Privacy-preserving monitoring and benchmarking of antibiotic prescriptions – NSE

Antibiotic resistance is an increasing worldwide public health problem that requires immediate attention. Misuse and overuse of antibiotics is one of the main drivers for antibiotic resistance. There are national and international initiatives that aim to reduce antibiotic resistance. Learning from electronic health record (EHR) data generated in routine clinical care at general practitioner (GP) offices could be a major driver of quality improvement. Feedback on clinical performance of a clinician, viewed in comparison with peers, has been shown to be an effective instigator of change in prescribing behaviour.

The reuse of health data is guided by a variety of controls from both legal and ethical perspectives, such as privacy rights, data protection regulations, and duties of confidentiality. Patient consent is often exempted for reuse of health data for public health and quality improvement purposes. However, clinicians are often reluctant to pursue it due to the privacy concerns of their patients, themselves, and their institutions. The challenge is to develop an infrastructure that can securely aggregate distributed data while preserving the privacy of patients and individual institutions.

The demonstrator

The demonstrator developed by the Norwegian Centre for E-health Research (NSE) enables clinicians to compare their antibiotic prescriptions with their peers across multiple healthcare institutions, which is especially more important for smaller healthcare institutions such as GP offices where few peers within an institution are available to make comparisons. Although the current focus is on antibiotic prescriptions of GPs, the techniques implemented are generic and can be applied for other clinical domains and specialties.

Feedback provided to a clinician contains statistics on the antibiotic prescriptions of the clinician and aggregated data from multiple health institutions (Figure 1), while the collection of sensitive data in any central location is avoided. Thus, individual healthcare institutions maintain control of their data while they contribute to aggregate statistics. Aggregated statistics are computed from the combined data of GP offices using Emnet, which is a privacy-preserving distributed data mining tool implemented in the project. However, personal data on antibiotics prescribing by any of the individual clinicians is computed within their institution and is only available to the clinician.

Impact

In many countries, GPs diagnose and treat a major proportion of patients and are responsible for the majority of antibiotic prescriptions for human use. For example, Norwegian GPs are responsible for approximately 80% of antibiotic prescriptions for human use. Therefore, behavioural changes by GPs may have a significant effect on reducing inappropriate antibiotic prescriptions. A privacy-preserving framework for aggregate analytics like ASCLEPIOS addresses effectively the several risks that arise from inefficient data control, including legal liability, malpractice lawsuits, intrusive marketing and inference or disclosure of income, performance, or competitive data, which consequently affect the willingness of institutions and individual practitioners to participate in data sharing initiatives.