Predicting invasive fungal disease due to Candida species in non-neutropenic, critically ill, adult patients in United Kingdom critical care units

Candida albicans-A systematic review to inform the World Health Organization Fungal Priority Pathogens List

Candida albicans is a common fungal pathogen and amongst the leading causes of invasive candidiasis globally. This systematic review examines the characteristics and global impact of invasive infections caused by C. albicans. We searched on PubMed and Web of Science for studies reporting on criteria such as mortality, morbidity, drug resistance, preventability, yearly incidence, and distribution/emergence during the period from 2016 to 2021. Our findings indicate that C. albicans is the most common Candida species causing invasive disease and that standard infection control measures are the primary means of prevention. However, we found high rates of mortality associated with infections caused by C. albicans. Furthermore, there is a lack of data on complications and sequelae. Resistance to commonly used antifungals remains rare. Although, whilst generally susceptible to azoles, we found some evidence of increasing resistance, particularly in middle-income settings-notably, data from low-income settings were limited. Candida albicans remains susceptible to echinocandins, amphotericin B, and flucytosine. We observed evidence of a decreasing proportion of infections caused by C. albicans relative to other Candida species, although detailed epidemiological studies are needed to confirm this trend. More robust data on attributable mortality, complications, and sequelae are needed to understand the full extent of the impact of invasive C. albicans infections.

Identification and assessment of antifungal susceptibility of Candida species based on bronchoalveolar lavage in immunocompromised and critically ill patients

Background and Objectives

The presence of fungi in the respiratory tract as mycobiome, particularly Candida species (spp.), remains a serious problem due to increasing numbers of immunocompromised patients. The confirmed reliable existence of these pathogens due to frequent colonization is essential. This investigation aimed to recognize Candida spp. among isolates from bronchoalveolar lavage of immunocompromised and critically ill patients and to evaluate their susceptibility to antimycotic drugs.

Materials and Methods

Bronchoalveolar lavage fluid was collected from 161 hospitalized patients presenting with suspected respiratory fungal infection /colonization. The specimens were examined by standard molecular and mycological assays. Candida spp. were recognized with sequence assessment of the D1-D2 section of the large subunit ribosomal DNA. The susceptibility of Candida isolates to common antimycotic drugs was distinguished by standard broth microdilution.

Results

Seventy-one clinical isolates of Candida spp. were recognized. Candida albicans was the most frequent, followed by C. glabrata, C. krusei (Pichia kudriavzevii), C. dubliniensis, C. parapsilosis, and C. tropicalis. We found 5.1% of C. albicans isolates and 8% of C. glabrata isolates to show resistance to fluconazole. The whole of the Candida spp. were sensitive to amphotericin B and caspofungin.

Conclusion

This study demonstrated that C. albicans and C. glabrata are the most common isolates of bronchoalveolar lavage fluid in patients, and the drug susceptibility screening confirmed that amphotericin B and caspofungin are effective against Candida spp. but some C. glabrata and C. albicans isolates showed resistance to fluconazole.

Machine Learning in Infectious Disease for Risk Factor Identification and Hypothesis Generation: Proof of Concept Using Invasive Candidiasis

Background

Machine learning (ML) models can handle large datasets without assuming underlying relationships and can be useful for evaluating disease characteristics; yet, they are more commonly used for predicting individual disease risk rather than identifying factors at the population level. We offer a proof of concept applying random forest (RF) algorithms to Candida-positive hospital encounters in an electronic health record database of patients in the U.S.

Methods

Candida-positive encounters were extracted from the Cerner HealthFacts database; invasive infections were laboratory positive sterile site Candida infections. Features included demographics, admission source, care setting, physician specialty, diagnostic and procedure codes, and medications received prior to the first positive Candida culture. We used RF to assess risk factors for three outcomes: any invasive candidiasis (IC) vs non-IC, within-species IC vs non-IC (e.g. invasive C. glabrata vs non-invasive C. glabrata), and between-species IC (e.g. invasive C. glabrata vs all other IC).

Results

14 of 169 (8%) variables were consistently identified as important features in the ML models. When evaluating within-species IC, for example invasive C. glabrata vs non-invasive C. glabrata, we identified known features like central venous catheters, ICU stay, and gastrointestinal operations. In contrast, important variables for invasive C. glabrata vs all other IC included renal disease and medications like diabetes therapeutics, cholesterol medications, and antiarrhythmics.

Conclusions

Known and novel risk factors for IC were identified using ML, demonstrating the hypotheses generating utility of this approach for infectious disease conditions about which less is known, specifically at the species-level or for rarer diseases.

A multisite evaluation of antifungal use in critical care: implications for antifungal stewardship

Background

ICUs are settings of high antifungal consumption. There are few data on prescribing practices in ICUs to guide antifungal stewardship implementation in this setting.

Methods

An antifungal therapy (AFT) service evaluation (15 May-19 November 2019) across ICUs at three London hospitals, evaluating consumption, prescribing rationale, post-prescription review, de-escalation and final invasive fungal infection (IFI) diagnostic classification.

Results

Overall, 6.4% of ICU admissions (305/4781) received AFT, accounting for 11.41 days of therapy/100 occupied bed days (DOT/100 OBD). The dominant prescribing mode was empirical (41% of consumption), followed by targeted (22%), prophylaxis (18%), pre-emptive (12%) and non-invasive (7%). Echinocandins were the most commonly prescribed drug class (4.59 DOT/100 OBD). In total, 217 patients received AFT for suspected or confirmed IFI; 12%, 10% and 23% were classified as possible, probable or proven IFI, respectively. Hence, in 55%, IFI was unlikely. Proven IFI (n = 50) was mostly invasive candidiasis (92%), of which 48% had been initiated on AFT empirically before yeast identification. Where on-site (1 → 3)-β-d-glucan (BDG) testing was available (1 day turnaround), in those with suspected but unproven invasive candidiasis, median (IQR) AFT duration was 10 (7-15) days with a positive BDG (≥80 pg/mL) versus 8 (5-9) days with a negative BDG (<80 pg/mL). Post-prescription review occurred in 79% of prescribing episodes (median time to review 1 [0-3] day). Where suspected IFI was not confirmed, 38% episodes were stopped and 4% de-escalated within 5 days.

Conclusions

Achieving a better balance between promptly treating IFI patients and avoiding inappropriate antifungal prescribing in the ICU requires timely post-prescription review by specialist multidisciplinary teams and improved, evidence-based-risk prescribing strategies incorporating rapid diagnostics to guide AFT start and stop decisions.

State of the Art of Machine Learning-Enabled Clinical Decision Support in Intensive Care Units: Literature Review

Background

Modern clinical care in intensive care units is full of rich data, and machine learning has great potential to support clinical decision-making. The development of intelligent machine learning–based clinical decision support systems is facing great opportunities and challenges. Clinical decision support systems may directly help clinicians accurately diagnose, predict outcomes, identify risk events, or decide treatments at the point of care.

Objective

We aimed to review the research and application of machine learning–enabled clinical decision support studies in intensive care units to help clinicians, researchers, developers, and policy makers better understand the advantages and limitations of machine learning–supported diagnosis, outcome prediction, risk event identification, and intensive care unit point-of-care recommendations.

Methods

We searched papers published in the PubMed database between January 1980 and October 2020. We defined selection criteria to identify papers that focused on machine learning–enabled clinical decision support studies in intensive care units and reviewed the following aspects: research topics, study cohorts, machine learning models, analysis variables, and evaluation metrics.

Results

A total of 643 papers were collected, and using our selection criteria, 97 studies were found. Studies were categorized into 4 topics—monitoring, detection, and diagnosis (13/97, 13.4%), early identification of clinical events (32/97, 33.0%), outcome prediction and prognosis assessment (46/97, 47.6%), and treatment decision (6/97, 6.2%). Of the 97 papers, 82 (84.5%) studies used data from adult patients, 9 (9.3%) studies used data from pediatric patients, and 6 (6.2%) studies used data from neonates. We found that 65 (67.0%) studies used data from a single center, and 32 (33.0%) studies used a multicenter data set; 88 (90.7%) studies used supervised learning, 3 (3.1%) studies used unsupervised learning, and 6 (6.2%) studies used reinforcement learning. Clinical variable categories, starting with the most frequently used, were demographic (n=74), laboratory values (n=59), vital signs (n=55), scores (n=48), ventilation parameters (n=43), comorbidities (n=27), medications (n=18), outcome (n=14), fluid balance (n=13), nonmedicine therapy (n=10), symptoms (n=7), and medical history (n=4). The most frequently adopted evaluation metrics for clinical data modeling studies included area under the receiver operating characteristic curve (n=61), sensitivity (n=51), specificity (n=41), accuracy (n=29), and positive predictive value (n=23).

Conclusions

Early identification of clinical and outcome prediction and prognosis assessment contributed to approximately 80% of studies included in this review. Using new algorithms to solve intensive care unit clinical problems by developing reinforcement learning, active learning, and time-series analysis methods for clinical decision support will be greater development prospects in the future.

Using Machine Learning Algorithms to Predict Candidaemia in ICU Patients With New-Onset Systemic Inflammatory Response Syndrome

Background: Distinguishing ICU patients with candidaemia can help with the precise prescription of antifungal drugs to create personalized guidelines. Previous prediction models of candidaemia have primarily used traditional logistic models and had some limitations. In this study, we developed a machine learning algorithm trained to predict candidaemia in patients with new-onset systemic inflammatory response syndrome (SIRS).

Methods: This retrospective, observational study used clinical information collected between January 2013 and December 2017 from three hospitals. The ICU patient data were used to train 4 machine learning algorithms–XGBoost, Support Vector Machine (SVM), Random Forest (RF), ExtraTrees (ET)–and a logistic regression (LR) model to predict patients with candidaemia.

Results: Of the 8,002 cases of new-onset SIRS (in 7,932 patients) included in the analysis, 137 new-onset SIRS cases (in 137 patients) were blood culture positive for candidaemia. Risk factors, such as fungal colonization, diabetes, acute kidney injury, the total number of parenteral nutrition days and renal replacement therapy, were important predictors of candidaemia. The XGBoost machine learning model outperformed the other models in distinguishing patients with candidaemia [XGBoost vs. SVM vs. RF vs. ET vs. LR; area under the curve (AUC): 0.92 vs. 0.86 vs. 0.91 vs. 0.90 vs. 0.52, respectively]. The XGBoost model had a sensitivity of 84%, specificity of 89% and negative predictive value of 99.6% at the best cut-off value.

Conclusions: Machine learning algorithms can potentially predict candidaemia in the ICU and have better efficiency than previous models. These prediction models can be used to guide antifungal treatment for ICU patients when SIRS occurs.

Clinical predictive models of invasive Candida infection: a systematic literature review

Clinical predictive models (CPM) serve to identify and categorize patients into risk categories to assist in treatment and intervention recommendations. Predictive accuracy and practicality of models varies depending on methods used for their development, and should be evaluated.The aim of this study was to summarize currently available CPM for invasive candidiasis, analyze their performance, and assess their suitability for use in clinical decision making.We identified studies that described the construction of a CPM for invasive candidiasis from PubMed/MEDLINE, EMBASE, SCOPUS, Web of Science, Cochrane Library databases and Clinicaltrials.gov. Data extracted included: author, data source, study design, recruitment period, characteristics of study population, outcome types, predictor types, number of study participants and outcome events, modelling method and list of predictors used in the final model. Calibration and discrimination in the derivative datasets were used to assess the performance of each model.Ten articles were identified in our search and included for full text review. Five models were developed using data from ICUs, and five models included all hospitalized patients. The findings of this review highlight the limitations of currently available models to predict invasive candidiasis, including lack of generalizability, difficulty in everyday clinical use, and overly optimistic performance.There are significant concerns regarding predictive performance and usability in every day practice of existing CPM to predict invasive candidiasis.

LAY SUMMARY

Clinical predictive models may assist in early identification of patients at risk for invasive candidiasis to initiate appropriate treatment. The findings of this systematic review highlight the limitations of currently available models to predict invasive candidiasis.

Epidemiology and Mycology of Candidaemia in non-oncological medical intensive care unit patients in a tertiary center in the United States: Overall analysis and comparison between non-COVID-19 and COVID-19 cases

The epidemiology and mycology of invasive candidiasis in the ICU is well‐described in certain types of critically ill patients but not in others. One population that has been scarcely studied is non‐neutropenic patients admitted specifically to medical ICUs. Even less is known about the broader category of medical ICU patients without active oncological disease. This group constitutes a very large share of the patients requiring critical care across the globe, especially in the era of the SARS‐CoV‐2 pandemic. We analysed medical ICU candidaemia episodes that occurred in non‐oncological patients in our tertiary academic centre in the United States from May 2014 to October 2020 to determine the incidence and species distribution of the associated isolates. We then separately considered non‐COVID‐19 and COVID‐19 cases and compared their characteristics. In the non‐COVID‐19 group, there were 38 cases for an incidence of 1.1% and rate of 11/1000 admissions. In the COVID‐19 group, there were 12 cases for an incidence of 5.1% and rate of 51/1000 admissions. In the entire sample, as well as separately in the non‐COVID‐19 and COVID‐19 groups,Candida albicans accounted for a minority of isolates. Compared to non‐COVID‐19 patients with candidaemia, COVID‐19 patients had lower ICU admission SOFA score but longer ICU length of stay and central venous catheter dwell time at candidaemia detection. This study provides valuable insight into the incidence and species distribution of candidaemia cases occurring in non‐oncological critically ill patients and identifies informative differences between non‐COVID‐19 and COVID‐19 patients.

Risk factors for candidemia: a prospective matched case-control study

Background

Candidemia is an opportunistic infection associated with high morbidity and mortality in patients hospitalized both inside and outside intensive care units (ICUs). Identification of patients at risk is crucial to ensure prompt antifungal therapy. We sought to assess risk factors for candidemia and death, both outside and inside ICUs.

Methods

This prospective multicenter matched case-control study involved six teaching hospitals in Switzerland and France. Cases were defined by positive blood cultures for Candida sp. Controls were matched to cases using the following criteria: age, hospitalization ward, hospitalization duration, and, when applicable, type of surgery. One to three controls were enrolled by case. Risk factors were analyzed by univariate and multivariate conditional regression models, as a basis for a new scoring system to predict candidemia.

Results

One hundred ninety-two candidemic patients and 411 matched controls were included. Forty-four percent of included patients were hospitalized in ICUs, and 56% were hospitalized outside ICUs. Independent risk factors for candidemia in the ICU population included total parenteral nutrition, acute kidney injury, heart disease, prior septic shock, and exposure to aminoglycoside antibiotics. Independent risk factors for candidemia in the non-ICU population included central venous catheter, total parenteral nutrition, and exposure to glycopeptides and nitroimidazoles. The accuracy of the scores based on these risk factors is better in the ICU than in the non-ICU population. Independent risk factors for death in candidemic patients included septic shock, acute kidney injury, and the number of antibiotics to which patients were exposed before candidemia.

Discussion

While this study shows a role for known and novel risk factors for candidemia, it specifically highlights important differences in their distribution according to the hospital setting (ICU versus non-ICU).

Conclusion

This study provides novel risk scores for candidemia accounting for the hospital setting and recent progress in patients’ management strategies and fungal epidemiology.

A clinical predictive model of candidaemia by Candida auris in previously colonized critically ill patients

OBJECTIVES

Candida auris is an emerging multidrug-resistant fungus that has been associated with nosocomial outbreaks with high rates of mortality and transmission. The aim of this study was to perform a retrospective cohort analysis of risk factors and to build a scoring method for estimating the risk of candidaemia in colonized critically ill patients.

METHODS

We performed a retrospective observational cohort study of patients aged ≥15 years colonized by C. auris in the 3-year period between March 2016 and March 2019. Epidemiological, clinical, laboratory and microbiological data were collected. We developed a predictive model for candidaemia using elastic net multivariable logistic regression techniques, assessed its discriminative capacity, and internally validated it using bootstrap resampling.

RESULTS

Two-hundred and six patients were enrolled in the cohort for derivation and internal validation. Thirty-seven out of 206 patients developed candidaemia. Total parenteral nutrition was the foremost risk factor (adjusted OR 3.73); previous surgery (adjusted OR 1.03), sepsis (adjusted OR 1.75), previous exposure to antifungal agents (adjusted OR 1.17), arterial catheters (adjusted OR 1.46), central venous catheters (adjusted OR 1.21), presence of advanced chronic kidney disease (adjusted OR 1.35) and multifocal colonization (adjusted OR of unifocal colonization 0.46) were proven to be independent predictors of candidaemia in our cohort. The corresponding area under the curve (AUC) of the elastic net regularized predictive model was 0.89 (95%CI 0.826; 0.951). After performing the internal validation by generating 500 bootstrap replications, the model still showed great accuracy, with a resulting AUC of 0.84.

CONCLUSION

Our study provides evidence on the independent predisposing factors for candidaemia. It may help predict its estimated risk and may identify a high-risk population that could benefit from early or prophylactic antifungal treatment after external validation in other cohorts.

Risk factors of invasive candidiasis in critical cancer patients after various gastrointestinal surgeries: A 4-year retrospective study

For early diagnosis and treatment of invasive candidiasis (IC), the well-known risk factors may not apply in the intensive care unit (ICU). This retrospective study identified the risk factors predicting IC and candidemia in cancer patients under intensive care after gastrointestinal surgery.Enrolled were 229 cancer patients admitted to our oncology surgical ICU after gastrointestinal surgery between January 1, 2010 and October 31, 2014.The most common types of solid gastrointestinal cancers were gastric (49.8%), colon (20.1%), and esophageal (18.3%). The percentage of patients with corrected Candida colonization index (CCI) ≥0.4 was 31.9%. IC was confirmed in 19 patients (8.3%), and the ICU mortality was 15.8%. Candida albicans accounted for 52.6% of the total number of pathogenic Candida isolates. Among patients with CCI ≥0.4, the cancers with the highest prevalence were cardiac (45%) and gastric (36%), with ICU mortalities of 20% and 4.9%, respectively. For the diagnosis of candidemia, (1-3)-β-D-glucan (BDG) ≥80 pg/mL showed a sensitivity and specificity of 25% and 82.7%, respectively, positive and negative predictive values 6.7% and 95.7%, and area under the receiver operating characteristic curve 0.512. CCI ≥0.4 was the only significant predictor of IC, and number of organ failures was the only predictor of candidemia (P = .000 and .026).CCI ≥0.4 was the only significant risk factor predicting IC, with greater prediction of intra-abdominal candidiasis but failure to predict candidemia. Blood culture and BDG detection are recommended to supplement diagnosis. Patients may have multifocal and high-grade Candida colonization after cardiac surgery, and; therefore, are at high risk of IC, which should be taken seriously.

Impact of a diagnostics-driven antifungal stewardship programme in a UK tertiary referral teaching hospital

Objectives

A concise invasive candidosis guideline (based on the ESCMID candidaemia guideline) utilizing an informative biomarker [serum β-1-3-d-glucan (BDG)] was developed in 2013 by an antifungal stewardship (AFS) team and implemented with the help of an AFS champion in 2014. The main aims of the AFS programme were to reduce inappropriate use of antifungals and improve patient outcomes. The aim of this project was to evaluate the compliance of the ICU teams with the invasive candidosis guideline and the impact of the AFS programme on mortality and antifungal consumption on the ICUs (total of 71 beds).

Methods

All patients who were prescribed micafungin for suspected or proven invasive candidosis during 4 month audit periods in 2014 and 2016 were included. Prescriptions and patient records were reviewed against the guideline. Antifungal consumption and mortality data were analysed.

Results

The number of patients treated for invasive candidosis decreased from 39 in 2014 to 29 in 2016. This was mainly due to the reduction in patients initiated on antifungal therapy inappropriately: 18 in 2014 and 2 in 2016. Antifungal therapy was stopped following negative biomarker results in 12 patients in 2014 and 10 patients in 2016. Crude mortality due to proven or probable invasive candidosis decreased to 19% from 45% over the period 2003-07. Antifungal consumption reduced by 49% from 2014 to 2016.

Conclusions

The AFS programme was successful in reducing the number of inappropriate initiations of antifungals by 90%. Concurrently, mortality due to invasive candidosis was reduced by 58%. BDG testing can guide safe cessation of antifungals in ICU patients at risk of invasive candidosis.

The Overlooked Immune State in Candidemia: A Risk Factor for Mortality

Lymphopenia has been related to increased mortality in septic patients. Nonetheless, the impact of lymphocyte count on candidemia mortality and prognosis has not been addressed. We conducted a retrospective study, including all admitted patients with candidemia from 2007 to 2016. We examined lymphocyte counts during the first 5 days following the diagnosis of candidemia. Multivariable logistic regression analysis was performed to determine the relationship between lymphocyte count and mortality. Classification and Regression Tree analysis was used to identify the best cut-off of lymphocyte count for mortality associated with candidemia. From 296 cases of candidemia, 115 died, (39.8% 30-day mortality). Low lymphocyte count was related to mortality and poor outcome (p < 0.001). Lymphocyte counts <0.703 × 10⁹ cells/L at diagnosis (area under the curve (AUC)-ROC, 0.783 ± 0.042; 95% confidence interval (CI), 0.700-0.867, p < 0.001), and lymphocyte count <1.272 × 10⁹ cells/L five days later (AUC-ROC, 0.791 ± 0.038; 95%CI, 0.716-0.866, p < 0.001) increased the odds of mortality five-fold (odds ratio (OR), 5.01; 95%CI, 2.39-10.93) at time of diagnosis, and three-fold (OR, 3.27; 95%CI, 1.24-8.62) by day 5, respectively. Low lymphocyte count is an independent predictor of mortality in patients with candidemia and might serve as a biomarker for predicting candidemia-associated mortality and poor outcome.

Fungal infections in adult patients on extracorporeal life support

BACKGROUND

Patients on extracorporeal membrane oxygenation (ECMO) are often among the most severely ill in the intensive care unit. They are often receiving broad-spectrum antibiotics; they have multiple entry points for pathogens; and their immune system is impaired by blood circuit interaction. These factors are thought to predispose them to fungal infections. We thus aimed to evaluate the prevalence, risk factors, and prognosis of fungal infections in adults on ECMO.

METHODS

We conducted a retrospective cohort study using the Extracorporeal Life Support Organization registry, which compiles data on ECMO use from hundreds of international centers. We included all adult patients from 2006 to 2016 on any mode of ECMO with either a diagnosis of fungal infection or a positive fungal culture.

RESULTS

Our study comprised 2129 adult patients (10.8%) with fungal colonization or infection. Aspergillus involvement (colonization or infection) was present in 272 patients (1.4%), of whom 35.7% survived to hospital discharge. There were 245 patients (1.2%) with Candida invasive bloodstream infection, with 35.9% survival. Risk factors for Aspergillus involvement included solid organ transplant (OR 1.83; p = 0.008), respiratory support (OR 2.75; p < 0.001), and influenza infection (OR 2.48; p < 0.001). Risk factors for candidemia included sepsis (OR 1.60; p = 0.005) and renal replacement therapy (OR 1.55; p = 0.007). In multivariable analysis, Aspergillus involvement (OR 0.40; p < 0.001) and candidemia (OR 0.47; p < 0.001) were both independently associated with decreased survival.

CONCLUSIONS

The prevalence of Aspergillus involvement and Candida invasive bloodstream infection were not higher in patients on ECMO than what has been reported in the general intensive care population. Both were independently associated with a reduced survival. Aspergillus involvement was strongly associated with ECMO for respiratory support and influenza.

Candida and invasive mould diseases in non-neutropenic critically ill patients and patients with haematological cancer

Critically ill patients and patients with haematological cancer are HIV-negative populations at high risk of invasive fungal infections. In intensive-care units, candidaemia and intra-abdominal candidiasis predominate, but aspergillosis has emerged as a lethal, under-recognised cause of pneumonia. In patients with haematological malignancies or who have undergone stem-cell transplantations, pulmonary disease due to aspergillus and other mould diseases predominate. In this Series paper, we provide an update on risk assessment, new diagnostic strategies, and therapeutic approaches. New concepts have emerged for use of risk prediction rules and an evidence base now exists for inclusion of biomarkers (eg, galactomannan, 1,3-β-D-glucan, and PCR assays for Aspergillus spp) into early diagnostic and therapeutic strategies. Imaging techniques remain helpful for early diagnosis of pulmonary mould diseases, with PET techniques offering potential improvements in diagnostic specificity and evaluation of clinical response. Echinocandins and triazoles have been validated extensively for prophylaxis, empirical therapy, and targeted therapy, but an increase in intrinsically resistant fungi and emergence of secondary resistance as a result of drug-induced selection pressure are of major concern. Echinocandins remain a major component of treatment of invasive candidiasis and new triazoles are the best alternative for prophylaxis and therapy of invasive aspergillosis.