Welcome back to another post from my personal biomedical data science blog, [VS]Codes! In today’s post, I will provide a brief overview of the concept of “learning health systems” and how this model is shifting the way that healthcare is approached today.
Translational informatics: advancing medical care from “bench to bedside”
The discipline of “translational informatics” focuses on the translation of data generated from research-based biomedical endeavors to applications in the clinic, allowing for the improvement of disease diagnosis, staging, prognosis, and treatment. Translational research will typically involve the development of machine learning methods to integrate multimodal, heterogeneous biomedical data for clinical decision support. Ultimately, translational informatics enables the study of increasingly large bodies of biomedical data to inform predictive, preventive, and personalized health applications. Translational informatics can also be deeply tied to the concept of “bench to bedside” biomedical discovery, where the results of research conducted in the laboratory are directly used to create actionable innovations for real-world healthcare settings.
Learning Health Systems: implementing the translational medicine paradigm
Learning Health Systems are intrinsically linked to the concepts of “bench to bedside” care and translational informatics. A Learning Health System (LHS) refers to a healthcare model that makes use of continual data collection and analysis for the improvement of patient care. The key idea driving LHSs is the cyclical nature of advancements in research and clinical care: real-time data coming from routine care and patient experiences inform new directions in biomedical research, improving the way that clinical practice is performed. This continuous learning feedback loop is crucial for the improvement of patient outcomes and longer-term healthcare innovation.
As defined by the National Academy of Medicine (NAM), the four key elements of LHSs are:
Generation, application, and improvement of scientific knowledge
Organizational infrastructure to support the engagement of patient communities, healthcare professionals, and researchers for the identification of evidence gaps between biomedical research and clinical care
Deployment of computational technologies and informatics approaches to organize and leverage large-scale electronic health data for use in research
Quality improvement at the point of care for each patient using new knowledge generated by research
Some examples of downstream applications of LHSs include:
Cohort development for clinical trials and research studies through the collection of patients with similar attributes to one another
Identification of both improved as well as sub-optimal examples of patient care and treatments as compared to standardized benchmarks
The creation of patient risk models for adverse events and outcomes
Clinical Decision Support (CDS) systems to recommend personalized treatment options
Automation of routine care processes
Surveillance monitoring for disease outbreaks and other treatment issues and complications
Benefits and challenges of Learning Health Systems
Key benefits of LHSs include better patient outcomes through personalized care and faster implementation of data-driven research advancements, as well as lowered healthcare costs through the implementation of improved processes and the reduction of ineffective treatments.
However, numerous challenges have also hindered the wide-spread adoption of LHS principles globally. Some of the biggest obstacles to the broader implementation of these guidelines include:
- Maintaining data interoperability and security: multimodal, heterogeneous data are often siloed across healthcare systems, making it challenging to integrate data across departments and organizations. Furthermore, these data typically do not follow consistent common data standards, further complicating the process for data integration. LHSs also need to ensure that a system for data democratization is in place, including appropriate data access rights as well as ethical data stewardship that promotes not only data sharing but also patient privacy.
- Overcoming cultural resistance and realigning healthcare incentives: Setting up an LHS requires significant investments in technology infrastructure, data management systems, and staff training. Healthcare organizations are typically slow to adopt such new practices, particularly when they require drastic changes in traditional workflows. Instead, gradual shifts must take place to bring researchers and healthcare providers to alignment on priorities and timelines. Overall healthcare incentives must be shifted from volume to value, and appropriate performance metrics will need to be developed to hold clinicians and their teams accountable toward patient care.
- Ensuring data quality and minimizing bias: The mantra of “garbage in, garbage out” is particularly pertinent in the LHS model of continuous data generation and ingestion - if the data used to train the healthcare system are biased, incomplete, or inaccurate, then the overall reliability and applicability of the system will be drastically reduced.
- Keeping pace with technological advancement and ensuring proper model validation: Rapid advancements in artificial intelligence and data analytics mean that healthcare organizations must constantly stay abreast of new tools and methodologies while continuing to fulfill their current care duties. Given the meteoric changes in the field of information technology, it is essential that a proper system is put in place to safely incorporate the latest shifts in technology into the workings of each LHS. Furthermore, appropriate global standards and metrics must be developed and agreed upon to help determine the efficacy of changes introduced to each LHS.
Looking toward the future: an iterative process
While many healthcare systems have begun to incorporate elements of LHSs into their workflows, the complete LHS is still very much an aspirational model. Becoming a complete LHS is an iterative process that requires a step-by-step cultural shift toward the effective use of data across both clinical and research settings. Future advancements toward the LHS model will necessitate progress in multiple categories, including technology, regulatory policies and funding, and the direct involvement and engagement of patients.
References
Johns Hopkins Berman Institute of Bioethics - What is a Learning Health System?
Agency for Healthcare Research and Quality - About Learning Health Systems
McLachlan et al. 2019 - LAGOS: learning health systems and how they can integrate with patient care
Stanford Medicine Center for Biomedical Informatics Research - Translational Informatics