A machine-learning algorithm for diagnosis of multisystem inflammatory syndrome in children and Kawasaki disease in the USA: a retrospective model development and validation study

Clinical prediction model: Multisystem inflammatory syndrome in children versus Kawasaki disease

BACKGROUND

Multisystem inflammatory syndrome in children (MIS-C) is a rare but serious complication of severe acute respiratory syndrome coronavirus 2 infection. Features of MIS-C overlap with those of Kawasaki disease (KD).

OBJECTIVE

The study objective was to develop a prediction model to assist with this diagnostic dilemma.

METHODS

Data from a retrospective cohort of children hospitalized with KD before the coronavirus disease 2019 pandemic were compared to a prospective cohort of children hospitalized with MIS-C. A bootstrapped backwards selection process was used to develop a logistic regression model predicting the probability of MIS-C diagnosis. A nomogram was created for application to individual patients.

RESULTS

Compared to children with incomplete and complete KD (N = 602), children with MIS-C (N = 105) were older and had longer hospitalizations; more frequent intensive care unit admissions and vasopressor use; lower white blood cell count, lymphocyte count, erythrocyte sedimentation rate, platelet count, sodium, and alanine aminotransferase; and higher hemoglobin and C-reactive protein (CRP) at admission. Left ventricular dysfunction was more frequent in patients with MIS-C, whereas coronary abnormalities were more common in those with KD. The final prediction model included age, sodium, platelet count, alanine aminotransferase, reduction in left ventricular ejection fraction, and CRP. The model exhibited good discrimination with AUC 0.96 (95% confidence interval: [0.94-0.98]) and was well calibrated (optimism-corrected intercept of -0.020 and slope of 0.99).

CONCLUSIONS

A diagnostic prediction model utilizing admission information provides excellent discrimination between MIS-C and KD. This model may be useful for diagnosis of MIS-C but requires external validation.

Clinical Timing-Sequence Warning Models for Serious Bacterial Infections in Adults Based on Machine Learning: Retrospective Study

BACKGROUND

Serious bacterial infections (SBIs) are linked to unplanned hospital admissions and a high mortality rate. The early identification of SBIs is crucial in clinical practice.

OBJECTIVE

This study aims to establish and validate clinically applicable models designed to identify SBIs in patients with infective fever.

METHODS

Clinical data from 945 patients with infective fever, encompassing demographic and laboratory indicators, were retrospectively collected from a 2200-bed teaching hospital between January 2013 and December 2020. The data were randomly divided into training and test sets at a ratio of 7:3. Various machine learning (ML) algorithms, including Boruta, Lasso (least absolute shrinkage and selection operator), and recursive feature elimination, were utilized for feature filtering. The selected features were subsequently used to construct models predicting SBIs using logistic regression (LR), random forest (RF), and extreme gradient boosting (XGBoost) with 5-fold cross-validation. Performance metrics, including the receiver operating characteristic (ROC) curve and area under the ROC curve (AUC), accuracy, sensitivity, and other relevant parameters, were used to assess model performance. Considering both model performance and clinical needs, 2 clinical timing-sequence warning models were ultimately confirmed using LR analysis. The corresponding predictive nomograms were then plotted for clinical use. Moreover, a physician, blinded to the study, collected additional data from the same center involving 164 patients during 2021. The nomograms developed in the study were then applied in clinical practice to further validate their clinical utility.

RESULTS

In total, 69.9% (661/945) of the patients developed SBIs. Age, hemoglobin, neutrophil-to-lymphocyte ratio, fibrinogen, and C-reactive protein levels were identified as important features by at least two ML algorithms. Considering the collection sequence of these indicators and clinical demands, 2 timing-sequence models predicting the SBI risk were constructed accordingly: the early admission model (model 1) and the model within 24 hours of admission (model 2). LR demonstrated better stability than RF and XGBoost in both models and performed the best in model 2, with an AUC, accuracy, and sensitivity of 0.780 (95% CI 0.720-841), 0.754 (95% CI 0.698-804), and 0.776 (95% CI 0.711-832), respectively. XGBoost had an advantage over LR in AUC (0.708, 95% CI 0.641-775 vs 0.686, 95% CI 0.617-754), while RF achieved better accuracy (0.729, 95% CI 0.673-780) and sensitivity (0.790, 95% CI 0.728-844) than the other 2 approaches in model 1. Two SBI-risk prediction nomograms were developed for clinical use based on LR, and they exhibited good performance with an accuracy of 0.707 and 0.750 and a sensitivity of 0.729 and 0.927 in clinical application.

CONCLUSIONS

The clinical timing-sequence warning models demonstrated efficacy in predicting SBIs in patients suspected of having infective fever and in clinical application, suggesting good potential in clinical decision-making. Nevertheless, additional prospective and multicenter studies are necessary to further confirm their clinical utility.

Quantifying the impact of AI recommendations with explanations on prescription decision making

The influence of AI recommendations on physician behaviour remains poorly characterised. We assess how clinicians' decisions may be influenced by additional information more broadly, and how this influence can be modified by either the source of the information (human peers or AI) and the presence or absence of an AI explanation (XAI, here using simple feature importance). We used a modified between-subjects design where intensive care doctors (N = 86) were presented on a computer for each of 16 trials with a patient case and prompted to prescribe continuous values for two drugs. We used a multi-factorial experimental design with four arms, where each clinician experienced all four arms on different subsets of our 24 patients. The four arms were (i) baseline (control), (ii) peer human clinician scenario showing what doses had been prescribed by other doctors, (iii) AI suggestion and (iv) XAI suggestion. We found that additional information (peer, AI or XAI) had a strong influence on prescriptions (significantly for AI, not so for peers) but simple XAI did not have higher influence than AI alone. There was no correlation between attitudes to AI or clinical experience on the AI-supported decisions and nor was there correlation between what doctors self-reported about how useful they found the XAI and whether the XAI actually influenced their prescriptions. Our findings suggest that the marginal impact of simple XAI was low in this setting and we also cast doubt on the utility of self-reports as a valid metric for assessing XAI in clinical experts.

Kawasaki disease, multisystem inflammatory syndrome in children, and adenoviral infection: a scoring system to guide differential diagnosis

Children with Kawasaki disease (KD), Multisystem Inflammatory Syndrome in Children (MIS-C), and Adenovirus infections (AI) of the upper respiratory tract show overlapping features. This study aims to develop a scoring system based on clinical or laboratory parameters to differentiate KD or MIS-C from AI patients. Ninety pediatric patients diagnosed with KD (n = 30), MIS-C (n = 26), and AI (n = 34) admitted to the Pediatric Emergency Unit of S.Orsola University Hospital in Bologna, Italy, from April 2018 to December 2021 were enrolled. Demographic, clinical, and laboratory data were recorded. A multivariable logistic regression analysis was performed, and a scoring system was subsequently developed. A simple model (clinical score), including five clinical parameters, and a complex model (clinic-lab score), resulting from the addition of one laboratory parameter, were developed and yielded 100% sensitivity and 80% specificity with a score ≥2 and 98.3% sensitivity and 83.3% specificity with a score ≥3, respectively, for MIS-C and KD diagnosis, as compared to AI.

CONCLUSION

This scoring system, intended for both outpatients and inpatients, might limit overtesting, contribute to a more effective use of resources, and help the clinician not underestimate the true risk of KD or MIS-C among patients with an incidental Adenovirus detection.

WHAT IS KNOWN

• Kawasaki Disease (KD), Multisystem Inflammatory Syndrome in Children (MIS-C) and adenoviral infections share overlapping clinical presentation in persistently febrile children, making differential diagnosis challenging. • Scoring systems have been developed to identify high-risk KD patients and discriminate KD from MIS-C patients.

WHAT IS NEW

• This is the first scoring model based on clinical criteria to distinguish adenoviral infection from KD and MIS-C. • The score might be used by general pediatricians before referring febrile children to the emergency department.

Expression of TRAIL, IP-10, and CRP in children with suspected COVID-19 and real-life impact of a computational signature on clinical decision-making: a prospective cohort study

PURPOSE

We evaluated the host-response marker score "BV" and its components TRAIL, IP-10, and CRP in SARS-CoV-2 positive children, and estimated the potential impact on clinical decision-making.

METHODS

We prospectively analyzed levels of TRAIL, IP-10, CRP, and the BV score, in children with suspected COVID-19. Classification of infectious etiology was performed by an expert panel. We used a 5-point-questionnaire to evaluate the intention to treat with antibiotics before and after receiving test results.

RESULTS

We screened 111 children, of whom 6 (5.4%) were positive for SARS-CoV-2. A total of 53 children were included for the exploratory analysis. Median age was 3.1 years (interquartile range [IQR] 1.3-4.3), and 54.7% (n = 29) were girls. A viral and a bacterial biomarker pattern was found in 27/53 (50.9%) and 15/53 (28.3%), respectively. BV scores differed between COVID-19, children with other viral infections, and children with bacterial infections (medians 29.5 vs. 9 vs. 66; p = 0.0006). Similarly, median TRAIL levels were different (65.5 vs. 110 vs. 78; p = 0.037). We found no differences in IP-10 levels (555 vs. 504 vs. 285; p = 0.22). We found a concordance between physicians' "unlikely intention to treat" children with a viral test result in most cases (n = 19/24, 79.2%). When physicians expressed a "likely intention to treat" (n = 15), BV test revealed 5 bacterial, viral, and equivocal scores each. Antibiotics were withheld in three cases (20%). Overall, 27/42 (64%) of pediatricians appraised the BV test positively, and considered it helpful in clinical practice.

CONCLUSION

Host-response based categorization of infectious diseases might help to overcome diagnostic uncertainty, support clinical decision-making and reduce unnecessary antibiotic treatment.

Endothelial Cell Response in Kawasaki Disease and Multisystem Inflammatory Syndrome in Children

Although Kawasaki disease (KD) and multisystem inflammatory syndrome in children (MIS-C) share some clinical manifestations, their cardiovascular outcomes are different, and this may be reflected at the level of the endothelial cell (EC). We performed RNA-seq on cultured ECs incubated with pre-treatment sera from KD (n = 5), MIS-C (n = 7), and healthy controls (n = 3). We conducted a weighted gene co-expression network analysis (WGCNA) using 935 transcripts differentially expressed between MIS-C and KD using relaxed filtering (unadjusted p < 0.05, >1.1-fold difference). We found seven gene modules in MIS-C, annotated as an increased TNFα/NFκB pathway, decreased EC homeostasis, anti-inflammation and immune response, translation, and glucocorticoid responsive genes and endothelial-mesenchymal transition (EndoMT). To further understand the difference in the EC response between MIS-C and KD, stringent filtering was applied to identify 41 differentially expressed genes (DEGs) between MIS-C and KD (adjusted p < 0.05, >2-fold-difference). Again, in MIS-C, NFκB pathway genes, including nine pro-survival genes, were upregulated. The expression levels were higher in the genes influencing autophagy (UBD, EBI3, and SQSTM1). Other DEGs also supported the finding by WGCNA. Compared to KD, ECs in MIS-C had increased pro-survival transcripts but reduced transcripts related to EndoMT and EC homeostasis. These differences in the EC response may influence the different cardiovascular outcomes in these two diseases.

Emerging Insights Into the Pathophysiology of Multisystem Inflammatory Syndrome Associated With COVID-19 in Children

Multisystem inflammatory syndrome in children (MIS-C) has emerged as a rare delayed hyperinflammatory response to SARS-CoV-2 infection and causes severe morbidity in the pediatric age group. Although MIS-C shares many clinical similarities to Kawasaki disease (KD), important differences in epidemiologic, clinical, immunologic, and potentially genetic factors exist and suggest potential differences in pathophysiology and points to be explored and explained. Epidemiologic features include male predominance, peak age of 6 to12 years, and specific racial or ethnicity predilections. MIS-C is characterized by fever, prominent gastrointestinal symptoms, mucocutaneous manifestations, respiratory symptoms, and neurologic complaints, and patients often present with shock. Cardiac complications are frequent and include ventricular dysfunction, valvular regurgitation, pericardial effusion, coronary artery dilation and aneurysms, conduction abnormalities, and arrhythmias. Emerging evidence regarding potential immunologic mechanisms suggest that an exaggerated T-cell response to a superantigen on the SARS-CoV-2 spike glycoprotein-as well as the formation of autoantibodies against cardiovascular, gastrointestinal, and endothelial antigens-are major contributors to the inflammatory milieu of MIS-C. Further studies are needed to determine both shared and distinct immunologic pathway(s) that underlie the pathogenesis of MIS-C vs both acute SARS-CoV-2 infection and KD. There is evidence to suggest that the rare risk of more benign mRNA vaccine-associated myopericarditis is outweighed by a reduced risk of more severe MIS-C. In the current review, we synthesize the published literature to describe associated factors and potential mechanisms regarding an increased risk of MIS-C and cardiac complications, provide insights into the underlying immunologic pathophysiology, and define similarities and differences with KD.

Effectiveness of Using the FreeStyle Libre® System for Monitoring Blood Glucose during the COVID-19 Pandemic in Diabetic Individuals: Systematic Review

BACKGROUND

Artificial Intelligence (AI) is an area of computer science/engineering that is aiming to spread technological systems. The COVID-19 pandemic caused economic and public health turbulence around the world. Among the many possibilities for using AI in the medical field is FreeStyle Libre^® (FSL), which uses a disposable sensor inserted into the user's arm, and a touchscreen device/reader is used to scan and retrieve other continuous monitoring of glucose (CMG) readings. The aim of this systematic review is to summarize the effectiveness of FSL blood glucose monitoring during the COVID-19 pandemic.

METHODS

This systematic review was carried out in accordance with the Preferred Reporting Items for Systematic Review and Meta-Analyses (PRISMA) and was registered in the international prospective register of systematic reviews (PROSPERO: CRD42022340562). The inclusion criteria considered studies involving the use of the FSL device during the COVID-19 pandemic and published in English. No publication date restrictions were set. The exclusion criteria were abstracts, systematic reviews, studies with patients with other diseases, monitoring with other equipment, patients with COVID-19, and bariatrics patients. Seven databases were searched (PubMed, Scopus, Embase, Web of Science, Scielo, PEDro and Cochrane Library). The ACROBAT-NRSI tool (A Cochrane Risk of Bias Assessment Tool for Non-Randomized Studies) was used to evaluate the risk of bias in the selected articles.

RESULTS

A total of 113 articles were found. Sixty-four were excluded because they were duplicates, 39 were excluded after reading the titles and abstracts, and twenty articles were considered for full reading. Of the 10 articles analyzed, four articles were excluded because they did not meet the inclusion criteria. Thus, six articles were included in the current systematic review. It was observed that among the selected articles, only two were classified as having serious risk of bias. It was shown that FSL had a positive impact on glycemic control and on reducing the number of individuals with hypoglycemia.

CONCLUSION

The findings suggest that the implementation of FSL during COVID-19 confinement in this population can be confidently stated to have been effective in diabetes mellitus patients.