EULAR-endorsed overarching principles and points to consider for the use of big data in RMDs, with levels of agreement and for the specific points, levels of evidence and strength
| Definitions | |||
| The term ‘big data’ refers to extremely large datasets which may be complex, multidimensional, unstructured and from heterogeneous sources, and which accumulate rapidly. Computational technologies, including artificial intelligence (eg, machine learning), are often applied to big data. Big data may arise from multiple data sources including clinical, biological, social and environmental data sources | |||
| Overarching principles | LoA, mean (SD) | ||
| A. For all big data use, ethical issues related to privacy, confidentiality, identity and transparency are key principles to consider | 9.6 (0.7) | ||
| B. Big data provides unprecedented opportunities to deliver transformative discoveries in RMD research and practice | 9.5 (1.2) | ||
| C. The ultimate goal of using big data in RMDs is to improve the health, lives and care of people including health promotion and assessment, prevention, diagnosis, treatment and monitoring of disease | 9.6 (0.5) | ||
| Points to consider | LoA, mean (SD) | LoE | SoR |
| 1. The use of global, harmonised and comprehensive standards should be promoted to facilitate interoperability of big data | 9.7 (0.6) | 4 | C |
| 2. Big data should be Findable, Accessible, Interoperable and Reusable (FAIR principle) | 9.6 (0.9) | 5 | D |
| 3. Open data platforms should be preferred for big data related to RMDs | 8.7 (1.2) | 5 | D |
| 4. Privacy by design must be applied to the collection, processing, storage, analysis and interpretation of big data | 9.6 (0.5) | 4 | C |
| 5. The collection, processing, storage, analysis and interpretation of big data should be underpinned by interdisciplinary collaboration, including biomedical/health/life scientists, computational and/or data scientists, relevant clinicians/health professionals and patients | 9.7 (0.6) | 4 | C |
| 6. The methods used to analyse big data must be reported explicitly and transparently in scientific publications | 10 (0) | 4 | C |
| 7. Benchmarking of computational methods for big data used in RMD research should be encouraged | 9.4 (1.2) | 5 | D |
| 8. Before implementation, conclusions and/or models drawn from big data should be independently validated | 9.1 (0.7) | 4 | C |
| 9. Researchers using big data should proactively consider the implementation of findings in clinical practice | 9.3 (0.8) | 5 | D |
| 10. Interdisciplinary training on big data methods in RMDs for clinicians/health professionals/health and life scientists and data scientists must be encouraged | 9.7 (0.6) | 5 | D |
Numbers in the column ‘LoA’ indicate the mean and SD (in parentheses) of the LoA, as well as the mean agreement of the 14 task force members on a 0–10 scale. LoE and strength based on the Oxford Centre for Evidence-Based Medicine classification, with ‘Level 1’ corresponding to meta-analysis or randomised controlled trials (RCTs) or high-quality RCTs; ‘Level 2’ to lesser quality RCT or prospective comparative studies; ‘Level 3’ to case–control studies or retrospective studies; ‘Level 4’ to case series without the use of comparison or control groups; ‘Level 5’ to case reports or expert opinion.27
LoA, level of agreement; LoE, level of evidence; RMD, rheumatic and musculoskeletal disorder; SoR, strength of recommendation.