Erratum to: Making sense of big data in health research: towards an EU action plan

What are the bottlenecks to health data sharing in Switzerland? An interview study

BACKGROUND

While health data sharing for research purposes is strongly supported in principle, it can be challenging to implement in practice. Little is known about the actual bottlenecks to health data sharing in Switzerland.

AIMS OF THE STUDY

This study aimed to assess the obstacles to Swiss health data sharing, including legal, ethical and logistical bottlenecks.

METHODS

We identified 37 key stakeholders in data sharing via the Swiss Personalised Health Network ecosystem, defined as being an expert on sharing sensitive health data for research purposes at a Swiss university hospital (or a Swiss disease cohort) or being a stakeholder in data sharing at a public or private institution that uses such data. We conducted semi-structured interviews, which were transcribed, translated when necessary, and de-identified. The entire research team discussed the transcripts and notes taken during each interview before an inductive coding process occurred.

RESULTS

Eleven semi-structured interviews were conducted (primarily in English) with 17 individuals representing lawyers, data protection officers, ethics committee members, scientists, project managers, bioinformaticians, clinical trials unit members, and biobank stakeholders. Most respondents felt that it was not the actual data transfer that was the bottleneck but rather the processes and systems around it, which were considered time-intensive and confusing. The templates developed by the Swiss Personalised Health Network and the Swiss General Consent process were generally felt to have streamlined processes significantly. However, these logistics and data quality issues remain practical bottlenecks in Swiss health data sharing. Areas of legal uncertainty include privacy laws when sharing data internationally, questions of "who owns the data", inconsistencies created because the Swiss general consent is perceived as being implemented differently across different institutions, and definitions and operationalisation of anonymisation and pseudo-anonymisation. Many participants desired to create a "culture of data sharing" and to recognise that data sharing is a process with many steps, not an event, that requires sustainability efforts and personnel. Some participants also stressed a desire to move away from data sharing and the current privacy focus towards processes that facilitate data access.

CONCLUSIONS

Facilitating a data access culture in Switzerland may require legal clarifications, further education about the process and resources to support data sharing, and further investment in sustainable infrastructureby funders and institutions.

The future regulation of artificial intelligence systems in healthcare services and medical research in the European Union

Despite its promising future, the application of artificial intelligence (AI) and automated decision-making in healthcare services and medical research faces several legal and ethical hurdles. The European Union (EU) is tackling these issues with the existing legal framework and drafting new regulations, such as the proposed AI Act. The EU General Data Protection Regulation (GDPR) partly regulates AI systems, with rules on processing personal data and protecting data subjects against solely automated decision-making. In healthcare services, (automated) decisions are made more frequently and rapidly. However, medical research focuses on innovation and efficacy, with less direct decisions on individuals. Therefore, the GDPR’s restrictions on solely automated decision-making apply mainly to healthcare services, and the rights of patients and research participants may significantly differ. The proposed AI Act introduced a risk-based approach to AI systems based on the principles of ethical AI. We analysed the complex connection between the GDPR and AI Act, highlighting the main issues and finding ways to harmonise the principles of data protection and ethical AI. The proposed AI Act may complement the GDPR in healthcare services and medical research. Although several years may pass before the AI Act comes into force, many of its goals will be realised before that.

Occupational Hearing Loss for Platinum Miners in South Africa: A Case Study of Data Sharing Practices and Ethical Challenges in the Mining Industry

BACKGROUND

The relevant legislation ensures confidentiality and has paved the way for data handling and sharing. However, the industry remains uncertain regarding big data handling and sharing practices for improved healthcare delivery and medical research.

METHODS

A semi-qualitative cross-sectional study was used which entailed analysing miners' personal health records from 2014 to 2018. Data were accessed from the audiometry medical surveillance database (n = 480), the hearing screening database (n = 24,321), and the occupational hygiene database (n = 15,769). Ethical principles were applied to demonstrate big data protection and sharing.

RESULTS

Some audiometry screening and occupational hygiene records were incomplete and/or inaccurate (N = 4675). The database containing medical disease and treatment records could not be accessed. Ethical challenges included a lack of clarity regarding permission rights when sharing big data, and no policy governing the divulgence of miners' personal and medical records for research.

CONCLUSION

This case study illustrates how research can be effectively, although not maliciously, obstructed by the strict protection of employee medical data. Clearly communicated company policies should be developed for the sharing of workers' records in the mining industry to improve HCPs.

Medical Informatics Platform (MIP): A Pilot Study Across Clinical Italian Cohorts

Introduction: With the shift of research focus to personalized medicine in Alzheimer's Dementia (AD), there is an urgent need for tools that are capable of quantifying a patient's risk using diagnostic biomarkers. The Medical Informatics Platform (MIP) is a distributed e-infrastructure federating large amounts of data coupled with machine-learning (ML) algorithms and statistical models to define the biological signature of the disease. The present study assessed (i) the accuracy of two ML algorithms, i.e., supervised Gradient Boosting (GB) and semi-unsupervised 3C strategy (Categorize, Cluster, Classify-CCC) implemented in the MIP and (ii) their contribution over the standard diagnostic workup. Methods: We examined individuals coming from the MIP installed across 3 Italian memory clinics, including subjects with Normal Cognition (CN, n = 432), Mild Cognitive Impairment (MCI, n = 456), and AD (n = 451). The GB classifier was applied to best discriminate the three diagnostic classes in 1,339 subjects, and the CCC strategy was used to refine the classical disease categories. Four dementia experts provided their diagnostic confidence (DC) of MCI conversion on an independent cohort of 38 patients. DC was based on clinical, neuropsychological, CSF, and structural MRI information and again with addition of the outcome from the MIP tools. Results: The GB algorithm provided a classification accuracy of 85% in a nested 10-fold cross-validation for CN vs. MCI vs. AD discrimination. Accuracy increased to 95% in the holdout validation, with the omission of each Italian clinical cohort out in turn. CCC identified five homogeneous clusters of subjects and 36 biomarkers that represented the disease fingerprint. In the DC assessment, CCC defined six clusters in the MCI population used to train the algorithm and 29 biomarkers to improve patients staging. GB and CCC showed a significant impact, evaluated as +5.99% of increment on physicians' DC. The influence of MIP on DC was rated from "slight" to "significant" in 80% of the cases. Discussion: GB provided fair results in classification of CN, MCI, and AD. CCC identified homogeneous and promising classes of subjects via its semi-unsupervised approach. We measured the effect of the MIP on the physician's DC. Our results pave the way for the establishment of a new paradigm for ML discrimination of patients who will or will not convert to AD, a clinical priority for neurology.

Big data analytics in medical engineering and healthcare: methods, advances and challenges

Big data analytics are gaining popularity in medical engineering and healthcare use cases. Stakeholders are finding big data analytics reduce medical costs and personalise medical services for each individual patient. Big data analytics can be used in large-scale genetics studies, public health, personalised and precision medicine, new drug development, etc. The introduction of the types, sources, and features of big data in healthcare as well as the applications and benefits of big data and big data analytics in healthcare is key to understanding healthcare big data and will be discussed in this article. Major methods, platforms and tools of big data analytics in medical engineering and healthcare are also presented. Advances and technology progress of big data analytics in healthcare are introduced, which includes artificial intelligence (AI) with big data, infrastructure and cloud computing, advanced computation and data processing, privacy and cybersecurity, health economic outcomes and technology management, and smart healthcare with sensing, wearable devices and Internet of things (IoT). Current challenges of dealing with big data and big data analytics in medical engineering and healthcare as well as future work are also presented.

Development of Supervised Learning Predictive Models for Highly Non-linear Biological, Biomedical, and General Datasets

In highly non-linear datasets, attributes or features do not allow readily finding visual patterns for identifying common underlying behaviors. Therefore, it is not possible to achieve classification or regression using linear or mildly non-linear hyperspace partition functions. Hence, supervised learning models based on the application of most existing algorithms are limited, and their performance metrics are low. Linear transformations of variables, such as principal components analysis, cannot avoid the problem, and even models based on artificial neural networks and deep learning are unable to improve the metrics. Sometimes, even when features allow classification or regression in reported cases, performance metrics of supervised learning algorithms remain unsatisfyingly low. This problem is recurrent in many areas of study as, per example, the clinical, biotechnological, and protein engineering areas, where many of the attributes are correlated in an unknown and very non-linear fashion or are categorical and difficult to relate to a target response variable. In such areas, being able to create predictive models would dramatically impact the quality of their outcomes, generating an immediate added value for both the scientific and general public. In this manuscript, we present RV-Clustering, a library of unsupervised learning algorithms, and a new methodology designed to find optimum partitions within highly non-linear datasets that allow deconvoluting variables and notoriously improving performance metrics in supervised learning classification or regression models. The partitions obtained are statistically cross-validated, ensuring correct representativity and no over-fitting. We have successfully tested RV-Clustering in several highly non-linear datasets with different origins. The approach herein proposed has generated classification and regression models with high-performance metrics, which further supports its ability to generate predictive models for highly non-linear datasets. Advantageously, the method does not require significant human input, which guarantees a higher usability in the biological, biomedical, and protein engineering community with no specific knowledge in the machine learning area.

A Scalable Data Access Layer to Manage Structured Heterogeneous Biomedical Data

This work presents a scalable data access layer, called PyEHR, designed to support the implementation of data management systems for secondary use of structured heterogeneous biomedical and clinical data. PyEHR adopts the openEHR's formalisms to guarantee the decoupling of data descriptions from implementation details and exploits structure indexing to accelerate searches. Data persistence is guaranteed by a driver layer with a common driver interface. Interfaces for two NoSQL Database Management Systems are already implemented: MongoDB and Elasticsearch. We evaluated the scalability of PyEHR experimentally through two types of tests, called "Constant Load" and "Constant Number of Records", with queries of increasing complexity on synthetic datasets of ten million records each, containing very complex openEHR archetype structures, distributed on up to ten computing nodes.