Table 4.
Responses to statements included in the six domains which sought agreed approaches to using big data in obesity research
| Round 1 (n = 36) | Round 2 (n = 29) | Round 3 (n = 26) | ||||
|---|---|---|---|---|---|---|
| Agree % | Disagree % | Agree % | Disagree % | Agree % | Disagree % | |
| Data Acquisition | ||||||
| 1. There is not equal access to big datasets for all academic researchers | 97.1% | 2.9% | 96.6% | 3.4% | 100.0% | 0.0% |
| 2. There is not equal access to big datasets across academic institutions or non-academic researchers | 97.1% | 3.0% | 96.6% | 3.4% | 100.0% | 0.0% |
| 3. I don’t know what big data are available to use for research purposes | 58.3% | 41.7% | 75.9% | 24.1% | 76.9% | 23.1% |
| 4. I don’t know how to access big data for research purposes | 47.2% | 52.8% | 48.3% | 51.7% | 57.7% | 42.3% |
| 5. Accessing big data for research purposes takes too long | 75.0% | 25.0% | 95.5% | 4.5% | 95.2% | 4.8% |
| 6. Timescales for access to big data limit their utility for obesity research | 55.2% | 44.8% | 72.0% | 28.0% | 73.9% | 26.1% |
| 7. Negotiating access to big data for obesity research is a challenge | 94.1% | 5.9% | 96.6% | 3.4% | 96.2% | 3.8% |
| 8. Access to big data should be provided via a third party centre/organisation that is independent both from the data owner and the researcher | 76.0%a | 24.0%a | 83.3% | 16.7% | 82.6% | 17.4% |
| 9. Third party organisations (i.e. those outside of a university) should be responsible for promoting the awareness of big data for use in obesity research | 46.2% | 53.8% | 20.8% | 79.2% | 25.0% | 75.0% |
| 10. It is the responsibility of data owners to make their data available | 65.7% | 34.3% | 69.0% | 31.0% | 73.1% | 26.9% |
| 11. Data owners are responsible for making others aware of the availability of their data | 48.5% | 51.5% | 35.7% | 64.3% | 36.0% | 64.0% |
| 12. It is the responsibility of individual research institutions to identify and negotiate access to big data sources | 56.7% | 43.3% | 63.0% | 37.0% | 75.0% | 25.0% |
| 13. The cost attached to the use of big data is a major barrier to its use | 62.1% | 37.9% | 79.2% | 20.8% | 81.0% | 19.0% |
| 14. Data protection regulations unduly restrict the use of big data in obesity research | – | – | 50.0% | 50.0% | 42.1% | 57.9% |
| 15. Government legislation is needed to encourage commercial organisations to share their data for obesity research | – | – | 80.8% | 19.2% | 84.0% | 16.0% |
| 16. Big data should be made available via third party organisations who should be responsible for protecting both commercially sensitive and individually sensitive data | – | – | 83.3% | 16.7% | 87.0% | 13.0% |
| Ethics | ||||||
| 1. It is unethical to use big data in obesity research when consent has not been obtained for this purpose | 12.9% | 87.1% | 11.1% | 88.9% | 7.7% | 92.3% |
| 2. Consent is a major ethical challenge for big data in obesity research | 77.4% | 22.6% | 85.2% | 14.8% | 84.0% | 16.0% |
| 3. Big data from commercial sources is a potential conflict of interest | 64.7% | 35.3% | 78.6% | 21.4% | 80.8% | 19.2% |
| 4. Ethical processes need reviewing in light of using big data in obesity research | 94.3% | 5.7% | 96.6% | 3.4% | 96.2% | 3.8% |
| 5. Ethical processes unduly restrict the use of big data for obesity research | 46.4% | 53.6% | 36.4% | 63.6% | 30.0% | 70.0% |
| 6. There are high confidentially risks when using big data for obesity research | 38.2% | 61.8% | 26.9% | 73.1% | 20.8% | 79.2% |
| 7. It is the responsibility of individual research institutions to ensure that big data is used ethically | 94.4% | 5.6% | 100.0% | 0.0% | 100.0% | 0.0% |
| 8. It is the responsibility of individual researchers to ensure that big data is used ethically | 97.2% | 2.8% | 100.0% | 0.0% | 100.0% | 0.0% |
| 9. It is the responsibility of data owners to ensure that big data is used ethically | 94.4% | 5.6% | 93.1% | 6.9% | 92.3% | 7.7% |
| 10. It is unethical of commercial companies to withhold big data sets that could be used to identify determinants of obesity and opportunities for intervention | 48.5% | 51.5% | 39.9% | 60.7% | 38.5% | 61.5% |
| 11. Using big data for obesity research doesn’t cause harm because no further contact with individuals or communities is made | 58.6% | 41.4% | 73.9% | 26.1% | 76.2% | 23.8% |
| 12. An ethical framework is required to review big data research proposals through formal research processes | 93.9% | 6.1% | 93.1% | 6.9% | 96.2% | 3.8% |
| 13. An ethical framework should be developed by independent bodies with no conflicts of interest | 79.4% | 20.6% | 86.2% | 13.8% | 92.3% | 7.7% |
| 14. Ethical processes should distinguish between open data already in the public domain and secondary data not already in the public domain, which may contain both commercially and individually sensitive data | – | – | 92.9% | 7.1% | 96.0% | 4.0% |
| 15. It is unethical NOT to use big data where it is available, even when informed consent has not been provided, if it will help address obesity | – | – | 30.4% | 69.6% | 14.3% | 85.7% |
| Data Governance | ||||||
| 1. The data governance requirements associated with using big data in obesity research are clear | 17.2% | 82.8% | 16.0% | 84.0% | 16.7% | 83.3% |
| 2. Data governance processes are clear for data controllers | 34.8%a | 65.2%a | 13.6% | 86.4% | 15.0% | 85.0% |
| 3. Data governance processes are clear for researchers | 25.8% | 74.2% | 12.0% | 88.0% | 12.0% | 88.0% |
| 4. Data governance processes are clear for data owners | 20.8%a | 79.2%a | 13.6% | 86.4% | 15.8% | 84.2% |
| 5. Ownership of big data can be ambiguous (e.g. for wearables/activity tracking technology the owner could be taken to be the organisation who collates/manages the data, or the individual people the data relates to) | 94.3% | 5.7% | 96.6% | 3.4% | 96.2% | 3.8% |
| Training and Infrastructure | ||||||
| 1. Big data requires novel/non-traditional analysis techniques | 80.0% | 20.0% | 92.9% | 7.1% | 96.0% | 4.0% |
| 2. Researchers need specialist training to link big data | 85.3% | 14.7% | 92.9% | 7.1% | 92.0% | 8.0% |
| 3. Researchers need specialist training to manage big data | 88.6% | 11.4% | 89.3% | 10.7% | 92.0% | 8.0% |
| 4. Researchers need specialist training to analyse big data | 83.3% | 16.7% | 89.7% | 10.3% | 88.5% | 11.5% |
| 5. There is insufficient training available to me, regarding the handling of big data and analysis | 59.4% | 40.6% | 61.5% | 38.5% | 59.1% | 40.9% |
| 6. The cost of training courses in big data analysis techniques prevents me from using these datasets | 23.3% | 76.7% | 19.2% | 80.8% | 17.4% | 82.6% |
| 7. My institution has limited equipment/systems necessary for handling big data (i.e. computer memory, secure networked systems etc.) | 41.9% | 58.1% | 37.0% | 63.0% | 37.5% | 62.5% |
| 8. It is the responsibility of individual universities to improve their training and infrastructure to use big data in obesity research | 80.6% | 19.4% | 93.1% | 6.9% | 88.5% | 11.5% |
| 9. It is the responsibility of professional organisations, including funding organisations, to provide more training around big data | 82.9% | 17.1% | 86.2% | 13.8% | 88.5% | 11.5% |
| 10. The time involved in preparing big datasets for analysis prevents me from using these datasets | 40.0% | 60.0% | 48.3% | 51.7% | 48.0% | 52.0% |
| 11. There are no training or infrastructure issues that prevent me from using big data for obesity research | 41.2% | 58.8% | 25.9% | 74.1% | 20.8% | 79.2% |
| 12. Collaboration that draws on varied skill sets is needed to appropriately handle big data in obesity research | – | – | 93.1% | 6.9% | 92.3% | 7.7% |
| Reporting and Transparency | ||||||
| 1. The provenance (source and date of collection) of big data is adequately reported in peer-reviewed literature | 25.0% | 75.0% | 12.5% | 87.5% | 4.2% | 95.8% |
| 2. The methods originally used to collect big data are adequately reported in peer-reviewed literature | 29.4% | 70.6% | 7.1% | 92.9% | 7.7% | 92.3% |
| 3. Procedures used to clean and process (e.g. re-code) big data are adequately reported in peer-reviewed literature | 8.6% | 91.4% | 7.1% | 92.9% | 8.0% | 92.0% |
| 4. The content of big data sources are adequately reported in peer-reviewed literature | 20.6% | 79.4% | 7.4% | 92.6% | 12.0% | 88.0% |
| 5. The processes used to link big data sources (e.g. geocoding techniques) are adequately reported in peer-reviewed literature | 19.4% | 80.6% | 11.1% | 88.9% | 8.3% | 91.7% |
| 6. Inadequate reporting of big data and associated methods in peer-reviewed literature means study findings cannot be usefully interpreted | 65.7% | 34.3% | 78.6% | 21.4% | 84.6% | 15.4% |
| 7. The costs associated with obtaining big data should be reported in peer-reviewed literature | 51.6% | 48.4% | 51.9% | 48.1% | 62.5% | 37.5% |
| 8. To improve big data related obesity research, standardised reporting frameworks are required | 84.8% | 15.2% | 89.3% | 10.7% | 92.3% | 7.7% |
| 9. Academic journals have a responsibility to enforce the use of reporting frameworks for big data | 82.9% | 17.1% | 86.2% | 13.8% | 92.3% | 7.7% |
| 10. Where contractual restrictions exist around the reporting of data, these should be noted when disseminating research findings | – | – | 100.0% | 0.0% | 100.0% | 0.0% |
| 11. Reporting needs to be independent of the data owner to reduce potential conflicts of interest | – | – | 72.0% | 28.0% | 79.2% | 20.8% |
| Quality and Inference | ||||||
| 1. Big data from commercial organisations results in an increased risk of bias | 58.8% | 41.2% | 73.1% | 26.9% | 80.0% | 20.0% |
| 2. Standardised quality checks of the data [i.e. how data was collected, missing data] are required from the data provider | 91.4% | 8.6% | 89.3% | 10.7% | 96.2% | 3.8% |
| 3. Big data should be used irrespective of quality in obesity research | 19.4% | 80.6% | 13.8% | 86.2% | 11.5% | 88.5% |
| 4. It is important to acknowledge methodological limitations of big data used in obesity research | 100.0% | 0.0% | 93.1% | 6.9% | 100.0% | 0.0% |
| 5. Statistically significant results need to be interpreted with caution when using big datasets in obesity research | 91.2% | 8.8% | 96.4% | 3.6% | 96.0% | 4.0% |
| 6. Outputs from research using big data are rarely misinterpreted | 11.1% | 88.9% | 8.3% | 91.7% | 9.1% | 90.9% |
| 7. There is an over reliance on big data in obesity research despite its potential bias | 17.2% | 82.8% | 12.0% | 88.0% | 16.7% | 83.3% |
| 8. The emergence of big data has negatively impacted the use of traditional data sources | 20.0% | 80.0% | 14.3% | 85.7% | 16.7% | 83.3% |
| 9. Big data is having an unhealthy steer on the obesity-related research agenda | 13.8% | 86.2% | 14.3% | 85.7% | 15.4% | 84.6% |
| 10. Researchers have a responsibility to ensure that their results are correctly interpreted in view of any limitations | 100.0% | 0.0% | 100.0% | 0.0% | 100.0% | 0.0% |
| 11. Big data obesity research should always consider inequalities in health or health behaviours as a measure of quality | 57.6% | 42.4% | 69.2% | 30.8% | 73.9% | 26.1% |
Note: Bold % denotes that 70% consensus was achieved
aProportion of ‘don’t know’ responses to this statement exceeded 30%