Abstract: PO0756
Enhanced Sentinel Surveillance System for COVID-19 Outbreak Prediction in a European Dialysis Clinics Network
Session Information
- COVID-19: Dialysis Patients
October 22, 2020 | Location: On-Demand
Abstract Time: 10:00 AM - 12:00 PM
Category: Coronavirus (COVID-19)
- 000 Coronavirus (COVID-19)
Authors
- Bellocchio, Francesco, Fresenius Medical Care Italia SpA, Palazzo Pignano, Lombardia, Italy
- Carioni, Paola, Fresenius Medical Care Italia SpA, Palazzo Pignano, Lombardia, Italy
- Garbelli, Mario, Fresenius Medical Care Italia SpA, Palazzo Pignano, Lombardia, Italy
- Martínez-Martínez, Francisco, Fresenius Medical Care Deutschland GmbH, Bad Homburg, Germany
- Larkin, John W., Fresenius Medical Care North America, Waltham, Massachusetts, United States
- Usvyat, Len A., Fresenius Medical Care North America, Waltham, Massachusetts, United States
- Maddux, Franklin W., Fresenius Medical Care North America, Waltham, Massachusetts, United States
- Stuard, Stefano, Fresenius Medical Care Deutschland GmbH, Bad Homburg, Germany
- Neri, Luca, Fresenius Medical Care Italia SpA, Palazzo Pignano, Lombardia, Italy
Background
Accurate predictions of epidemic dynamics may enable timely organizational interventions in high risk regions.
We exploited the interconnection of the EMEA Fresenius Medical Care (FMC) dialysis clinic network to establish a sentinel surveillance system where the occurrence of new cases in a clinic propagates distance-weighted risk estimates to proximal dialysis units. The surveillance system is embedded in an artificial intelligence model which predicts COVID-19 outbreak occurrence in HD clinics from trends in clinical practice patterns and regional COVID-19 epidemic metrics. The system stratifies clinics by their risk of new local outbreak.
Methods
The risk prediction model is computed considering a cohort of 640 clinics belonging to the FMC network. We trained a model to predict outbreak in each clinic in a 2-week prediction horizon (i.e. two or more COVID-19 cases). In addition to sentinel distance-weighted risk estimates, the model included 73 variables (i.e. regional-level epidemic data from open source datasets and clinical practice data from the EuCliD® database). We generated the training set on data available on 04/01/2020 and tested prediction accuracy at 4/15/2020 and 4/20/2020.
Results
In the training set there were 58 (9.1%) clinics with two or more patients with COVID-19 infection in the two-week prediction window. In the validation samples there were 27 (4.2%) and 12 (1.9%) clinics with two or more patients with COVID-19 infection during the two-week prediction window. The performance of the model was suitable in both testing windows (AUC=0.86 and 0.80 respectively). The model is used to construct risk maps highlighting geographical clusters of clinics at risk (figure).
Conclusion
A sentinel surveillance system together with the wealth of information collected in EuCliD® and state of the art modeling strategies allows prompt risk assessment and timely response to COVID-19 epidemic challenges throughout networked European clinics.
Funding
- Commercial Support – Fresenius Medical Care