Rationale for a Swedish cohort consortium

Figures & data

Table 1. Examples of Swedish official registries.

Registry	Contents
Swedish Total Population Registry (14)	Place of residency; country of own and parents’ birth; marital status; date of death or emigration
Swedish Censuses	Socio-economic group; education; income; sick leave
Swedish National Insurance Agency	Sick leave; pensions
Swedish Education Registry	Highest education
Swedish 9th Grade Registry	Junior high school grades
Swedish Multi-Generation Registry (15)	Number of children and siblings; identity of parents if born after 1932 and alive in 1961
Swedish Medical Birth Registry (since 1973) (16)	Numbers of pregnancies and births; pregnancy outcomes
Swedish Prescribed Drug Registry (since 2005) (17)	Pharmacy-expedited drug prescriptions
Swedish Inpatient Registry (since 1964, with complete national coverage since 1987) (18)	Diagnoses of all hospitalizations; surgical and other procedures
Swedish Cancer Registry (since the 1950s) (19)	All cancer diagnoses
Swedish Cause-of-Death Registry (20)	Causes of death, including contributing factors
Swedish Out-Patient Registries (day-care surgery since 1997, all others since 2001)	All diagnoses. Hospital-based mandatory; primary care voluntary

Figure 1. Time (h) spent on processes in the pilot study. SAP: statistical analysis plan.

Figure 2. Resubmission to the ethics review board.

Table 2. Proposal for a Swedish Cohort Consortium (Cohorts.se).

Component	Description
Data management	Some Swedish cohorts are kept at the highest possible standard, with secure storage of data in modern data formats, including backups, controlled data access, and access logs. Many also have curated data with complete and documented procedures for data curation as a result of e.g. logical checks, monitoring, adjudication, or outlier detection; and with data curation traceable through an auditing feature in the data warehouse. Unfortunately, many valuable cohorts are kept under suboptimal circumstances in some or all of these respects. Sharing experiences and tools for efficient data management among cohort owners is an important feature of a cohort infrastructure. The Maelstrom Research offers guidelines (23) and a suite of open source software (24) (Opal, Onyx, Mica, and Agate) to support management of cohort data. Optimally, all cohorts in the infrastructure would run a local instance of Opal in conjunction with their other software stack (24). Opal can be used to achieve data management and harmonization. It is integrated with R, and, using the DataSHIELD approach, such an infrastructure could support advanced statistical data analysis across cohorts without having to share or disclose individual-level data (11).
Cataloguing	Cohort steering groups hold detailed knowledge and documentation on the cohorts’ design, sampling procedures, data collection methods, losses to follow-up, data quality, and other cumulative knowledge about the data gathered through a multitude of research projects. Given the complexity in the structure of the existing datasets and the rich content of the data, the experience of the cohort holders is invaluable for effective use of the data. This experience needs to be condensed into a useful metadata repository. In Cohorts.se, the Maelstrom approach for study and variables cataloguing is proposed. A key activity in joint cohort projects is searching for harmonization potential among cohorts with similar data. A prerequisite for this is a metadata catalogue containing structured details about the cohort design and the variables and samples collected in the cohort study. In Cohorts.se, it is proposed to use the Mica software offered by Maelstrom. Mica is a software application for web portals for individual cohorts or for consortia. When used in conjunction with the Maelstrom component Opal, for data management, Mica also allows authenticated users to perform distributed queries on the content of study databases hosted on remote servers and retrieve summary statistics (24). A pilot catalogue including 6 studies and 10,000 variables is accessible on www.maelstrom-research.org/mica/network/cohorts.se
Enrichment	In order to stay relevant, epidemiology needs collection of new cohorts, and re-investigation and expansion of existing cohorts. Cohorts.se proposes to customize and use the Maelstrom’s Onyx software to support high-quality collection of new cohorts. Onyx is a web-based application that manages participant logistics at cohort assessment clinics, including appointment management, controlling stage availability and dependencies, consents, questionnaires, sample collection (barcode scanning), linkage to sensory equipment such as ECGs or scales, producing personalized reports for participants, and exporting encrypted data to multiple destinations. In addition, Cohorts.se will aid in variable selection for new investigations of cohorts, guided by knowledge of usefulness of existing variable lists in other cohorts in Cohorts.se. Cohorts.se is also proposed to provide standardized integration of commonly used data sources like public national registries.
Project management	For researchers to conduct projects using the consortium, some information needs to be readily available, including cohort metadata, variable lists, definitions of variables, and standardized application forms for research proposals. We propose that Cohorts.se provides basic support for submission of research proposals, coordinates decisions to join a collaborative project, coordinates data access, provides statistical support as needed, and facilitates critical steps in the publication process. Further, Cohorts.se should facilitate compliance with legal and ethical requirements, and ensure complete anonymization of data before delivery to researchers, facilitate meta-analysis or, where relevant, use federated analysis such as DataSHIELD.
Harmonization	In order to make joint analyses, variables from different cohorts need to be measured on the same scale. Using the Maelstrom approach, harmonization of variables is facilitated and documented in stored scripts (25). Harmonized data can then be analysed within the Maelstrom system, or exported for analysis on other analysis platforms. The harmonization process is unique for each research project, but parts will be re-usable in subsequent projects. These parts will grow over time with shorter and shorter time needed for reprocessing.
Power calculations	Cohorts.se should be able to provide data useful for statistical power calculations for all cohorts storing data on an Opal server connected to Mica. At the moment, user-friendly tools for determining adequate sample size for more complex analyses that include clustered data, family designs, or multiple interactive effects are not readily available. Power calculations require detailed knowledge about the structure and contents of data in specific cohorts, including information on missing data for combinations of variables. Assuring that the existing data are adequate for a proposed analysis will likely be a common question.
Statistical analysis	Analysing harmonized data from different cohorts today often means transferring and pooling of individual data into a single large database, but that practice may be infeasible because of time, ethical, or legal issues. DataSHIELD offers a possibility to analyse data as if they were pooled, though still stored at local Opal servers behind firewalls (11). Cohorts.se can set up analysis servers at several participating universities, permitting different analyses of the joint cohort data from multiple access points.
Local cohort operations	The backbone of Cohorts.se is the participating cohorts and all persons involved in the management of the cohorts. For a successful infrastructure, incentives must be developed for the cohort holders to keep their cohorts safely stored in modern formats on adequate hardware, accessible for research projects, up to date in the Cohorts.se catalogue, sufficiently staffed with data management personnel, and governed by a functioning steering committee or other body. Within Cohorts.se, a charter with recommendations of levels of governance, maintenance, data storage, and data curation will be proposed.

Table 3. International infrastructures with which collaboration may be sought.

Name	Website
The Asia-Pacific Cohort Studies Collaboration	www.apcsc.net
Emerging Risk Factors Collaboration	www.phpc.cam.ac.uk/ceu/research/erfc
Prospective Studies Collaboration	www.ctsu.ox.ac.uk/research/meta-studies/psc/psc-website
Monica Risk, Genetics, Archiving and Monograph (MORGAM)	www.thl.fi/morgam
Biomarker for Cardiovascular Risk Assessment in Europe (BiomarCaRE)	www.biomarcare.eu
National Cancer Institute (NCI) Cohort Consortium	http://epi.grants.cancer.gov/Consortia/cohort.html
European Cohort Consortium (follow-up to BBMRI-LPC)	www.bbmri-lpc.org
Research on European Children and Adults born Preterm (RECAP)	https://recap-preterm.eu/

The Maelstrom software suite (23) is used for several of these collaborations, including the NCI Cohort Consortium, BBMRI-LPC, and RECAP. Using the same suite of tools for Cohorts.se will facilitate collaboration with those international consortia. Some Swedish cohorts are already catalogued in Maelstrom, including ULSAM, EpiHealth, LifeGene and TwinGene. The continuation of the BBMRI-LPC is uncertain, but there are initiatives to transform it into a European Cohort Consortium, which will most likely be open for inclusion of more Swedish cohorts. Some of the other networks that use Maelstrom are described on https://www.maelstrom-research.org.

Ludvigsson JF, Almqvist C, Bonamy AE, Ljung R, Michaelsson K, Neovius M, et al. Registers of the Swedish total population and their use in medical research. Eur J Epidemiol. 2016;31:125–36.

 PubMed Web of Science ®Google Scholar

Ekbom A. The Swedish multi-generation register. Methods Mol Biol. 2011;675:215–20.

 PubMedGoogle Scholar

Cnattingius S, Ericson A, Gunnarskog J, Kallen B. A quality study of a medical birth registry. Scand J Soc Med. 1990;18:143–8.

 PubMedGoogle Scholar

Wettermark B, Hammar N, Fored CM, Leimanis A, Otterblad Olausson P, Bergman U, et al. The new Swedish Prescribed Drug Register–opportunities for pharmacoepidemiological research and experience from the first six months. Pharmacoepidemiol Drug Saf. 2007;16:726–35.

 PubMed Web of Science ®Google Scholar

Ludvigsson JF, Andersson E, Ekbom A, Feychting M, Kim JL, Reuterwall C, et al. External review and validation of the Swedish national inpatient register. BMC Public Health. 2011;11:450.

 PubMed Web of Science ®Google Scholar

Barlow L, Westergren K, Holmberg L, Talback M. The completeness of the Swedish Cancer Register: a sample survey for year 1998. Acta Oncol. 2009;48:27–33.

 PubMed Web of Science ®Google Scholar

Brooke HL, Talback M, Hornblad J, Johansson LA, Ludvigsson JF, Druid H, et al. The Swedish cause of death register. Eur J Epidemiol. 2017;32:765–73.

 PubMed Web of Science ®Google Scholar

Fortier I, Raina P, Van den Heuvel ER, Griffith LE, Craig C, Saliba M, et al. Maelstrom research guidelines for rigorous retrospective data harmonization. Int J Epidemiol. 2016;46:103–5.

 Web of Science ®Google Scholar

Doiron D, Marcon Y, Fortier I, Burton P, Ferretti V. Software application profile: Opal and Mica: open-source software solutions for epidemiological data management, harmonization and dissemination. Int J Epidemiol. 2017;46:1372–8.

 PubMed Web of Science ®Google Scholar

Wolfson M, Wallace SE, Masca N, Rowe G, Sheehan NA, Ferretti V, et al. DataSHIELD: resolving a conflict in contemporary bioscience–performing a pooled analysis of individual-level data without sharing the data. Int J Epidemiol. 2010;39:1372–82.

 PubMed Web of Science ®Google Scholar

Fortier I, Doiron D, Little J, Ferretti V, L’Heureux F, Stolk RP, et al. Is rigorous retrospective harmonization possible? Application of the DataSHaPER approach across 53 large studies. Int J Epidemiol. 2011;40:1314–28.

 PubMed Web of Science ®Google Scholar