mechanical ventilation and hospital stay. Multivariable logistic regression and propensity score matching were used to investigate any association. RESULTS: The study population included 836 fungi-positive patients (16.6%) and 4191 fungi-negative patients (83.4%). Fungi-positive patients had more congestive heart failure and chronic pulmonary, higher sequential organ failure assessment (SOFA), and more need for renal replacement therapy on day one than fungi-negative patients. There was no correlation between antifungal therapy and hospital mortality (adjusted odds ratio = 1.03, 95% CI [0.89, 1.20]; P=0.676 (cid:0) . Hospital mortality was lower in the fungi-negative group (25.5%) than in the fungi-positive group (37.3%, P<0.001). After propensity score matching, 613 cases from each group were matched. The hospital mortality remained signicantly higher in the fungi-positive group (167/613 vs. 216/613, p=0.003). CONCLUSIONS: Although residual confounding cannot be excluded, signicant differences between fungi-positive and fungi-negative sepsis are identied, with the former group having more comorbidities, worse severity of illness, longer hospitalizations, and higher mortality. Antifungal therapy does not affect the outcome.

Hence, our study aimed to compare the characteristics and outcomes of fungi culture-positive versus fungi culture-negative sepsis and analyzed the effect of antifungal therapy on mortality.

Database introduction
We extracted the data from an online international database-Medical Information Mart for Intensive Care III (MIMIC III), with approval from the review boards of the Massachusetts Institute of Technology and Beth Israel Deaconess Medical Center. We did not need informed consent because all the patients in the database were de-identi ed for privacy protection. One author (Zhiye Zou) obtained access to this database (certi cation number 35951237) and was responsible for data extraction.

Study design Inclusion and exclusion criteria
We included all patients who met the criteria for Martin sepsis, consisting of codes that imply a disseminated bloodstream infection (septicemia, bacteremia, and fungemia) (11,12). Patients who were younger than 18 years in the ICU were excluded. For patients admitted to the ICU more than once, only the rst ICU stay was considered.

Data collection
Data collected were baseline variables on entry to the ICU including patient demographics, source of admission, comorbidities, vital signs and blood investigations (white blood cell count, and Neutrophils where available), and variables on the rst day of ICU admission including the Sequential Organ Failure Assessment (SOFA) score. We de ned organ failures as a SOFA score of >2 for the organs concerned (13).
To ensure that any fungi and bacteria isolated were associated with sepsis that resulted in ICU admission, we recorded results of all fungi and bacteria cultures collected within the two days before and the two days after ICU admission. The use of antifungal drugs, including azole, echinocandins, nystatin, and amphotericin, was also extracted.
The primary outcome variable was hospital mortality, while the secondary outcome variables were duration of mechanical ventilation, hospital stays, and the effect of antifungal agents on mortality.

Statistical analyses
Continuous variables were expressed as the mean± standard deviation or median (interquartile range) as appropriate. The Student's t-test and Wilcoxon rank-sum test were used as appropriate. Categorical data were expressed as proportions and compared using the chi-square test. The included patients were divided into two subgroups according to whether the fungi were positive or negative. Hierarchical chisquare analysis was used to test for homogeneity between the two subgroups. Multivariable logistic regression was used for covariate adjustment. The logistic models were built using the stepwise backward method.
PSM(Propensity score matching) (14) was used to minimize confounding factors such as Comorbid conditions and disease severity, which may lead to outcome bias. A one-to-one nearest neighbor matching algorithm was applied using a caliper width of 0.05. The following variables were selected to generate the propensity score: age, male, hypertension, congestive heart failure, chronic pulmonary, liver disease, chronic kidney disease, and SOFA score on ICU admission. Kernel density plots of the p score were used to examine the PSM degree. Finally, 613 matched pairs were generated and applied to further analyses.

Results baseline characteristics
The study population included 836 fungi-positive patients (16.6%) and 4191 fungi-negative patients (83.4%). Table 1 describes their characteristics at baseline and on day one of the ICU stay. Fungi-positive patients had more congestive heart failure and chronic pulmonary, less hypertension and chronic kidney disease, higher white blood cell and neutrophils, higher sequential organ failure assessment, more respiratory, Cardiovascular, renal and hepatic failure, and more need for renal replacement therapy on day one than fungi-negative patients. In addition, antifungal therapy was used more frequently in patients with positive fungal culture (48.0% vs. 25.2%, P<0.001).
As shown in Table 2, More than half of the fungi 487 (58.3%) were cultured in sputum, and the mortality rate was 38.2%. Bronchoalveolar lavage uid, urine, and blood also had many fungi, while other sources of the fungi are less. Table 3 details the variables associated with Positive Fungi. Multivariable analysis revealed the following independent predictors of positive fungi: age, sex, chronic pulmonary, congestive heart failure, neutrophils, SOFA score, respiratory failure. Table 4 details the variables associated with hospital mortality. Fungi positivity was associated with higher mortality both in univariate and multivariate analyses. Fungi in sputum (including other yeast fungi, aspergillus spp. and candida spp.), bronchoalveolar lavage, and urine were also associated with higher mortality. Multivariable analysis revealed the following independent predictors of mortality: age, maximum respiratory rate, liver disease, SOFA score, organ failure. In addition, there was no correlation between antifungal therapy and hospital mortality (adjusted odds ratio = 1.03, 95% CI [0.89, 1.20]; P=0.676). Fungus positive in the lungs, with or without antifungal therapy, is a risk factor for death.
Fungus negative in the lungs, the use of antifungal therapy had no effect on hospital mortality, while the absence of antifungal therapy was a protective factor for death (adjusted odds ratio = 0.79, 95% CI [0.68, 0.90]; P=0.001). The same results appear in the urine. Fungus negative in the blood is a protective factor for death.
Patient outcomes are presented in Table 5. Fungi-positive patients had a longer duration of mechanical ventilation, longer duration of hospital stay (14.20days (7.80 to 24.56) versus 10.42 (5.70 to 20.25), P<0.001) and higher 28-day mortality. Hospital mortality was lower in the fungi-negative group (25.5%) than in the fungi-positive group (37.3%, P<0.001).

Figure1
In both bacteria-negative and bacteria-positive subgroups, the mortality rate of fungi-positive patients was signi cantly higher than that of fungi-negative patients.

Discussion
To the best of our knowledge, no previous study has focused on the differences between fungi-positive and fungi-negative sepsis. This study's main ndings are that patients with fungi-positive sepsis had more comorbidities, more organ failure, a longer duration of mechanical ventilation, a longer length of hospital stay, and higher hospital mortality. Antifungal therapy did not affect the outcome. After multivariate analysis and PSM, the results remained unchanged.
Fungal culture-positive sepsis has increased signi cantly (15,16). A study of the epidemiology of sepsis in the United States (U.S.) found that the annual number of sepsis cases caused by fungal organisms increased by 207% between 1979 and 2000 (11). Candida spp. positive occurs in up to 80% of critically ill patients after one week in intensive care (17). The Extended Prevalence of Infection in Intensive Care (EPIC II) study found that Candida spp. were the second most frequent cause of infection (18.2% of all infections) in North American intensive care units (ICUs) (18). In our retrospective study, 16.7% of patients with sepsis were fungal culture-positive. In short, fungal infection is still a big challenge (19).
Compared with fungal culture-negative in sepsis patients, positive patients had a worse outcome. After ICU admission, 1-year mortality was signi cantly higher in fungal culture-positive patients with sepsis with an APACHE II score less than 25 than in those with fungal culture-negative patients (66.7% versus 50.0%) (20). Some studies even gave a 30-day mortality of up to 60% for fungal positive patients, while mortality in septic shock was almost 90% in a retrospective case series from the USA (21). A lot of studies have shown that patients with positive fungus have more complications and long hospital stays, so we should pay attention to them(4).
However, antifungal treatments do not always work. (22) In a multicentre, randomized, double-blind clinical trial, empirical antifungal therapy in sepsis patients with nonneutropenic did not improve 28-day mortality.(23) Antifungal therapy also did not signi cantly improve prognosis in burn patients. (9) Therefore, in the early stage of sepsis, antifungal agents had no signi cant effect on death regardless of the patients' fungal culture status.
This study has several limitations. First, sepsis patients are divided into two groups according to fungi culture status. In reality, both groups are a mixed bag of diagnoses. Culture-negative patients include many non-fungal sepsis or even non-septic patients. Second, bacteria are the main pathogenic microorganisms of sepsis, so both fungal positive and negative patients contain a large number of positive bacterial patients, so it is likely to affect the outcome. Third, the fungi species had not been subdivided, most of which were yeast, which affected further analysis. Fourth, there is no speci c time for medication. The use of antifungal agents in this article is indicative of the presence of these agents throughout the treatment process. These can also affect the e cacy of antifungal drugs and our outcomes.

Conclusions
In conclusion, by the analysis of a large clinical database, our study shows that signi cant differences between fungi-positive and fungi-negative sepsis, with the former group having more comorbidities, worse severity of illness, longer hospitalizations, and higher mortality. Antifungal therapy does not affect hospital mortality. However, more research on this topic needs to be undertaken before the association between antifungal therapy and mortality is more clearly understood.

Declarations
Ethics approval and consent to participate: We extracted the data from an online international database-Medical Information Mart for Intensive Care III (MIMIC III), with approval from the review boards of the Massachusetts Institute of Technology and Beth Israel Deaconess Medical Center. We did not need informed consent because all the patients in the database were de-identi ed for privacy protection.

Consent for publication:
Written informed consent for publication was obtained from all participants.
Availability of data and materials: One author (Zhiye Zou) obtained access to this database (certi cation number 35951237) and was responsible for data extraction.
Author Contributions: