No abstract

Zusammenfassung Hintergrund und Ziele Schon in der frühen Phase der global sehr verschieden verlaufenden COVID-19-Pandemie zeigten sich Hinweise auf den Einfluss sozioökonomischer Faktoren auf die Ausbreitungsdynamik der Erkrankung, die vor allem ab der zweiten Phase (September 2020) Menschen mit geringerem sozioökonomischen Status stärker betraf. Solche Effekte können sich auch innerhalb einer Großstadt zeigen. Die vorliegende Studie visualisiert und untersucht die zeitlich-räumliche Verbreitung aller in Köln gemeldeten COVID-19-Fälle (Februar 2020–Oktober 2021) auf Stadtteilebene und deren mögliche Assoziation mit sozioökonomischen Faktoren. Methoden Pseudonymisierte Daten aller in Köln gemeldeten COVID-19-Fälle wurden geocodiert, deren Verteilung altersstandardisiert auf Stadtteilebene über 4 Zeiträume kartiert und mit der Verteilung von sozialen Faktoren verglichen. Der mögliche Einfluss der ausgewählten Faktoren wird zudem in einer Regressionsanalyse in einem Modell mit Fallzuwachsraten betrachtet. Ergebnisse Das kleinräumige lokale Infektionsgeschehen ändert sich im Pandemieverlauf. Stadtteile mit schwächeren sozioökonomischen Indizes weisen über einen großen Teil des pandemischen Verlaufs höhere Inzidenzzahlen auf, wobei eine positive Korrelation zwischen den Armutsrisikofaktoren und der altersstandardisierten Inzidenz besteht. Die Stärke dieser Korrelation ändert sich im zeitlichen Verlauf. Schlussfolgerung Die zeitnahe Beobachtung und Analyse der lokalen Ausbreitungsdynamik lassen auch auf der Ebene einer Großstadt die positive Korrelation von nachteiligen sozioökonomischen Faktoren auf die Inzidenzrate von COVID-19 erkennen und können dazu beitragen, lokale Eindämmungsmaßnahmen zielgerecht zu steuern.

The COVID-19 pandemic and the high numbers of infected individuals pose major challenges for public health departments. To overcome these challenges, the health department in Cologne has developed a software called DiKoMa. This software offers the possibility to track contact and index persons, but also provides a digital symptom diary. In this work, the question of whether these can also be used for diagnostic purposes will be investigated. Machine learning makes it possible to identify infections based on early symptom profiles and to distinguish between the predominant dominant variants. Focusing on the occurrence of the symptoms in the first week, a decision tree is trained for the differentiation between contact and index persons and the prevailing dominant variants (Wildtype, Alpha, Delta, and Omicron). The model is evaluated, using sex- and age-stratified cross-validation and validated by symptom profiles of the first 6 days. The variants achieve an AUC-ROC from 0.89 for Omicron and 0.6 for Alpha. No significant differences are observed for the results of the validation set (Alpha 0.63 and Omicron 0.87). The evaluation of symptom combinations using artificial intelligence can determine the individual risk for the presence of a COVID-19 infection, allows assignment to virus variants, and can contribute to the management of epidemics and pandemics on a national and international level. It can help to reduce the number of specific tests in times of low labor capacity and could help to early identify new virus variants.

Background High rate of people infected with SARS-CoV-2 and their contacts in Cologne, Germany required innovative tools for notification, monitoring and reporting. The digital tool for COVID19 (DiKoMa) provides self-service symptom diaries allowing (a) the stratification for prioritized telephone contact by the health authority and (b) training a machine learning (ML) model that predicts infections with prevailing dominant variant (PDV) from early symptom profiles (SP). Methods Pseudononymized SP covering the first week of diary recordings were included for training (16646 index, 11582 contacts). A balanced random forest (BRF) model was trained to differentiate early predictive symptom patterns of different PDV and contact persons. Model evaluation was performed using sex and age stratified cross validation (CV), the model was validated on SP recorded from days 1 and 6. Results From 03/20 to 02/22, 90478 indeces and 75444 contact persons reported symptoms and health status, covering 46% and 42% of all reported cases, respectively. Diaries contained between 1-52 entries (566791, median 2). Daily analysis of entries, prioritized according to age, prevalent co-morbidities and detoriation of symptoms allowed risk adjusted follow up even during phases with high case notification rates. The top 5 predictive factors of the BRF were immunization, cough, dysgeusia and dysnosmia, fatigue, and sniffles to differentiate infection between wildtype, three PDV and contact persons (CV AUC 80.6%, Validation AUC 77.1%). Conclusions The use of digital symptom diary surveillance helps to provide appropriate medical support for patients on a large scale. Machine learning shows potential for symptom based risk assessment to differentiate PDV for future outbreaks and can thus become a valuable tool alongside specific laboratory diagnostics. Key messages • Digital symptom diaries are a powerful and widely accepted tool to attend COVID19 patients in isolation. They allow risk stratification for follow up and are a low-threshold service. • Machine learning supports index case identification by symptom analysis and can thus become a valuable tool alongside specific laboratory diagnostics.

BackgroundDiagnosis and treatment of PsA and axSpA is often delayed due to missing clear diagnostic criteria and limitations in resources for referral to rheumatologist including high numbers of incorrect referrals. Primary care is usually provided by either general practitioner, dermatologists, or orthopedics. Clinical discriminators with a high specificity for rheumatic conditions include morning stiffness (MST; peripheral or axial, >30min). Artificial intelligence (AI) and natural language processing (NLP) methods offer algorithms for learning systems to recognize disease associated terms and classify clinical phenotypes using large data sets that may support early identification of patients with suspected diagnosis of PsA or axSpA.ObjectivesAI and NLP methods are used to identify patients with typical attributes for inflammation by using morning stiffness as one potential discriminating pattern, which can be detected automatically and might help to prioritize referral for rheumatologist appointments.MethodsWithin a multicentre observational study, patients with visits at the rheumatologist with a suspected diagnosis of PsA or axSpA from the referral primary care provider were recruited. All data on clinical examinations and findings were collected and evaluated by rheumatologists in focus on criteria for diagnosis of PsA/axSpA (gold standard for evaluation). Unstructured text from the patient history was used to extract diagnosis-relevant characteristics. The information extraction algorithms used NLP models to detect expert curated “morning stiffness” (MST) keywords and puts them into a contextualized framework that allows to capture possible negations.ResultsA total of 116 patients were recruited (73 female, 63%) with a median age of 42 (IQR: 34-54). 51 patients were referred as axSpA (44%) and 60 as PsA (52%) by primary care providers. After preselection for PsA and axSpA patients, we observed a 23% rate of referrals without rheumatic diagnosis. Only 7.1% of patients were admitted without signs of MST, 29% with axial MST, 35% with peripheral MST and 28% with both MST types. Average morning stiffness duration was recorded as 35 minutes; patients with a finally confirmed rheumatic diagnosis had a higher average MST duration reported (36 minutes) compared to patients without a confirmed diagnosis. Our AI assisted extraction of MST identified MST in 82.7% of patient history texts. In 75% NLP methods correctly identified the negation of MST symptoms (6 of 8), and 94% of MST was detected when both axial and peripheral joints were affected (30 of 32). Manual inspection of 20 patient history reports where MST was not detected by our automated algorithm revealed that 17 reports did not contain information about MST and three mention unspecific early morning discomfort, without mention of MST.ConclusionThe high rate of correct detection of MST from patient history text using NLP methods allowed us to assess the potential for NLP models to support automated analysis of patient reports to facilitate intelligent patient referral.AcknowledgementsWe thank the Fraunhofer Excellence Cluster for Immune-Mediated Diseases CIMD for the financial support.Disclosure of InterestsNone declared