Contextual Anonymization for Secondary Use of Big Data in Biomedical Research: Proposal for an Anonymization Matrix

Rumbold, John; Pierscionek, Barbara K.

doi:10.2196/medinform.7096

Cited by 11 publications

(4 citation statements)

References 48 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The reasons may be related to issues facing many data scientists and researchers in this area: the limited availability of or access to data or the readiness of health care institutions to share data. Privacy concerns over personal data, and in particular health care data, means that although the data exist, they are deemed too sensitive for public release [ 1 ], even for research purposes.…”

Section: Introductionmentioning

confidence: 99%

Reliability of Supervised Machine Learning Using Synthetic Data in Health Care: Model to Preserve Privacy for Data Sharing

Rankin¹,

Black²,

Bond³

et al. 2020

JMIR Med Inform

112

View full text Add to dashboard Cite

Background The exploitation of synthetic data in health care is at an early stage. Synthetic data could unlock the potential within health care datasets that are too sensitive for release. Several synthetic data generators have been developed to date; however, studies evaluating their efficacy and generalizability are scarce. Objective This work sets out to understand the difference in performance of supervised machine learning models trained on synthetic data compared with those trained on real data. Methods A total of 19 open health datasets were selected for experimental work. Synthetic data were generated using three synthetic data generators that apply classification and regression trees, parametric, and Bayesian network approaches. Real and synthetic data were used (separately) to train five supervised machine learning models: stochastic gradient descent, decision tree, k-nearest neighbors, random forest, and support vector machine. Models were tested only on real data to determine whether a model developed by training on synthetic data can used to accurately classify new, real examples. The impact of statistical disclosure control on model performance was also assessed. Results A total of 92% of models trained on synthetic data have lower accuracy than those trained on real data. Tree-based models trained on synthetic data have deviations in accuracy from models trained on real data of 0.177 (18%) to 0.193 (19%), while other models have lower deviations of 0.058 (6%) to 0.072 (7%). The winning classifier when trained and tested on real data versus models trained on synthetic data and tested on real data is the same in 26% (5/19) of cases for classification and regression tree and parametric synthetic data and in 21% (4/19) of cases for Bayesian network-generated synthetic data. Tree-based models perform best with real data and are the winning classifier in 95% (18/19) of cases. This is not the case for models trained on synthetic data. When tree-based models are not considered, the winning classifier for real and synthetic data is matched in 74% (14/19), 53% (10/19), and 68% (13/19) of cases for classification and regression tree, parametric, and Bayesian network synthetic data, respectively. Statistical disclosure control methods did not have a notable impact on data utility. Conclusions The results of this study are promising with small decreases in accuracy observed in models trained with synthetic data compared with models trained with real data, where both are tested on real data. Such deviations are expected and manageable. Tree-based classifiers have some sensitivity to synthetic data, and the underlying cause requires further investigation. This study highlights the potential of synthetic data and the need for further evaluation of their robustness. Synthetic data must ensure individual privacy and data utility are preserved in order to instill confidence in health care departments when using such data to inform policy decision-making.

show abstract

Section: Introductionmentioning

confidence: 99%

Reliability of Supervised Machine Learning Using Synthetic Data in Health Care: Model to Preserve Privacy for Data Sharing

Rankin¹,

Black²,

Bond³

et al. 2020

JMIR Med Inform

112

View full text Add to dashboard Cite

show abstract

“…Classifying data into sensitivity levels was previously recommended to limit access to data and limit the type of data shared 4. Sensitive personal data as defined in data protection legislation should include specific protections for highly stigmatized diseases, such as sexually transmitted and psychiatric illnesses 24. Current results show that in less than half of the cases, more limitations were put on sensitive data than with regular data.…”

Section: Discussionmentioning

confidence: 96%

<p>Patient Data Sharing and Confidentiality Practices of Researchers in Jordan</p>

Karasneh

Al‐Azzam

Alzoubi

et al. 2019

RMHP

View full text Add to dashboard Cite

PurposeThe main focus of this study is to assess the knowledge and practices of healthcare practitioners regarding data sharing, security, and confidentiality, with a focus on the use of health data retrieved from electronic health records (EHRs) for research purposes.MethodsA descriptive, cross-sectional, questionnaire-based survey study was conducted across all academic institutions including all researchers in the medical field in Jordan. Personal and administrative practices in data sharing were assessed through collecting data from respondents.ResultsThe response rate was 22% with an average of 10.25 years of experience in publications. Almost 60% had published at least 1 to 3 studies using EHRs. The prevalence of researchers who “Always” used antivirus software and preserved patient’s information was 75.5% and 92.2%, respectively. However, other personal security and confidentiality measures were not satisfactory. Less than half of health data used in the research was “Always” anonymised or encrypted and only around 44.0% had “Always” used sensitive data with more specificity than normal data.ConclusionConfidentiality and data sharing practices of healthcare practitioners and researchers were generally less than optimal. Efforts from healthcare providers, health institutions, and lawmakers should be put in place to protect the security and confidentiality of electronic patient data.

show abstract

“…However, setting up such networks usually takes quite some time, as common tools and policies are needed to achieve interoperability and enable the required flows of data and biosamples [ 6 7 ]. One area in which this challenge is frequently encountered is the handling of personal data and the related data protection issues, which can arise in all processing steps, from collection [ 8 ] to sharing [ 9 ] and even analysis and visualization [ 10 ].…”

Section: Introductionmentioning

confidence: 99%

A Scalable Pseudonymization Tool for Rapid Deployment in Large Biomedical Research Networks: Development and Evaluation Study

Abu Attieh,

Neves,

Guedes

et al. 2024

JMIR Med Inform

View full text Add to dashboard Cite

Background The SARS-CoV-2 pandemic has demonstrated once again that rapid collaborative research is essential for the future of biomedicine. Large research networks are needed to collect, share, and reuse data and biosamples to generate collaborative evidence. However, setting up such networks is often complex and time-consuming, as common tools and policies are needed to ensure interoperability and the required flows of data and samples, especially for handling personal data and the associated data protection issues. In biomedical research, pseudonymization detaches directly identifying details from biomedical data and biosamples and connects them using secure identifiers, the so-called pseudonyms. This protects privacy by design but allows the necessary linkage and reidentification. Objective Although pseudonymization is used in almost every biomedical study, there are currently no pseudonymization tools that can be rapidly deployed across many institutions. Moreover, using centralized services is often not possible, for example, when data are reused and consent for this type of data processing is lacking. We present the ORCHESTRA Pseudonymization Tool (OPT), developed under the umbrella of the ORCHESTRA consortium, which faced exactly these challenges when it came to rapidly establishing a large-scale research network in the context of the rapid pandemic response in Europe. Methods To overcome challenges caused by the heterogeneity of IT infrastructures across institutions, the OPT was developed based on programmable runtime environments available at practically every institution: office suites. The software is highly configurable and provides many features, from subject and biosample registration to record linkage and the printing of machine-readable codes for labeling biosample tubes. Special care has been taken to ensure that the algorithms implemented are efficient so that the OPT can be used to pseudonymize large data sets, which we demonstrate through a comprehensive evaluation. Results The OPT is available for Microsoft Office and LibreOffice, so it can be deployed on Windows, Linux, and MacOS. It provides multiuser support and is configurable to meet the needs of different types of research projects. Within the ORCHESTRA research network, the OPT has been successfully deployed at 13 institutions in 11 countries in Europe and beyond. As of June 2023, the software manages data about more than 30,000 subjects and 15,000 biosamples. Over 10,000 labels have been printed. The results of our experimental evaluation show that the OPT offers practical response times for all major functionalities, pseudonymizing 100,000 subjects in 10 seconds using Microsoft Excel and in 54 seconds using LibreOffice. Conclusions Innovative solutions are needed to make the process of establishing large research networks more efficient. The OPT, which leverages the runtime environment of common office suites, can be used to rapidly deploy pseudonymization and biosample management capabilities across research networks. The tool is highly configurable and available as open-source software.

show abstract

Contextual Anonymization for Secondary Use of Big Data in Biomedical Research: Proposal for an Anonymization Matrix

Cited by 11 publications

References 48 publications

Reliability of Supervised Machine Learning Using Synthetic Data in Health Care: Model to Preserve Privacy for Data Sharing

Reliability of Supervised Machine Learning Using Synthetic Data in Health Care: Model to Preserve Privacy for Data Sharing

<p>Patient Data Sharing and Confidentiality Practices of Researchers in Jordan</p>

A Scalable Pseudonymization Tool for Rapid Deployment in Large Biomedical Research Networks: Development and Evaluation Study

Contact Info

Product

Resources

About