The learning public health system: A new paradigm for disease surveillance and prevention

The burden and impact of chronic diseases continue to remain the global health priority it has been for the last several decades.[1] While the world has not yet fully recovered from what was clearly, a devastating recent COVID-19 pandemic that took the lives of millions across nations, it clearly demonstrated that those with noncommunicable diseases (NCDs), remain the most vulnerable to COVID-19 and its potential myriad complications,[2] and that health systems around the world need to be forever vigilant against such catastrophic events in the future, with everything at their disposal. Apart from the remarkably rapid development and deployment of mRNA vaccines, the pandemic also sparked many other innovations in health care, or forced rapid implementation/expansion of existing technologies (e.g., telemedicine), and those that enabled both the detection (e.g., kits for rapid home COVID-19 testing) and the tracking of disease outbreaks in real time, and, after some latency, rapid creation of active surveillance systems by public health systems and many academic health systems throughout the world, through advances in health information technology, rapid establishment of both local, regional, national, and international registries or data systems, where very few existed, thanks to the power of the internet, advances in computer science and the global realization regarding the value of “big data.” The vital importance of monitoring the pandemic (i.e., disease surveillance) in informing resource allocation, disease management, and disease prevention all at the same time was rapidly deployed within the few months in the first wave of the pandemic. Most in health care and public health were reminded of the fundamental importance of such monitoring systems, alongside more obvious disease containment strategies, such as lockouts, masking and social distancing, screening, isolation of those with symptoms, strong emergency, robust healthcare systems’ response, key strategies for active disease management, and care of the critically ill. The idea of disease surveillance, originating from the world of communicable diseases, was formally defined in 1988 by Thacker and Berkelman,[3] at the US Centers for Disease Control and Prevention. It entails the collection, assembly, analysis, and interpretation of health data, followed by dissemination to those that need to know – (e.g., policymakers, healthcare providers, researchers, educators, and the public, with the expectation that action will result and that accumulating data will continue to be collected and monitored over time, so as to continually inform policy and action toward prevention through evidence-based interventions in health systems and through public health initiatives by public health agencies. While it is described and depicted as a cyclical process, in practice, disease surveillance is traditionally slower in pace and linear in its implementation, relying mostly on periodic surveys, screening of samples of the population, and use of administrative healthcare billing data (healthcare claims), vital records, etc., Only recently, have data derived from electronic health records, laboratories, or pharmacy databases, begun to be routinely included in some surveillance systems, beyond health services or clinical research. Reporting through traditional channels such as government reports, surveillance websites, and peer-reviewed publications tends to be the norm. It is not considered a vehicle for evidence generation, although the same data can be utilized for traditional public health or health services research. Typically, public health surveillance projects are deemed exempt from the need for institutional review board review, unless they involve direct patient contact for data collection. The concept of the learning health system (LHS) was formally enunciated in 2007 in an Institute of Medicine (IOM) report in the United States, following a workshop that discussed the need for rapid generation of evidence from ongoing clinical practice that could inform and improve patient care.[4] The IOM recognized that there was an unacceptably long latency (median of 17 years) between evidence generation (through traditional channels of peer-reviewed publications/research) and its implementation in actual practice. The age of the electronic medical record and big data, with ongoing advances in computer science and health informatics, led to the realization that with the creation of a data infrastructure, along with the establishment of a culture of and ecosystem for more rapid-cycle iterative quality improvement in health care, and the coalescence of key stakeholders (considered a fundamental requirement in LHS) in the spirit of kaizen (a Japanese business philosophy of continuous improvement), that it were possible to learn rapidly from the experience and rapidly accumulating data of patients in a given healthcare system that could inform local practice as well as identify local barriers or conditions that were influencing care, whilst also potentially creating generalizable knowledge. The cycle of LHS,[5] therefore, includes data-to-knowledge, knowledge-to-practice, and then practice-to-data. If tapped into by a trained portfolio of data and implementation scientists, can iteratively improve care, and generate more knowledge over time, thereby reducing the long time it takes to generate evidence by the traditional research enterprise. Thus, from the same data infrastructure, multiple clinical problems of interest could be tackled by different teams of stakeholders (clinicians, nursing staff, researchers, health system leadership, and patients) dedicated to their areas of interest. Over the last decade, the LHS schema, previously restricted to traditional healthcare environment, has been applied to the field of public health and the expanded notion of a Learning Public Health System (LPHS) has emerged.[6] This has the potential to be implemented locally (a group of health systems, along with public health services), regionally (a network of health systems aligned with the public health system and committed to public health priorities, perhaps partnering with the regional public health department in areas of mutual interest), nationally (federated data systems across the nation that agree to share data and can by queried by researchers at all participating institutions, without releasing their patients’ identity), or internationally, etc. LPHS can take on disease prevention projects,[7] that start by interdisciplinary stakeholders coming together, in the spirit of serving a common purpose of tackling one or more chronic conditions. In the US, there are currently a few examples of “federated” data systems include (i) the Federal Drug Agency’s – Sentinel system[8] for postmarketing surveillance of drugs, (ii) the National Institutes of Health funded National COVID-19 Collaborative (N3C),[9] and the Patient Centered Outcomes Institute’s PCORnet data[10] and LHS infrastructure. The idea here is that each health system site keeps its data locally but shares data with other sites without revealing patient identity. This allows large-scale data queries that are more likely to lead to generalizable evidence generation. The integrated Veterans Health Administration’s health system data (serving US Veterans in over 150 Veterans Health Administration (VHA) hospitals and over 1000 community-based outpatient clinics) is a functioning LHS.[11] It has a well-established corporate data warehouse and a long track record of funding and training its workforce in clinical and health services research, VA-wide cooperative studies, pragmatic clinical trials, and quality improvement efforts. While many high-income nations are increasingly adopting learning healthcare system (LHS) and learning public health system (LPHS) concepts, many challenges remain, and universal acceptance and adoption remain a distant vision. However, the concepts underlying a LHS, or a learning public health system (LPHS), are very promising indeed for health care, public health, implementation science, quality improvement, or pragmatic trials. These newer concepts if implemented thoughtfully and collaboratively, leveraging advances in computer and data science, along with traditional disciplines of biostatistics, epidemiology, and the clinical sciences, have the potential to bridge the gap between health systems and the community/public health domain, and have measurable high impact on health and well-being of communities, while preventing NCDs, slowing their progression, and addressing health equity worldwide.