European HPC cloud infrastructure for managing SARS-CoV-2 data in compliance with GDPR

Abstract   The Connecting European SARS-CoV-2 Cohorts to Increase Common and Effective Response to SARS-CoV-2 Pandemic (ORCHESTRA) consortium, led by University of Verona (Italy), brings together key European academic experts and research institutions in infectious diseases, data management and High Performance Computing (HPC) from 26 organizations (extending to 37 partners) from 15 countries. The project aims to create a new pan-European cohort built on existing and new large-scale population cohorts in European and non-European countries to significantly impact on the responsiveness to SARS-CoV-2. The integration and analysis of the very heterogeneous characteristics of SARS-CoV-2 data coming from many different sources such as EHR, retrospective and prospective patient registries, and related ‘omics’ data (incl. genomics, proteomics and transcriptomics) can benefit of data analytics enabled by HPC, where both high compute performance and fast storage capabilities are immensely important. During the first year of the project, a dedicated HPC cloud infrastructure have been designed and partially deployed to fulfill the functional requirements for data management ensuring healthcare data confidentiality/privacy, integrity and security in compliance with the European GDPR regulations. The result is an infrastructure for Data Management composed by three main layers: National Data Providers; National Hubs (one for each HPC center involved: CINECA - Italy, CINES - France and HLRS - Germany), to centralize data at national level and to support data storage, sharing and analysis on data ingested from the National Data Providers; ORCHESTRA Data Portal: the pan-European portal for sharing aggregated data and results. Currently data collection is on going; at the end of the project, clinical centers are expected to have enrolled more than 10.000 patients with about 50.000 samples for the prospective studies. Key messages • The SARS-CoV-2 crisis made evident the need to manage and analyse very heterogeneous health data coming from many different resources across different countries. • The HPC cloud infrastructure released for the Orchestra project can act as a model to manage future public health threats.


Background:
There is increasing recognition that Public Health Institutes need to build on the traditional value for money approach, to find ways to capture, measure and show the full range of their outcomes, impacts and related value. As part of a drive to measure value and impact in public health and demonstrate how investment in health can contribute to an Economy of Well-being, Public Health Wales has developed an interactive database to capture and illustrate the social value of public health services and interventions. Methods: Scoping reviews of both academic and grey literature were undertaken to populate a database of health economics evaluations of public health interventions, focusing on Social Return on Investment (SROI). In addition, a simulated methodology was developed which allows the evidence to be manipulated and made relevant to individual contexts to help inform investment decisions at a local level.

Results:
To date, the database has accumulated an excess of 50 SROI evaluations of various public health interventions, across areas including mental health, behaviour change, physical activity, nutrition, employment and primary care. The evaluations are based on European and International contexts, are published in both grey and academic sources, and are of varying quality. Conclusions: SROI is a credible method for measuring the value of wider social, economic and environment outcomes achieved from public health interventions. The Social Value Database and Simulator presents a collation of studies and analysis utilising innovative health economics methods.

Background:
The validity of self-reported disease prevalence estimates in health surveys may be low when compared to data from medical records in administrative registers. Such discrepancies reflect a low content validity of the survey question, which may ultimately compromise the application of these survey data for public health purposes. The aim of the present study was to examine the agreement of self-reports of seven diseases with data from administrative registers, both overall and by sociodemographic characteristics.

Methods:
Prevalence estimates of self-reported current and/or previous diabetes, asthma, rheumatoid arthritis, osteoporosis, myocardial infarction, apoplexy, and cancer, respectively, were derived from the Danish National Health Survey in 2017 (n = 183,372 adults aged 16 years). Individual-level data were linked to registry data on the same diseases. Sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), kappa, and total agreement between self-reported and registry-documented prevalence estimates were examined.

Results:
For all included diseases, the specificity was >92%, and the sensitivity varied between 59% (cancer) and 95% (diabetes). NPV was >94% for all diseases and PPV varied between 13% (rheumatoid arthritis) and 93% (cancer). Total agreement varied between 91 % (asthma) and 99% (diabetes), whereas kappa was lowest for rheumatoid arthritis (0.21) and highest for diabetes (0.88). Sociodemographic variables were significantly associated with total agreement with sex, age, and educational level exhibiting the strongest associations.

Conclusions:
Overall, total agreement, specificity, and NPV between selfreported and registry-documented disease prevalence estimates are high, but PPV and kappa vary greatly between diseases. The latter findings reflect a low content validity of the applied survey question for specific diseases. This should be taken into account when interpreting similar results from surveys.

Key messages:
The validity of self-reported disease prevalence estimates may be low when compared to data from medical records. We found positive predictive values and kappa to vary greatly between diseases. Future studies should aim at designing survey questions properly in order to ensure a high content validity of the applied question.

Background:
Up to 7% of the Swedish population meets criteria for harmful use or alcohol dependency but only 10-20% seek treatment. One of the most recommended psychological treatments for controlled drinking is Motivational Enhancement Therapy (MET). Behavioural Self-Control Training (BSCT) is another treatment that is unique in that it is based on the psychology of learning and specifically focused on skills training. To our knowledge, no previous studies exist that evaluated the costeffectiveness of BSCT for alcohol use disorders (AUD). The aim of this study is to assess the cost-effectiveness of BSCT compared to MET for patients with AUD aiming for controlled drinking over the longer-term from a societal perspective.

Methods:
We modelled a cohort of patients with AUD who aim for controlled drinking, over a 10 year time horizon, and estimated the expected costs and outcomes of BSCT and MET. The model reflects the epidemiological transitions between drinking states, which reflect different levels of daily alcohol intake. Each drinking state is connected to temporary or long-term complications attributable to alcohol consumption, different costs and utilities. The data was sourced from a randomized trial evaluating the effectiveness of MET vs BSCT. Risks for complications and associated costs, utilities and mortality were sourced from the literature.

Methods:
From a mapping exercise of policies and programs in PM in EU and China, we identified 20 priority items for shaping sustainable healthcare. Such items were submitted to several Chinese and European experts in PM involved in a 3-round Delphi survey. Experts were asked to review the items' content and rate their validity and relevance on a 5-point Likert scale. Priorities reaching a Content Validity Index of more than 79% were included, between 70 and 79% were revised, and less than 70% were excluded.

Results:
Of 20 priorities submitted, 9 reached consensus. The priorities hinge on the resources allocation, defining in advance priority investment, and identifying new payment models for public reimbursement, health technology impact, and assessment importance, while integrating end-user perceptions into the whole innovation process. In addition, the pivotal role of multidisciplinary and cross-sectorial collaborations emerged. Ethical, legal, and social implications and the related costs should be always considered in policymaking, evaluation, and management of technological innovation.

Conclusions:
Integrating resources and setting a clear agenda for the implementation of PM would lead to a faster and more efficient translation into clinical practice. Developing policies valuing all the stakeholders' contributions would implement PM adoption.

Key messages:
Healthcare systems sustainability is a priority and PM could make the population healthier and help allocate resources more efficiently, hence reducing the overall costs of healthcare.