Privacy-preserving federated machine learning on FAIR health data: A real-world application

Objective

This paper introduces a privacy-preserving federated machine learning (ML) architecture built upon Findable, Accessible, Interoperable, and Reusable (FAIR) health data. It aims to devise an architecture for executing classification algorithms in a federated manner, enabling collaborative model-building among health data owners without sharing their datasets.

Materials and methods

Utilizing an agent-based architecture, a privacy-preserving federated ML algorithm was developed to create a global predictive model from various local models. This involved formally defining the algorithm in two steps: data preparation and federated model training on FAIR health data and constructing the architecture with multiple components facilitating algorithm execution. The solution was validated by five healthcare organizations using their specific health datasets.

Results

Five organizations transformed their datasets into Health Level 7 Fast Healthcare Interoperability Resources via a common FAIRification workflow and software set, thereby generating FAIR datasets. Each organization deployed a Federated ML Agent within its secure network, connected to a cloud-based Federated ML Manager. System testing was conducted on a use case aiming to predict 30-day readmission risk for chronic obstructive pulmonary disease patients and the federated model achieved an accuracy rate of 87%.

Discussion

The paper demonstrated a practical application of privacy-preserving federated ML among five distinct healthcare entities, highlighting the value of FAIR health data in machine learning when utilized in a federated manner that ensures privacy protection without sharing data.

Conclusion

This solution effectively leverages FAIR datasets from multiple healthcare organizations for federated ML while safeguarding sensitive health datasets, meeting legislative privacy and security requirements.

Machine Learning-Based Prediction of Changes in the Clinical Condition of Patients With Complex Chronic Diseases: 2-Phase Pilot Prospective Single-Center Observational Study

BACKGROUND

Functional impairment is one of the most decisive prognostic factors in patients with complex chronic diseases. A more significant functional impairment indicates that the disease is progressing, which requires implementing diagnostic and therapeutic actions that stop the exacerbation of the disease.

OBJECTIVE

This study aimed to predict alterations in the clinical condition of patients with complex chronic diseases by predicting the Barthel Index (BI), to assess their clinical and functional status using an artificial intelligence model and data collected through an internet of things mobility device.

METHODS

A 2-phase pilot prospective single-center observational study was designed. During both phases, patients were recruited, and a wearable activity tracker was allocated to gather physical activity data. Patients were categorized into class A (BI≤20; total dependence), class B (2060; moderate or mild dependence, or independent). Data preprocessing and machine learning techniques were used to analyze mobility data. A decision tree was used to achieve a robust and interpretable model. To assess the quality of the predictions, several metrics including the mean absolute error, median absolute error, and root mean squared error were considered. Statistical analysis was performed using SPSS and Python for the machine learning modeling.

RESULTS

Overall, 90 patients with complex chronic diseases were included: 50 during phase 1 (class A: n=10; class B: n=20; and class C: n=20) and 40 during phase 2 (class B: n=20 and class C: n=20). Most patients (n=85, 94%) had a caregiver. The mean value of the BI was 58.31 (SD 24.5). Concerning mobility aids, 60% (n=52) of patients required no aids, whereas the others required walkers (n=18, 20%), wheelchairs (n=15, 17%), canes (n=4, 7%), and crutches (n=1, 1%). Regarding clinical complexity, 85% (n=76) met patient with polypathology criteria with a mean of 2.7 (SD 1.25) categories, 69% (n=61) met the frailty criteria, and 21% (n=19) met the patients with complex chronic diseases criteria. The most characteristic symptoms were dyspnea (n=73, 82%), chronic pain (n=63, 70%), asthenia (n=62, 68%), and anxiety (n=41, 46%). Polypharmacy was presented in 87% (n=78) of patients. The most important variables for predicting the BI were identified as the maximum step count during evening and morning periods and the absence of a mobility device. The model exhibited consistency in the median prediction error with a median absolute error close to 5 in the training, validation, and production-like test sets. The model accuracy for identifying the BI class was 91%, 88%, and 90% in the training, validation, and test sets, respectively.

CONCLUSIONS

Using commercially available mobility recording devices makes it possible to identify different mobility patterns and relate them to functional capacity in patients with polypathology according to the BI without using clinical parameters.

Mobile Health Requirements for the Occupational Health Assessment of Health Care Professionals: Delphi Study

BACKGROUND

In recent years, owing to the COVID-19 pandemic, awareness of the high level of stress among health care professionals has increased, and research in this area has intensified. Hospital staff members have historically been known to work in an environment involving high emotional demands, time pressure, and workload. Furthermore, the pandemic has increased the strain experienced by health care professionals owing to the high number of people they need to manage and, on many occasions, the limited available resources with which they must carry out their functions. These psychosocial risks are not always well dealt with by the organization or the professionals themselves. Therefore, it is necessary to have tools to assess these psychosocial risks and to optimize the management of this demand from health care professionals. Digital health, and more specifically, mobile health (mHealth), is presented as a health care modality that can contribute greatly to respond to these unmet needs.

OBJECTIVE

We aimed to analyze whether mHealth tools can provide value for the study and management of psychosocial risks in health care professionals, and assess the requirements of these tools.

METHODS

A Delphi study was carried out to determine the opinions of experts on the relevance of using mHealth tools to evaluate physiological indicators and psychosocial factors in order to assess occupational health, and specifically, stress and burnout, in health care professionals. The study included 58 experts with knowledge and experience in occupational risk prevention, psychosocial work, and health-related technology, as well as health professionals from private and public sectors.

RESULTS

Our data suggested that there is still controversy about the roles that organizations play in occupational risk prevention in general and psychosocial risks in particular. An adequate assessment of the stress levels and psychosocial factors can help improve employees' well-being. Moreover, making occupational health evaluations available to the team would positively affect employees by increasing their feelings of being taken into account by the organization. This assessment can be improved with mHealth tools that identify and quickly highlight the difficulties or problems that occur among staff and work teams. However, to achieve good adherence and participation in occupational health and safety evaluations, experts consider that it is essential to ensure the privacy of professionals and to develop feelings of being supported by their supervisors.

CONCLUSIONS

For years, mHealth has been used mainly to propose intervention programs to improve occupational health. Our research highlights the usefulness of these tools for evaluating psychosocial risks in a preliminary and essential phase of approaches to improve the health and well-being of professionals in health care settings. The most urgent requirements these tools must meet are those aimed at protecting the confidentiality and privacy of measurements.

A Data Transformation Methodology to Create Findable, Accessible, Interoperable, and Reusable Health Data: Software Design, Development, and Evaluation Study

BACKGROUND

Sharing health data is challenging because of several technical, ethical, and regulatory issues. The Findable, Accessible, Interoperable, and Reusable (FAIR) guiding principles have been conceptualized to enable data interoperability. Many studies provide implementation guidelines, assessment metrics, and software to achieve FAIR-compliant data, especially for health data sets. Health Level 7 (HL7) Fast Healthcare Interoperability Resources (FHIR) is a health data content modeling and exchange standard.

OBJECTIVE

Our goal was to devise a new methodology to extract, transform, and load existing health data sets into HL7 FHIR repositories in line with FAIR principles, develop a Data Curation Tool to implement the methodology, and evaluate it on health data sets from 2 different but complementary institutions. We aimed to increase the level of compliance with FAIR principles of existing health data sets through standardization and facilitate health data sharing by eliminating the associated technical barriers.

METHODS

Our approach automatically processes the capabilities of a given FHIR end point and directs the user while configuring mappings according to the rules enforced by FHIR profile definitions. Code system mappings can be configured for terminology translations through automatic use of FHIR resources. The validity of the created FHIR resources can be automatically checked, and the software does not allow invalid resources to be persisted. At each stage of our data transformation methodology, we used particular FHIR-based techniques so that the resulting data set could be evaluated as FAIR. We performed a data-centric evaluation of our methodology on health data sets from 2 different institutions.

RESULTS

Through an intuitive graphical user interface, users are prompted to configure the mappings into FHIR resource types with respect to the restrictions of selected profiles. Once the mappings are developed, our approach can syntactically and semantically transform existing health data sets into HL7 FHIR without loss of data utility according to our privacy-concerned criteria. In addition to the mapped resource types, behind the scenes, we create additional FHIR resources to satisfy several FAIR criteria. According to the data maturity indicators and evaluation methods of the FAIR Data Maturity Model, we achieved the maximum level (level 5) for being Findable, Accessible, and Interoperable and level 3 for being Reusable.

CONCLUSIONS

We developed and extensively evaluated our data transformation approach to unlock the value of existing health data residing in disparate data silos to make them available for sharing according to the FAIR principles. We showed that our method can successfully transform existing health data sets into HL7 FHIR without loss of data utility, and the result is FAIR in terms of the FAIR Data Maturity Model. We support institutional migration to HL7 FHIR, which not only leads to FAIR data sharing but also eases the integration with different research networks.