Cluster analysis to define distinct clinical phenotypes among septic patients with bloodstream infections

Bloodstream infection: Derivation and validation of a reliable and multidimensional prognostic score based on a machine learning model (BLISCO)

BACKGROUND

A bloodstream infection (BSI) prognostic score applicable at the time of blood culture collection is missing.

METHODS

In total, 4,327 patients with BSIs were included, divided into a derivation (80%) and a validation dataset (20%). Forty-two variables among host-related, demographic, epidemiological, clinical, and laboratory extracted from the electronic health records were analyzed. Logistic regression was chosen for predictive scoring.

RESULTS

The 14-day mortality model included age, body temperature, blood urea nitrogen, respiratory insufficiency, platelet count, high-sensitive C-reactive protein, and consciousness status: a score of ≥ 6 was correlated to a 14-day mortality rate of 15% with a sensitivity of 0.742, a specificity of 0.727, and an area under the curve of 0.783. The 30-day mortality model further included cardiovascular diseases: a score of ≥ 6 predicting 30-day mortality rate of 15% with a sensitivity of 0.691, a specificity of 0.699, and an area under the curve of 0.697.

CONCLUSIONS

A quick mortality score could represent a valid support for prognosis assessment and resources prioritizing for patients with BSIs not admitted in the intensive care unit.

Risk Factors for Bacteremia and Its Clinical Impact on Complicated Community-Acquired Urinary Tract Infection

Bacteremia has been associated with severity in some infections; however, its impact on the prognosis of urinary tract infections (UTIs) is still disputed. Our goal is to determine the risk factors for bacteremia and its clinical impact on hospitalized patients with complicated community-acquired urinary tract infections. We conducted a prospective observational study of patients admitted to the hospital with complicated community-acquired UTIs. Clinical variables and outcomes of patients with and without bacteremia were compared, and multivariate analysis was performed to identify risk factors for bacteremia and mortality. Of 279 patients with complicated community-acquired UTIs, 37.6% had positive blood cultures. Risk factors for bacteremia by multivariate analysis were temperature ≥ 38 °C (p = 0.006, OR 1.3 (95% CI 1.1-1.7)) and procalcitonin ≥ 0.5 ng/mL (p = 0.005, OR 8.5 (95% CI 2.2-39.4)). In-hospital and 30-day mortality were 9% and 13.6%, respectively. Quick SOFA (p = 0.030, OR 5.4 (95% CI 1.2-24.9)) and Barthel Index <40% (p = 0.020, OR 4.8 (95% CI 1.3-18.2)) were associated with 30-day mortality by multivariate analysis. However, bacteremia was not associated with 30-day mortality (p = 0.154, OR 2.7 (95% CI 0.7-10.3)). Our study found that febrile community-acquired UTIs and elevated procalcitonin were risk factors for bacteremia. The outcomes in patients with bacteremia were slightly worse, but without significant differences in mortality.

Clinical phenotypes of chronic cough categorised by cluster analysis

BACKGROUND

Chronic cough is a heterogeneous disease with various aetiologies that are difficult to determine. Our study aimed to categorise the phenotypes of chronic cough.

METHODS

Adult patients with chronic cough were assessed based on the characteristics and severity of their cough using the COugh Assessment Test (COAT) and the Korean version of the Leicester Cough Questionnaire. A cluster analysis was performed using the K-prototype, and the variables to be included were determined using a correlation network.

RESULTS

In total, 255 participants were included in the analysis. Based on the correlation network, age, score for each item, and total COAT score were selected for the cluster analysis. Four clusters were identified and characterised as follows: 1) elderly with mild cough, 2) middle-aged with less severe cough, 3) relatively male-predominant youth with severe cough, and 4) female-predominant elderly with severe cough. All clusters had distinct demographic and symptomatic characteristics and underlying causes.

CONCLUSIONS

Cluster analysis of age, score for each item, and total COAT score identified 4 distinct phenotypes of chronic cough with significant differences in the aetiologies. Subgrouping patients with chronic cough into homogenous phenotypes could provide a stratified medical approach for individualising diagnostic and therapeutic strategies.

Cluster analysis categorizes five phenotypes of pulmonary tuberculosis

Tuberculosis (TB) has a heterogeneous phenotype, which makes it challenging to diagnose. Our study aimed to identify TB phenotypes through cluster analysis and compare their initial symptomatic, microbiological and radiographic characteristics. We systemically collected data of notified TB patients notified in Korea and constructed a prospective, observational cohort database. Cluster analysis was performed using K-means clustering, and the variables to be included were determined by correlation network. A total of 4,370 subjects with pulmonary TB were enrolled in the study. Based on the correlation network, age and body mass index (BMI) were selected for the cluster analysis. Five clusters were identified and characterised as follows: (1) middle-aged overweight male dominance, (2) young-aged relatively female dominance without comorbidities, (3) middle-aged underweight male dominance, (4) overweight elderly with comorbidities and (5) underweight elderly with comorbidities. All clusters had distinct demographic and symptomatic characteristics. Initial microbiologic burdens and radiographic features also varied, including the presence of cavities and bilateral infiltration, which reflect TB-related severity. Cluster analysis of age and BMI identified five phenotypes of pulmonary TB with significant differences at initial clinical presentations. Further studies are necessary to validate our results and to assess their clinical implications.

Clinical applications of machine learning in the survival prediction and classification of sepsis: coagulation and heparin usage matter

BACKGROUND

Sepsis is a life-threatening syndrome eliciting highly heterogeneous host responses. Current prognostic evaluation methods used in clinical practice are characterized by an inadequate effectiveness in predicting sepsis mortality. Rapid identification of patients with high mortality risk is urgently needed. The phenotyping of patients will assistant invaluably in tailoring treatments.

METHODS

Machine learning and deep learning technology are used to characterize the patients' phenotype and determine the sepsis severity. The database used in this study is MIMIC-III and MIMIC-IV ('Medical information Mart for intensive care') which is a large, public, and freely available database. The K-means clustering is used to classify the sepsis phenotype. Convolutional neural network (CNN) was used to predict the 28-day survival rate based on 35 blood test variables of the sepsis patients, whereas a double coefficient quadratic multivariate fitting function (DCQMFF) is utilized to predict the 28-day survival rate with only 11 features of sepsis patients.

RESULTS

The patients were grouped into four clusters with a clear survival nomogram. The first cluster (C_1) was characterized by low white blood cell count, low neutrophil, and the highest lymphocyte proportion. C_2 obtained the lowest Sequential Organ Failure Assessment (SOFA) score and the highest survival rate. C_3 was characterized by significantly prolonged PTT, high SIC, and a higher proportion of patients using heparin than the patients in other clusters. The early mortality rate of patients in C_3 was high but with a better long-term survival rate than that in C_4. C_4 contained septic coagulation patients with the worst prognosis, characterized by slightly prolonged partial thromboplastin time (PTT), significantly prolonged prothrombin time (PT), and high septic coagulation disease score (SIC). The survival rate prediction accuracy of CNN and DCQMFF models reached 92% and 82%, respectively. The models were tested on an external dataset (MIMIC-IV) and achieved good performance. A DCQMFF-based application platform was established for fast prediction of the 28-day survival rate.

CONCLUSION

CNN and DCQMFF accurately predicted the sepsis patients' survival, while K-means successfully identified the phenotype groups. The distinct phenotypes associated with survival, and significant features correlated with mortality were identified. The findings suggest that sepsis patients with abnormal coagulation had poor outcomes, abnormal coagulation increase mortality during sepsis. The anticoagulation effects of appropriate heparin sodium treatment may improve extensive micro thrombosis-caused organ failure.

Development and validation of BLOOMY prediction scores for 14-day and 6-month mortality in hospitalised adults with bloodstream infections: a multicentre, prospective, cohort study

BACKGROUND

The burden of bloodstream infections remains high worldwide and cannot be confined to short-term in-hospital mortality. We aimed to develop scores to predict short-term and long-term mortality in patients with bloodstream infections.

METHODS

The Bloodstream Infection due to Multidrug-resistant Organisms: Multicenter Study on Risk Factors and Clinical Outcomes (BLOOMY) study is a prospective, multicentre cohort study at six German tertiary care university hospitals to develop and validate two scores assessing 14-day and 6-month mortality in patients with bloodstream infections. We excluded patients younger than 18 years or who were admitted to an ophthalmology or psychiatry ward. Microbiological, clinical, laboratory, treatment, and survival data were prospectively collected on day 0 and day 3 and then from day 7 onwards, weekly. Participants were followed up for 6 months. All patients in the derivation cohort who were alive on day 3 were included in the analysis. Predictive scores were developed using logistic regression and Cox proportional hazards models with a machine-learning approach. Validation was completed using the C statistic and predictive accuracy was assessed using sensitivity, specificity, and predictive values.

FINDINGS

Between Feb 1, 2017, and Jan 31, 2019, 2568 (61·5%) of 4179 eligible patients were recruited into the derivation cohort. The in-hospital mortality rate was 23·75% (95% CI 22·15-25·44; 610 of 2568 patients) and the 6-month mortality rate was 41·55% (39·54-43·59; 949 of 2284). The model predictors for 14-day mortality (C statistic 0·873, 95% CI 0·849-0·896) and 6-month mortality (0·807, 0·784-0·831) included age, body-mass index, platelet and leukocyte counts, C-reactive protein concentrations, malignancy (ie, comorbidity), in-hospital acquisition, and pathogen. Additional predictors were, for 14-day mortality, mental status, hypotension, and the need for mechanical ventilation on day 3 and, for 6-month mortality, focus of infection, in-hospital complications, and glomerular filtration rate at the end of treatment. The scores were validated in a cohort of 1023 patients with bloodstream infections, recruited between Oct 9, 2019, and Dec 31, 2020. The BLOOMY 14-day score showed a sensitivity of 61·32% (95% CI 51·81-70·04), a specificity of 86·36% (83·80-88·58), a positive predictive value (PPV) of 37·57% (30·70-44·99), and a negative predictive value (NPV) of 94·35% (92·42-95·80). The BLOOMY 6-month score showed a sensitivity of 69·93% (61·97-76·84), a specificity of 66·44% (61·86-70·73), a PPV of 40·82% (34·85-47·07), and a NPV of 86·97% (82·91-90·18).

INTERPRETATION

The BLOOMY scores showed good discrimination and predictive values and could support the development of protocols to manage bloodstream infections and also help to estimate the short-term and long-term burdens of bloodstream infections.

FUNDING

DZIF German Center for Infection Research.

TRANSLATION

For the German translation of the abstract see Supplementary Materials section.

Pediatric sepsis phenotypes for enhanced therapeutics: An application of clustering to electronic health records

OBJECTIVE

The heterogeneity of pediatric sepsis patients suggests the potential benefits of clustering analytics to derive phenotypes with distinct host response patterns that may help guide personalized therapeutics. We evaluate the relative performance of latent class analysis (LCA) and K-means, 2 commonly used clustering methods toward the derivation of clinically useful pediatric sepsis phenotypes.

METHODS

Data were extracted from anonymized medical records of 6446 pediatric patients that presented to 1 of 6 emergency departments (EDs) between 2013 and 2018 and were thereafter admitted. Using International Classification of Diseases (ICD)-9 and ICD-10 discharge codes, 151 patients were identified with a sepsis continuum diagnosis that included septicemia, sepsis, severe sepsis, and septic shock. Using feature sets used in related clustering studies, LCA and K-means algorithms were used to derive 4 distinct phenotypic pediatric sepsis segmentations. Each segmentation was evaluated for phenotypic homogeneity, separation, and clinical use.

RESULTS

Using the 2 feature sets, LCA clustering resulted in 2 similar segmentations of 4 clinically distinct phenotypes, while K-means clustering resulted in segmentations of 3 and 4 phenotypes. All 4 segmentations identified at least 1 high severity phenotype, but LCA-identified phenotypes reflected superior stratification, high entropy approaching 1 (eg, 0.994) indicating excellent separation between estimated phenotypes, and differential treatment/treatment response, and outcomes that were non-randomly distributed across phenotypes (P < 0.001).

CONCLUSION

Compared to K-means, which is commonly used in clustering studies, LCA appears to be a more robust, clinically useful statistical tool in analyzing a heterogeneous pediatric sepsis cohort toward informing targeted therapies. Additional prospective studies are needed to validate clinical utility of predictive models that target derived pediatric sepsis phenotypes in emergency department settings.

Circulatory trajectories after out-of-hospital cardiac arrest: a prospective cohort study

BACKGROUND

Circulatory failure frequently occurs after out-of-hospital cardiac arrest (OHCA) and is part of post-cardiac arrest syndrome (PCAS). The aim of this study was to investigate circulatory disturbances in PCAS by assessing the circulatory trajectory during treatment in the intensive care unit (ICU).

METHODS

This was a prospective single-center observational cohort study of patients after OHCA. Circulation was continuously and invasively monitored from the time of admission through the following five days. Every hour, patients were classified into one of three predefined circulatory states, yielding a longitudinal sequence of states for each patient. We used sequence analysis to describe the overall circulatory development and to identify clusters of patients with similar circulatory trajectories. We used ordered logistic regression to identify predictors for cluster membership.

RESULTS

Among 71 patients admitted to the ICU after OHCA during the study period, 50 were included in the study. The overall circulatory development after OHCA was two-phased. Low cardiac output (CO) and high systemic vascular resistance (SVR) characterized the initial phase, whereas high CO and low SVR characterized the later phase. Most patients were stabilized with respect to circulatory state within 72 h after cardiac arrest. We identified four clusters of circulatory trajectories. Initial shockable cardiac rhythm was associated with a favorable circulatory trajectory, whereas low base excess at admission was associated with an unfavorable circulatory trajectory.

CONCLUSION

Circulatory failure after OHCA exhibits time-dependent characteristics. We identified four distinct circulatory trajectories and their characteristics. These findings may guide clinical support for circulatory failure after OHCA.

TRIAL REGISTRATION

ClinicalTrials.gov: NCT02648061.

Coagulation phenotypes in sepsis and effects of recombinant human thrombomodulin: an analysis of three multicentre observational studies

Background

A recent randomised trial showed that recombinant thrombomodulin did not benefit patients who had sepsis with coagulopathy and organ dysfunction. Several recent studies suggested presence of clinical phenotypes in patients with sepsis and heterogenous treatment effects across different sepsis phenotypes. We examined the latent phenotypes of sepsis with coagulopathy and the associations between thrombomodulin treatment and the 28-day and in-hospital mortality for each phenotype.

Methods

This was a secondary analysis of multicentre registries containing data on patients (aged ≥ 16 years) who were admitted to intensive care units for severe sepsis or septic shock in Japan. Three multicentre registries were divided into derivation (two registries) and validation (one registry) cohorts. Phenotypes were derived using k-means with coagulation markers, platelet counts, prothrombin time/international normalised ratios, fibrinogen, fibrinogen/fibrin-degradation-products (FDP), D-dimer, and antithrombin activities. Associations between thrombomodulin treatment and survival outcomes (28-day and in-hospital mortality) were assessed in the derived clusters using a generalised estimating equation.

Results

Four sepsis phenotypes were derived from 3694 patients in the derivation cohort. Cluster dA (n = 323) had severe coagulopathy with high FDP and D-dimer levels, severe organ dysfunction, and high mortality. Cluster dB had severe disease with moderate coagulopathy. Clusters dC and dD had moderate and mild disease with and without coagulopathy, respectively. Thrombomodulin was associated with a lower 28-day (adjusted risk difference [RD]: − 17.8% [95% CI − 28.7 to − 6.9%]) and in-hospital (adjusted RD: − 17.7% [95% CI − 27.6 to − 7.8%]) mortality only in cluster dA. Sepsis phenotypes were similar in the validation cohort, and thrombomodulin treatment was also associated with lower 28-day (RD: − 24.9% [95% CI − 49.1 to − 0.7%]) and in-hospital mortality (RD: − 30.9% [95% CI − 55.3 to − 6.6%]).

Conclusions

We identified four coagulation marker-based sepsis phenotypes. The treatment effects of thrombomodulin varied across sepsis phenotypes. This finding will facilitate future trials of thrombomodulin, in which a sepsis phenotype with high FDP and D-dimer can be targeted.

Supplementary Information

The online version contains supplementary material available at 10.1186/s13054-021-03541-5.

Early Detection of Sepsis With Machine Learning Techniques: A Brief Clinical Perspective

Sepsis is a major cause of death worldwide. Over the past years, prediction of clinically relevant events through machine learning models has gained particular attention. In the present perspective, we provide a brief, clinician-oriented vision on the following relevant aspects concerning the use of machine learning predictive models for the early detection of sepsis in the daily practice: (i) the controversy of sepsis definition and its influence on the development of prediction models; (ii) the choice and availability of input features; (iii) the measure of the model performance, the output, and their usefulness in the clinical practice. The increasing involvement of artificial intelligence and machine learning in health care cannot be disregarded, despite important pitfalls that should be always carefully taken into consideration. In the long run, a rigorous multidisciplinary approach to enrich our understanding in the application of machine learning techniques for the early recognition of sepsis may show potential to augment medical decision-making when facing this heterogeneous and complex syndrome.

Exploration of critical care data by using unsupervised machine learning

BACKGROUND AND OBJECTIVE

Identification of subgroups may be useful to understand the clinical characteristics of ICU patients. The purposes of this study were to apply an unsupervised machine learning method to ICU patient data to discover subgroups among them; and to examine their clinical characteristics, therapeutic procedures conducted during the ICU stay, and discharge dispositions.

METHODS

K-means clustering method was used with 1503 observations and 9 types of laboratory test results as features.

RESULTS

Three clusters were identified from this specific population. Blood urea nitrogen, creatinine, potassium, hemoglobin, and red blood cell were distinctive between the clusters. Cluster Three presented the highest blood products transfusion rate (19.8%), followed by Cluster One (15.5%) and cluster Two (9.3%), which was significantly different. Hemodialysis was more frequently provided to Cluster Three while bronchoscopy was done to Cluster One and Two. Cluster Three showed the highest mortality (30.4%), which was more than two-fold compared to Cluster One (14.1%) and Two (12.2%).

CONCLUSION

Three subgroups were identified and their clinical characteristics were compared. These findings may be useful to anticipate treatment strategies and probable outcomes of ICU patients. Unsupervised machine learning may enable ICU multi-dimensional data to be organized and to make sense of the data.

Cluster-Based Analysis of Infectious Disease Occurrences Using Tensor Decomposition: A Case Study of South Korea

For a long time, various epidemics, such as lower respiratory infections and diarrheal diseases, have caused serious social losses and costs. Various methods for analyzing infectious disease occurrences have been proposed for effective prevention and proactive response to reduce such losses and costs. However, the results of the occurrence analyses were limited because numerous factors affect the outbreak of infectious diseases and there are complex interactions between these factors. To alleviate this limitation, we propose a cluster-based analysis scheme of infectious disease occurrences that can discover commonalities or differences between clusters by grouping elements with similar occurrence patterns. To do this, we collect and preprocess infectious disease occurrence data according to time, region, and disease. Then, we construct a tensor for the data and apply Tucker decomposition to extract latent features in the dimensions of time, region, and disease. Based on these latent features, we conduct k-means clustering and analyze the results for each dimension. To demonstrate the effectiveness of this scheme, we conduct a case study on data from South Korea and report some of the results.

Different Types of Coagulase Are Associated With 28-Day Mortality in Patients With Staphylococcus aureus Bloodstream Infections

Background:Staphylococcus aureus (S. aureus), a leading cause of bacteremia and infective endocarditis, exploits the human coagulation system by using a wide range of specific virulence factors. However, the impact of these host-pathogen interactions on the outcome of patients with Staphylococcus aureus bacteremia (SAB) remains unclear.

Methods: A total of 178 patients with S. aureus bacteremia were included and analyzed regarding bacterial factors (coa gene size, vWbp, clfA, clfB, fnbA, fnbB, fib) and clinical parameters. A stepwise multivariate Cox regression model and a Partitioning Around Medoids (PAM) cluster algorithm were used for statistical analysis.

Results: Patients' risk factors for 28-day mortality were creatinine (OR 1.49, p < 0.001), age (OR 1.9, p < 0.002), fibrinogen (OR 0.44, p < 0.004), albumin (OR 0.63, p < 0.02), hemoglobin (OR 0.59, p < 0.03), and CRP (OR 1.72, p < 0.04). Five distinct bacterial clusters with different mortality rates were unveiled, whereof two showed a 2-fold increased mortality and an accumulation of specific coagulase gene sizes, 547-base pairs and 660-base pairs.

Conclusions: Based on the data obtained in the present study an association of coagulase gene size and fib regarding 28-day mortality was observed in patients with S. aureus bloodstream infections. Further animal and prospective clinical studies are needed to confirm our preliminary findings.

Phenotypic heterogeneity by site of infection in surgical sepsis: a prospective longitudinal study

Background

The role of site of infection in sepsis has been poorly characterized. Additionally, sepsis epidemiology has evolved. Early mortality has decreased, but many survivors now progress into chronic critical illness (CCI). This study sought to determine if there were significant differences in the host response and current epidemiology of surgical sepsis categorized by site of infection.

Study design

This is a longitudinal study of surgical sepsis patients characterized by baseline predisposition, insult characteristics, serial biomarkers, hospital outcomes, and long-term outcomes. Patients were categorized into five anatomic sites of infection.

Results

The 316 study patients were predominantly Caucasian; half were male, with a mean age of 62 years, high comorbidity burden, and low 30-day mortality (10%). The primary sites were abdominal (44%), pulmonary (19%), skin/soft tissue (S/ST, 17%), genitourinary (GU, 12%), and vascular (7%). Most abdominal infections were present on admission and required source control. Comparatively, they had more prolonged proinflammation, immunosuppression, and persistent organ dysfunction. Their long-term outcome was poor with 37% CCI (defined as > 14 in ICU with organ dysfunction), 49% poor discharge dispositions, and 30% 1-year mortality. Most pulmonary infections were hospital-acquired pneumonia. They had similar protracted proinflammation and organ dysfunction, but immunosuppression normalized. Long-term outcomes are similarly poor (54% CCI, 47% poor disposition, 32% 1-year mortality). S/ST and GU infections occurred in younger patients with fewer comorbidities, less perturbed immune responses, and faster resolution of organ dysfunction. Comparatively, S/ST had better long-term outcomes (23% CCI, 39% poor disposition, 13% 1-year mortality) and GU had the best (10% CCI, 20% poor disposition, 10% 1-year mortality). Vascular sepsis patients were older males, with more comorbidities. Proinflammation was blunted with baseline immunosuppression and organ dysfunction that persisted. They had the worst long-term outcomes (38% CCI, 67% poor disposition, 57% 1-year mortality).

Conclusion

There are notable differences in baseline predisposition, host responses, and clinical outcomes by site of infection in surgical sepsis. While previous studies have focused on differences in hospital mortality, this study provides unique insights into the host response and long-term outcomes associated with different sites of infection.

Machine learning in infection management using routine electronic health records: tools, techniques, and reporting of future technologies

BACKGROUND

Machine learning (ML) is increasingly being used in many areas of health care. Its use in infection management is catching up as identified in a recent review in this journal. We present here a complementary review to this work.

OBJECTIVES

To support clinicians and researchers in navigating through the methodological aspects of ML approaches in the field of infection management.

SOURCES

A Medline search was performed with the keywords artificial intelligence, machine learning, infection∗, and infectious disease∗ for the years 2014-2019. Studies using routinely available electronic hospital record data from an inpatient setting with a focus on bacterial and fungal infections were included.

CONTENT

Fifty-two studies were included and divided into six groups based on their focus. These studies covered detection/prediction of sepsis (n = 19), hospital-acquired infections (n = 11), surgical site infections and other postoperative infections (n = 11), microbiological test results (n = 4), infections in general (n = 2), musculoskeletal infections (n = 2), and other topics (urinary tract infections, deep fungal infections, antimicrobial prescriptions; n = 1 each). In total, 35 different ML techniques were used. Logistic regression was applied in 18 studies followed by random forest, support vector machines, and artificial neural networks in 18, 12, and seven studies, respectively. Overall, the studies were very heterogeneous in their approach and their reporting. Detailed information on data handling and software code was often missing. Validation on new datasets and/or in other institutions was rarely done. Clinical studies on the impact of ML in infection management were lacking.

IMPLICATIONS

Promising approaches for ML use in infectious diseases were identified. But building trust in these new technologies will require improved reporting. Explainability and interpretability of the models used were rarely addressed and should be further explored. Independent model validation and clinical studies evaluating the added value of ML approaches are needed.