Welcome to the ESPID 2021 Meeting Calendar
The Meeting will officially run on CEST (GMT + 2 / UTC + 2)
To convert the meeting times to your local time Click Here
The viewing of sessions cannot be accessed from this conference calendar. All sessions are accessible via the Main Lobby at the Virtual Platform.
Icons Legend: 
- Live Session 
- On Demand Session 
- On Demand with Live Q&A 
- Live ESPID Show![]()
Introduction (ID 2329)
Challenges of Clinical Decision Support Tools for Infectious Diseases (ID 201)
Artificial Intelligence in the Diagnosis of Respiratory Infections (ID 202)
A NEW INTEGRATED TOOL TO AID DIAGNOSIS OF FEVER WITHOUT SOURCE IN CHILDREN AT PEDIATRIC EMERGENCY DEPARTMENTS (ID 1380)
Abstract
Background
Management of febrile children presenting to the Emergency Department is challenging, as accurate tests that identify children with bacterial infection requiring antimicrobial therapy are not available. Diagnostic tests that identify pathogens have limitations for which host transcriptomic biomarkers may provide a promising complementary solution.
Methods
We developed a host response panel on Filmarray® platform that discriminates between viral and bacterial infection in less than 1 hour. We selected 7 bacterial, 5 viral markers and 3 housekeeping genes. We tested the panel on pediatric PAXgene blood samples prospectively collected from two independent cohorts, (1) 467 febrile patients recruited to a French multi-center study and (2) 339 febrile patients recruited in the European PERFORM study. A training set, test-set and a validation set were employed to optimally build a classifier and assess its clinical performance. The classification was compared to that of the C Reactive Protein (CRP).
Results
Using these 12 genes, we constructed a classifier for bacterial infection. We identified one third of the samples as very likely bacterial with a specificity of 97% and one third of the samples as very likely viral with a sensitivity at 99%. Two other classes were defined to classify the rest of the samples as likely bacterial and likely viral with a specificity at 84% and a sensitivity at 82% respectively. The performance obtained was better than that of CRP.
Conclusions
Our data highlight a new promising multiplex qPCR tool for a rapid discrimination between bacterial and viral infection in children with fever without source. This approach is currently being validated on the whole PERFORM cohort including more than 5,000 samples.
Clinical Trial Registration
Clinical trial registration: ANTOINE study: NCT03163628 and PERFORM study: NCT03502993
USE OF DIGITAL DEVICES TO ASSESS VACCINE HESITANCY AND PROMOTE PERTUSSIS VACCINATION AMONG PREGNANT WOMEN (ID 621)
Abstract
Background
Vaccination against pertussis in pregnancy is the main strategy to prevent the disease in the first trimester of life. An effective communication is essential to successfully engage pregnant women. The use of digital devices within the outpatient setting may be helpful to engage patients before and during the consultation. The aim of this study was to develop and pilot test an e-health tool to assess vaccine hesitancy and to deliver tailored information and education interventions to raise awareness and promote vaccine acceptance.
Methods
One-hundred-and-five participants were recruited in 4 gynaecological outpatients. Participants were invited to complete a self-administered psychometric questionnaire to assess vaccine hesitancy, disease beliefs and self-efficacy perception on health behaviours on a tablet. Participants were randomly allocated to three communication-format types providing equivalent content: 1)a single video simulating a patient-doctor conversation on the topic; 2)an interactive platform with five infographics videos; 3)a paper leaflet followed by a brief consultation with the physician. The intention to get vaccinated during pregnancy was assessed through a specific question before and after the intervention.
Results
In the pre-intervention phase there was no difference observed between groups in terms of the variable “intention” to get vaccinated. After the intervention, participants of groups 1 and 3 showed a higher intention to get vaccinated than group 2 at the Kruskal-Wallis test (H(2)=6.008, p<0.05). Post-intervention intention to vaccinate correlated with Individual Self-Efficacy (rs(105)=0.30, p<0.001) and was inversely associated with vaccine hesitancy (rs(105)=0.34, p<0.001).
Conclusions
We implemented and assessed the impact of different communication strategies to promote vaccine uptake among pregnant women. Our results suggest comparable effect may be obtained using simulated versus live patient-physician communication.
DEEPBREATH: THE POTENTIAL FOR INTELLIGENT LUNG AUSCULTATION IN PEDIATRIC TRIAGE (ID 898)
Abstract
Background
In children, respiratory disease is the leading cause of preventable deaths and also the primary context for antibiotic misuse. Lung auscultation is an established clinical exam in the assessment of respiratory disease, but interpretation is subjective with considerable inter-user bias and poor accuracy. Deep learning has the potential to provide more objective interpretation to improve predictive performance of this fundamental clinical exam.
Methods
We present DeepBreath: a series of deep learning models able to discriminate key clinical and diagnostic signatures from digital lung sounds for incorporation into a novel multi-parameter intelligent stethoscope, named Pneumoscope. Algorithms are derived from systematically collected digital lung auscultations on 133 pediatric (1-16 years) outpatients with acute respiratory disease (asthma n=51, pneumonia n=33, bronchiolitis/bronchitis n=40) and 101 healthy controls in Brazil and Geneva. For each patient, 30-second audio clips were recorded at 8 thoracic sites (apical and basal positions in anteroposterior and lateral planes) amounting to over 11 hours of sounds accompanied by clinical data on signs, symptoms, paraclinical tests, diagnoses and (for asthma) clinical outcomes during a 30-day follow-up.
Results
Among several interesting findings, we show that deep learning can detect dyspnea, reliably calculate respiratory rate, and has over 80% sensitivity and specificity in automatically discriminating pathological from healthy lung sounds as well as correctly identifying asthma. These models match or outperform human expert analyses and we explore the potential of this work to better standardise the evaluation of pediatric respiratory disease to guide clinical care, improve antibiotic stewardship and even automate analyses for remote patient-led monitoring of chronic conditions like asthma.
Conclusions
Artificial intelligence has the potential to better standardize and improve the interpretation of digital lung auscultation.
Clinical Trial Registration
ClinicalTrials.gov Identifier: NCT04528342