Retrospective Harmonisation of Mental Health and Resilience Data: A Federated Network of Covid-19 Cohorts with 40,000 participants (Co-RESPOND)
Wed-01
Presented by: Papoula Petri-Romão
Background
Psychological and epidemiological studies require large sample sizes to increase the power of their analyses, one way to achieve this is to harmonising several studies in an individual participant data meta-analysis. Specifically, the Covid-19 pandemic has dramatically illustrated the urge to ameliorate the standards of data management and stewardship in order to be able to collaboratively exploit existing data sets in an optimal way. While the FAIR publication guidelines that call for the improvement of the findability, accessibility, interoperability and reusability of digital assets have been published as early as 20161[, little has translated into research practice In light of this, the Co-RESPOND network is a collaboration of European cohort studies on mental health which has been initiated to identify resilient and vulnerable groups and inform European health authorities about inequities (EU Horizon 2020 programme, grant no. 101016127).
Objectives
This talk aims to illustrate advantages and challenges of implementing a federated approach to harmonising longitudinal cohort data. Co-RESPOND highlights the opportunities that lie in retrospective merging of datasets and offers practical solutions which conform with GDPR and FAIR guidelines.
Research question(s) and/or hypothesis/es
How can longitudinal data on mental health retrospectively be harmonized in a reliable, valid, transparent and sustainable way? Which technical solutions are available to facilitate the harmonisation process and support the FAIR publication of the harmonisation results?
Method/Approach
The harmonisation process was done in accordance with guidance provided by the Maelstrom initiative21 The Maelstrom research group has published practical guidance on data harmonisation. Moreover, the open-source OBiBa software stack3 complements this approach by providing tools for data documentation, co-analysis, and dissemination. Using DataSHIELD4, a secure, federated network for data analysis (FND) allowing for remote analyses without physically sharing data and in line with data protection requirements can be build.
Results/Findings (expected)
At this point, nine European cohorts have joined the Co-RESPOND network. In close collaboration, the data harmonisation potential of the cohorts has been settled, relevant variables have been identified, and the data transformation of individual data sets has been finalized and documented. Quality checks of the transformed data sets are currently under way. In parallel, the partners have installed local servers that will be made (selectively) accessible via OBiBa once the final data sets are uploaded. The harmonised dataset will be analysed to understand trajectories of mental health and resilience between 2020 and 2022.
Conclusions and implications (expected).
The Maelstrom guidelines and the OBiBA software stack along with DataShield provide an excellent, free-of-charge solution to guide the retrospective harmonisation of existing data sets and facilitate its joint use within a living federated network of data. However, ongoing and future cohorts should consider international standards of variable codings more properly in order to make the data collected available for joint analyses in advance. This network enables sustainable collaboration between research groups and offer the opportunity for more studies and collaborators to be added in the future.
Psychological and epidemiological studies require large sample sizes to increase the power of their analyses, one way to achieve this is to harmonising several studies in an individual participant data meta-analysis. Specifically, the Covid-19 pandemic has dramatically illustrated the urge to ameliorate the standards of data management and stewardship in order to be able to collaboratively exploit existing data sets in an optimal way. While the FAIR publication guidelines that call for the improvement of the findability, accessibility, interoperability and reusability of digital assets have been published as early as 20161[, little has translated into research practice In light of this, the Co-RESPOND network is a collaboration of European cohort studies on mental health which has been initiated to identify resilient and vulnerable groups and inform European health authorities about inequities (EU Horizon 2020 programme, grant no. 101016127).
Objectives
This talk aims to illustrate advantages and challenges of implementing a federated approach to harmonising longitudinal cohort data. Co-RESPOND highlights the opportunities that lie in retrospective merging of datasets and offers practical solutions which conform with GDPR and FAIR guidelines.
Research question(s) and/or hypothesis/es
How can longitudinal data on mental health retrospectively be harmonized in a reliable, valid, transparent and sustainable way? Which technical solutions are available to facilitate the harmonisation process and support the FAIR publication of the harmonisation results?
Method/Approach
The harmonisation process was done in accordance with guidance provided by the Maelstrom initiative21 The Maelstrom research group has published practical guidance on data harmonisation. Moreover, the open-source OBiBa software stack3 complements this approach by providing tools for data documentation, co-analysis, and dissemination. Using DataSHIELD4, a secure, federated network for data analysis (FND) allowing for remote analyses without physically sharing data and in line with data protection requirements can be build.
Results/Findings (expected)
At this point, nine European cohorts have joined the Co-RESPOND network. In close collaboration, the data harmonisation potential of the cohorts has been settled, relevant variables have been identified, and the data transformation of individual data sets has been finalized and documented. Quality checks of the transformed data sets are currently under way. In parallel, the partners have installed local servers that will be made (selectively) accessible via OBiBa once the final data sets are uploaded. The harmonised dataset will be analysed to understand trajectories of mental health and resilience between 2020 and 2022.
Conclusions and implications (expected).
The Maelstrom guidelines and the OBiBA software stack along with DataShield provide an excellent, free-of-charge solution to guide the retrospective harmonisation of existing data sets and facilitate its joint use within a living federated network of data. However, ongoing and future cohorts should consider international standards of variable codings more properly in order to make the data collected available for joint analyses in advance. This network enables sustainable collaboration between research groups and offer the opportunity for more studies and collaborators to be added in the future.