Public Health Informatics and the Perioperative Physician: Looking to the Future

In August 1854, London was amid a major cholera outbreak. Leading physicians thought that the population’s waste and dirty living conditions were getting into the air and somehow making people ill. John Snow, a London physician and one of the founding fathers of both public health and anesthesiology, was convinced that something other than the air might be responsible. Cholera is an extremely virulent disease that can cause severe acute watery diarrhea, affects both children and adults, and can kill within hours if untreated. Snow used government death-registration data and house-to-house inquiries to map the victims’ residences to determine that everyone affected had a single common connection: they all had retrieved water from the local water pump at Broad Street (Figure 1).

Snow was able to gather data and make inferences for an epidemic of his time that improved public health. He supported his hypothesis with a detailed statistical analysis to show the correlation between the quality of the water source and cholera cases. The potential benefits of this epidemiologic approach are only starting to make their way into perioperative care. Anesthesiologists have made tremendous strides to decrease intraoperative mortality and morbidity from anesthesia itself and have continuously worked to improve pre- and postoperative period patient safety. However, the perioperative period can be an important gateway to track and affect public or population-level health as physicians consider longer time horizons for patients. Potential public health issues as they relate to the surgical and critically ill patients are diverse and can include the opioid crisis and chronic diseases like opioid use disorder; pandemics and infectious diseases like coronavirus disease 2019 (COVID-19); emergencies, injuries, environmental health problems, and a host of other health threats.^2–4

Perioperative physicians will need to analyze and interpret the large amounts of data generated from public health agencies and efficiently and effectively develop informatics solutions to population-based issues. Public health informatics leverages concepts in data science to capture, manage, and analyze existing data to improve perioperative and population-level health outcomes (Figures 2 and 3). It is the “systematic application of information and computer science and technology to Public Health (PH) practice, research, and learning and public health data collection” in the United States.⁵ Regardless of the public health issue, a standard approach is often taken by public health informaticians who apply 4 general questions (Table 1)^6,7:

1. Surveillance: “What is the problem?” Surveillance systems are used to monitor public health events and behaviors and identify the problem occurring among a population.
2. Risk Factor Identification: “What is the cause of the problem?” Are there factors that might make certain populations more susceptible to disease (eg, something in the environment or certain behaviors that people practice)?
3. Intervention or Evaluation: “What intervention works to address the problem?” Interventions that have worked in the past to address a problem are evaluated and determine whether a proposed intervention makes sense with the affected population.
4. Implementation: “How can we implement the intervention? Will a proposed intervention work given the available resources and what is known about the affected population?”

The steps are often iterative and rely on a variety of informatics sources (Figure 4). Surveillance sources within the United States include the Behavioral Risk Factor Surveillance System (BRFSS) and the Electronic Surveillance System for the Early Notification of Community-based Epidemics (ESSENCE) that play essential roles to track worrisome public health concerns early.^8,9 The electronic health record (EHR) remains a cornerstone for the subsequent Protected Health Information (PHI) process steps that include risk factor identification, evaluation, and implementation. The Health Information Technology for Economic and Clinical Health (HITECH) Act drove the near-universal adoption of certified EHRs.^10–13 EHR data are routinely used to aggregate diagnoses and treatment data at the population level (national, state, and county levels). Historically, EHR data aggregation is for billing and claims, but the HITECH Act required health care systems to share data with public health agencies. The extent to which this succeeded during the COVID pandemic still needs to be determined, but recent research indicates that significant data and informatics skills gaps persist among the broader public health workforce.¹⁴ Perhaps the greatest opportunity for our specialties to lead and influence policy is to educate and support informaticists with surgical or critical care domain expertise, individuals who are exceedingly rare in public health agencies at the present time.¹⁵

Over the next decade, modernization of data systems and Information Technology (IT) practices to improve US public health are critical steps that will require the same, if not greater, herculean effort and focus that HITECH fostered in the digitization of care delivery over the past decade. United States public health is currently managed by a complex patchwork that includes federal agencies that include the Department of Health and Human Services, the United States Public Health Service, and the Centers for Disease Control; state, tribal, county, and territorial health departments; and large health care delivery systems. The quality of data exchange between these various entities is often slow, hierarchical, and impeded by different data standards across geographic areas, if there is a digital exchange at all.

The Department of Health and Human Services has proposed a framework for the next generation of public health informatics systems: Public Health 3.0.^16–18 Public Health 3.0 is meant to advance the practice of public health and outcomes for populations with a focus on Social Determinants of Health (SDOH) and data modernization. SDOH reflects the environments and situations where individuals live, work, study and that can impact an enormous range of health outcomes and risks.^19–21 SDOH includes numerous factors that include race, economic status, education, housing, and environmental risks (ie, pollution) that can affect population-level health risks.^19–21 SDOH measures include zip codes, incomes, education, access to health care, housing, and nutrition. Incorporating SDOH data into public health initiatives can help tailor to local community needs and promote effective prevention, treatment, and management strategies. Data modernization initiatives comprise early warning system enhancements for real-time, linked data on emerging health threats; early pandemic solution identification; skilled workforce development; and most importantly, common nationwide data standards for data access and exchange.

In this review, the authors present their views on how public health informatics will intersect with the practice of perioperative medicine, anesthesia, and critical care over the next 25 years. We examine the relevance of public health informatics to perioperative physicians, regardless of subspecialty, through the lens of the opioid and COVID-19 epidemics. We further propose that the perioperative informatician will need to incorporate public health informatics into their practices and provide timely, reliable, granular (ie, subcounty), and actionable surgical and intensive care unit (ICU) patient data, with clear metrics, to document success in public health practice.

RELEVANCE OF PUBLIC HEALTH INFORMATICS TO PERIOPERATIVE INFORMATICIANS

The current exponential growth in population-level data requires that we continue to learn and improve our ability to analyze data to apply these methods to population health in a meaningful fashion. This will involve the integration of advanced data science techniques and predictive modeling that uses reliable data sources. Anesthesiologists and critical care physicians will also need to better understand the methodologies behind public health informatics and identify reliable data sources. Our current day opioid and COVID-19 epidemics have clearly demonstrated their impacts on anesthesiologists and critical care physicians’ practice, and conversely, how we can impact them.^22,23

The Opioid Epidemic and Anesthesiologists

The United States’ opioid epidemic has taken hundreds of thousands of lives over the past 2 decades.^24–27 Perioperative prescription opioids are 1 potential gateway to sustained opioid use, and anesthesiologists have played important roles to prevent prolonged postoperative opioid use,⁴ decrease opioid prescriptions, influence surgeon opioid prescription practices, and limit the transition to illicit opioid use.²² Prescription opioids, especially chronic opioid use after surgical procedures, may contribute to this epidemic with up to a 3% incidence of previously naive patients continuing to use opioids months after surgery.²⁸ Furthermore, the incidence of persistent opioid use among hip and knee arthroplasty has been estimated to be between 10% and 40% after surgery.²⁹

As perioperative practitioners, we can allocate resources to address a portion of the opioid crisis. A personalized medicine approach that helps prevent and treat chronic postsurgical pain would be beneficial to reduce opioid use and direct limited resources to the patients at greatest risk and who would benefit most. The use of large datasets for predictive modeling may be a valuable approach to optimize a precision medicine-based approach to resource allocation.³⁰ Several studies have described the implementation of machine learning to predict opioid use after the acute period that follows joint arthroplasty.^31–33 Optimized data and machine learning parameters are essential to improve the predictive ability of these models. For example, ensemble-type machine learning and oversampling techniques (for unbalanced datasets) led to improved predictive models performance for persistent opioid use after lower extremity joint arthroplasty.³³ In another study of ~9000 patients who underwent total knee arthroplasty, various machine learning approaches were used, including neural networks, to develop predictive models to identify patients who might experience extended postoperative opioid use.³² Using an elastic-net penalized logistic regression model, authors developed a predictive model for persistent opioid use following total hip arthroplasty.³¹ Some notable predictors of persistent opioid use among these studies included preoperative opioid use, depression, age, duration of opioid exposure, and hospital length of stay.^31–33 Other predictive models for postoperative opioid outcomes have been developed for a variety of surgical populations including ambulatory surgery,³⁴ anterior cruciate ligament repair,³⁴ knee and hip arthroscopy,^35–37 spine surgery,^38,39 surgical patients with depression,⁴⁰ and pediatric patients.⁴¹

Informatics Aid in the Care of COVID-19 Patients

Health care systems are environments with finite resources, and the COVID-19 pandemic required implementation of a coordinated, data-informed approach tracking trends in infection rates and hospital bed capacity to optimize patient care.⁴² In response to anticipated COVID-19 surges, anesthesiologists and intensivists were involved in a key policy response by the Centers for Medicare & Medicaid Services (CMS) to delay elective surgeries and nonessential procedures. These policies were developed using data collected by public health agencies on the potential impact of COVID-19 on hospital resources; and that in turn led to a profound impact on elective surgery scheduling and ICU care.^43–47

Early in the COVID-19 pandemic, much of the virus’ pathogenesis was unknown. Data scientists leveraged large amounts of data produced by the worldwide clinical research to develop scientifically-guided health care pathways to address the health care crisis plans. The World Health Organization’s Clinical Progression Scale is a COVID-19 ordinal clinical severity score that may be used to standardize and compare clinical progression and hospital burden across a health care institution.⁴⁸ In a bi-institutional study, health care systems automated data acquisition and clinical decision aid support to calculate the severity score for COVID-19 patients.⁴⁹ Daily, the system queried data from the institution’s EHR and included demographic data, hospital length of stay, vital settings, ventilator settings, oxygen utilization, and dialysis use to produce a daily score. A time span of 13,386 patient days was analyzed and authors demonstrated successful implementation of such scoring system in 100% of inpatient days.⁴⁹ While the mortality and morbidity benefit of these systems is still pending, this initiative demonstrates the ability of perioperative physicians to study the clinical characteristics of COVID-19 to contribute data and to recruit patients to trials across informatics platforms. Reeves et al⁵⁰ describe the implementation of an Incidence Command Center that helped identify EHR-based tools to support clinical care during a COVID-19 surge. There were several tools developed at this institution and included outbreak management for COVID-19 patients (eg triaging, check-in, secure messaging, real-time data analytics). Timely integration of IT services to health care infrastructure to support clinic response efficacy related to COVID-19 and the use of telemedicine were initial priorities at the start of the epidemic.^51,52 Integration involved design and implementation of EHR-based rapid screening processes, laboratory testing, clinical decision support, reporting tools, and patient-facing technology.

ROLE OF PERIOPERATIVE INFORMATICIAN IN PUBLIC HEALTH PRACTICES

Starting with “Meaningful Use” (MU), a series of CMS-backed initiatives have incentivized hospitals and practitioners to adopt EHRs with public health reporting capabilities in recent years.⁵³ While MU has since been updated, MU involved the utilization of a verified EHR system to improve quality, safety, efficiency, and reduce health disparities, improve care coordination, improve population and public health, engage patients and their families in their own health care, and ensuring that patient privacy and security was maintained according to the Health Insurance Portability and Accountability Act (HIPAA) Privacy Rule. In 2011, 3 requirements were added to MU with direct relevance to the COVID-19 pandemic: the ability to electronically transmit laboratory results, immunization reporting, and key elements of disease surveillance.¹² These requirements were refined slightly with the 2018 transition to “Promoting Interoperability,” which replaced MU and required more evidence of active participation in data sharing.⁵⁴

As a result of these financial incentives, a huge proportion of anesthesia practitioners are able to use the EHR to relatively seamlessly report data to public health agencies.⁵⁵ Unfortunately, these same incentives were not applied to the public health agencies who are intended to receive these data. These agencies remain underfunded and are unable to keep up with the influx of data, which has led to a well-documented failure of coordination at a national level when public health challenges are faced (eg, COVID-19 pandemic).^56,57

Regional Health Information Exchanges (HIEs) have been proposed to facilitate regional data sharing and fill much of this gap. The initial enthusiasm for regional HIEs has since waned. Their adoption has generally been poor and may actually be declining in recent years.^55,56,58,59 The decline may ultimately increase hospitals’ burden to report to public health reporting agencies; in addition, the impetus for public health agencies will be to develop additional informatics infrastructure to process data efficiently from diverse sources (Figure 5).⁵⁶

The requirement to transmit data to clinical data registries is an equally-important component of Promoting Interoperability. Anesthesiology and other perioperative specialties have led in this field and demonstrate the utility of Promoting Interoperability. Registries including the Multicenter Perioperative Outcomes Group (MPOG),⁶⁰ the National Anesthesia Clinical Outcomes Registry,⁶¹ the Society for Thoracic Surgery Registry,⁶² and the American College of Surgery-National Surgery Quality Improvement Project (ACS-NSQIP)⁶³ have helped to facilitate data sharing to meet the Promoting Interoperability requirements. The surgical specialty-based registries in general demonstrate the potential for informatics collaboration with perioperative physicians.

As new, unanticipated public health crises arise, we would anticipate that anesthesiology and perioperative informaticians will continue to provide new and innovative solutions to leverage tools embedded within the EHR and aggregate and transmit critical public health data from the perioperative space to public health reporting agencies. CMS continues to require hospitals to do more, and it appears that EHR vendors will continue to offer increased reporting capabilities to meet demand and greatly facilitate increased adoption and dissemination.

The urgent needs currently are improvements on the public health agency side and improved coordination between clinical informaticians and EHR vendors to better facilitate public health data transmission, receipt, processing, and improved dissemination strategies.

EHR vendors need incentives to motivate them to engage with local, state, and federal agencies to ensure that data can be meaningfully used to generate usable data to guide public health policy. We propose a multipronged approach. First, federal support with earmarked funding specifically designated to federal, state, and local public health reporting agencies to address inadequacies and critical gaps revealed throughout the COVID-19 and more recent monkeypox pandemics. Second, CMS should offer incentives that require evidence that data are successfully both submitted to and successfully used by relevant reporting agencies on a sustained basis. This is a necessary next step beyond current Promoting Interoperability Stage 3 requirements because the ability to simply transmit data to public health reporting agencies and “active engagement” are inadequate. Finally, perioperative clinicians should engage and guide public health agencies on how data should best be presented and disseminated to ensure maximum relevance and utility.

PUBLIC HEALTH DATA RESOURCES AVAILABLE TO PERIOPERATIVE PHYSICIANS AND INFORMATICIANS

There are several public health resources available to perioperative physicians and informaticians. Initially, physicians may contact local, state, and federal public health agencies, the Agency for Healthcare Research and Quality, or seek guidance from physician-led societies such as the American Medical Association or the American Public Health Association. These organizations routinely collect data, develop analyses, and present prevailing guidelines and opinions on relevant public health issues. However, as public health data becomes increasingly “Big Data” in terms of volume, variety, and velocity, the underlying data and modeling will need to become integral to practices at all levels.¹⁶

One of the most likely sources of dependable information to practicing physicians will be facilitated via HIEs. There have been a few successful government-run and society-run HIE endeavors. In the United Kingdom and the Netherlands, large multicenter intensive care databases were successfully developed for public health tracking and pointed a way forward for deployment in the United States.^64,65 The underlying goals of these databases were to generate deidentified data while preserving privacy. Two current examples in the United States of perioperative data repositories are the MPOG,⁶⁰ and National Anesthesia Clinical Outcomes Registry.⁶¹ While the intended focus and structure of these lie primarily within research and quality improvement, these data repositories could be used for public health activities.

A purposeful restructuring of data repositories, with an eye on rapid curation and analysis of novel incoming data streams, will make these effective public health resources. Examples of automated public health resources in the infectious disease space include HealthMap, BioCaster, and Semantic Processing and Integration of Distributed Electronic Resources for Epidemiology (EpiSPIDER) that combine multiple data sources (eg, natural language processing of social media data) to provide real-time data on emerging diseases, notably without a reliance on HIE data aggregation.^66–68 Another example is the rapid local development of similar infectious disease dashboards during the COVID-19 epidemic in some states that proved to be useful for disease tracking.⁵⁹While some HIEs have found success, overall HIE use has stagnated or declined as health care organizations face challenges to collect and submit data and public health organizations encounter data aggregation difficulties.⁶⁹ Methods exist to develop health care organization pipelines to share data, but these are often limited by financial requirements and technological barriers that include interoperability (ie, lack of data standards), ease of data sharing, and also encompass nontechnological barriers such as network membership or public versus private hospital designation.^70,71

Ideally, with advances in technological capability, adequate resources for seamless data transfer, and coordination between data generators and data aggregators, there will be widespread availability of public health data through properly curated HIEs that are combined with social media, government records, meteorological data, and other information sources. Once these data are available, distinct teams could perform analyses manually or autonomously to provide timely guidance to practicing physicians.

There are numerous use cases relevant to perioperative physicians dealing with patient safety. State and national opioid registries are useful at the population level to influence policy through practice pattern analysis and identification of opioid-related overdoses and also at the patient level during preoperative assessment to detect potential misuse and influence pain management prescriptions^72–74 New York State’s Internet System for Tracking Over-Prescribing (I-STOP) program, for example, includes real-time opioid prescription tracking and has contributed to lower opioid prescribing.⁷⁵ Public gun registries can analyze gun violence–related injuries that allow trauma centers to estimate potential patient volume.⁷⁶ Similarly, HIEs with perioperative data that includes preoperative medications, patient demographics, and intraoperative parameters can be intermittently or continuously analyzed for physician practice patterns, documented medication adverse reactions, or new medication prevalence in a population. Early trend awareness can aid physicians to develop education and policies as public health issues emerge. For example, a hospital catchment area with worsening air quality could anticipate an increase in pulmonary disease exacerbations that could predispose patients to perioperative pulmonary complications.⁷⁷ As another example, an increase in adverse reactions related to a new perioperative medication may be the result of genetic factors or found only in select populations. Finally, public health initiatives may be instrumental to address health disparities and focus on SDOH that include health literacy and access to healthy diets.^21,78 Ultimately, the key drivers for any successful public health initiative will be the development of data pipelines for the vast amount of data, analysis protocols for each target, and identification of relevant outcomes from practicing physicians. This conceptual framework requires continued, active engagement by physicians in clinical informatics and public policy.

THE IMPORTANCE OF PHI IN PHI: PHI AND PUBLIC HEALTH INFORMATICS

Past and Current State of PHI

Privacy and protection of the individual is a central consideration in public health informatics. Advancing public health and public health informatics would benefit greatly from access to more sources of information across health systems and areas where critical patient data may be stored. However, this increased access to information may not come at the expense of compromising the individual patients’ privacy, security, and confidentiality. Beneficence and autonomy are central ethical principles in the individual versus public health information discussion and are needed to understand where public health informatics has been and where its future may lie to advance the discipline.

Privacy of health information from an informatics perspective is centered around PHI. PHI includes all personally identifiable health information. PHI can include any information that is transmitted or stored in any media format that “relates to past, present, or future physical or mental health or condition of an individual, the provision of health care to an individual, or the past, present, or future payment for the provision of health care for the individual,” where the information can be used to subsequently identify an individual.⁷⁹ Currently, there are 18 identifiers that designate information as PHI (Table 2). Health information is considered PHI when it includes one or more of the individual identifiers.

In the United States, patient privacy, the security of health data, and PHI are guarded by the patient privacy component of the HIPAA. Passed in 1996, HIPAA remains the most direct and important law protecting the right to privacy for all patients even though patient privacy was not the main intent of HIPAA legislation.^80,81 Specifically, the Privacy Rule of HIPAA applies to individuals and organizations that transmit health information during normal health care practice. For example, health care providers, health systems and health plans, are required to abide by the Privacy Rule. HIPAA violations can have both legal and financial ramifications, in addition to the critical breach of trust of patients that occurs. While a benefit to protecting the individual, regulations and historically limited technology and informatics abilities to address effective and efficient data sharing has often been viewed as an impediment to patient care.⁸²

HIPAA and the Privacy Rule provide guidance on how public health information should be shared. In general, public health needs are exempt from HIPAA regulations.⁸⁰ The Privacy Rule allows for the disclosure of health information to public health authorities without the patient’s authorization when required by another law or when the purpose is to prevent or control disease. A recent example of this is public health disease surveillance of COVID-19 and test result disclosures. Knowledge and information sharing of infectious disease burden that includes important individual PHI factors, such as location, could lead to more effective public health measures. On the other hand, potential individual privacy breaches and the risk of medical information theft as data is shared highlight the importance of technological advancements needed in public health informatics. The tension and balance between individual patient privacy and the need to protect the public’s health will continue to be a conversation for years to come in health care and information security.⁸³

Future State of PHI and Public Health Informatics

The future of public health informatics will need to include significant enhancements to how PHI is used, transmitted, and stored to maximize the value and sharing of data without compromising privacy. Current methods to select which PHI variables are needed or excluded from larger public health registries and data centers are a largely manual process and eliminate useful data that could be leveraged if protections could be ensured. The enhancement of PHI protection to ensure individuals’ privacy will require new models of data masking to allow data to be shared for the purposes of public health that will benefit from a larger pool and granularity of data to gain deeper analytical insights.

One of the most likely technologies that will become pervasive in the future of public health informatics involves data tokenization. Data tokenization is a process that replaces sensitive data with unique identification symbols that retain all the essential information without the potential security risks. Tokenization is currently used to help support credit card and e-commerce transaction data security. Its use in health care is not yet widely adopted. Tokenization assigns hashes and tokens to a group of identifier fields to encrypt a data file that contains PHI before it is shared with another organization and linked to a larger dataset.⁸⁴ The data linkage would then occur by comparing the hashes or tokens, which must match exactly, between the linked files.⁸⁵ Through the replacement of individual private health information with random inputs as tokens, the ability to link individual data across datasets (ie, public health data) to gain greater analytical insights and maximize privacy and reverse identification of patients can be achieved.

Tokenization is different from the data encryption data security methods used widely in health care currently (Figure 6). Encryption is a process that transforms sensitive data into an unreadable format using a specific, unique algorithm, and a password, or “key,” is required to decrypt the data. Encryption can be reversed by unauthorized individuals when the key is compromised. The privacy and security advantages of tokenization are derived from a nontraditional algorithm to mask the data.⁸⁷ Sensitive data, PHI, for example, is replaced with random data. A one-to-one mapping between sensitive PHI data and the corresponding random data effectively restricts data analysis only to individuals with access to the token who can reverse the process. Tokens can also be destroyed so that there will never be a way to reverse and reidentify the data’s sensitive information.

The future of PHI in public health informatics will require the discovery of technological solutions to enhance privacy of the individual and increase the volume and granularity of relevant health information in a timely fashion for public health. Tokenization and blockchain are emerging technologies that can be applied to health care data for this specific purpose.⁸⁸ It is vitally important for health care professionals and those working with PHI for public health to understand the appropriate application of these technologies.

ROLE OF KNOWLEDGE MANAGEMENT FRAMEWORKS FOR DECISION MAKING

Just as John Snow developed new techniques to map out the 1850s cholera outbreak, health officials have adapted how they collect, analyze, and think about data. In contrast to the 1850s, however, today’s physicians and public health informaticists are inundated with high volumes of data and will require new skills to make decisions quickly. Thought or knowledge management frameworks are neither recent developments nor limited only to medicine, but can help in decision making. Three useful frameworks are presented:

1. Data-Information-Knowledge-Wisdom (DIKW): Russell L. Ackoff, a well-known operational systems scientist, introduced what is now known as the knowledge hierarchy, or knowledge pyramid, in 1988. This DIKW hierarchy gave structure to how we handle concepts in informatics. Data are symbols that represent properties of objects, events, and their environments—they are products of observation. Information is extracted from data by analysis and viewed in context.
2. Therefore, the difference between data and information is functional, not structural. Knowledge is know-how—it is what makes possible the transformation of information into instructions. Dammann⁸⁹ suggests de-emphasizing wisdom in the framework and to insert evidence between information and knowledge (DIEK, Data, Information, Evidence, Knowledge). This framework defines data as raw symbols, which become information when viewed in context. Information achieves the status of evidence when compared to relevant standards. Finally, evidence is used to test hypotheses and is transformed into knowledge by success and consensus (peer review).
3. Information Chaos: The COVID-19 pandemic further shifted the dynamics of public health informatics. The urgency of the situation, the sheer volume from multiple sources of data, and the velocity at which the data arrives necessitated new strategies in public health informatics. Monkman et al updated Beasley et al’s⁹⁰ information chaos framework to understand the context of peoples’ experiences with information during the COVID-19 pandemic (Figure 7). This adapted framework can be used to characterize information-associated challenges observed during this time and the possible impact of information chaos on peoples’ cognition and behaviors.
4. Real-world evidence (RWE): Every practicing anesthesiologist likely felt the impact of the COVID-19 pandemic as it evolved in early 2020. The initial unknown, followed by the overload of information, deeply affected the specialty of anesthesia. Many likely thought that anesthesiologists were immune to, or at least overlooked, by the implications of public health issues. Yet anesthesiologists quickly found themselves combing through copious amounts of information and anecdotal experiences from colleagues around the world. The pandemic also introduced many of us to the concept and reality of RWE.
5. The 21st Century Cures Act (Cures Act) was designed to accelerate medical product development and led to the US Food and Drug Administration (FDA) to create a framework for evaluating the potential use of RWE. The FDA defines RWE as “data regarding the usage, or the potential benefits or risks, of a drug derived from sources other than traditional clinical trials.” Real-world data (RWD) are data relating to patient health status or the delivery of health care routinely collected from a variety of sources. RWE is the clinical evidence about the usage and potential benefits or risks of a medical product derived from analysis of RWD.⁹¹ Informaticists have known the benefits of RWD and RWE for some time, and it is very likely that all physicians will soon see its impact in their clinical careers (Figure 8). FDA Commissioner Scott Gottlieb states “As the breadth and reliability of RWE increases, so do the opportunities for FDA to make use of this information.”⁹¹
6. Strengths, Weaknesses, Opportunities, Threats (SWOT) analysis: As needs and opportunities for new policies, technology, and research arise within public health informatics, there will always be a need for a sound implementation strategy in PHI (Figure 5). Much like the knowledge frameworks above, there are strategic framework evaluation tools to assist with new idea evaluations from multiple perspectives. SWOT analysis is a common business strategy tool to assess how an organization compares to its competition (Figure 9). The concept of strategic fit, a ubiquitous objective sought by all organizations, can be explained by how well the internally related factors fit with the externally-related factors. SWOT may be extended to public health informatics organizations and perioperative informaticists as they implement interventions. An example related to weaknesses in the context of a digital health tool is that many overlapping data standards hinder interoperability and impede data aggregation.

DATA STANDARDS AND AGGREGATION PROCESSES

There are substantial technical challenges inherent to sharing data between organizations. One major challenge relates to data standardization requirements, where all parties must agree to use the same language when describing an element (eg, patient history, a vital sign, a physical examination element, a medication, aspects of a procedure, etc). Minor and major differences between how data are stored and represented in each institution’s electronic medical records (EMR) are not uncommon whether 2 institutions have adopted different EMRs, or use different implementations of the same EMR. Consider a comparison between a freestanding children’s hospital and a multispecialty trauma center that serves primarily adults. The children’s hospital may routinely capture data elements (eg, anesthetic induction type) not recorded at the trauma center, and vice versa (eg, Mallampati score). Even if every data element were precisely matched 1:1 between the 2 centers, the underlying representation is virtually guaranteed to be different.

Fortunately, there is a high degree of maturity in standardized vocabulary development that represents medical concepts. These vocabularies include representations of drugs and medications (National Drug Code [NDC]), laboratory and clinical observations (Logical Observation Identifiers Names and Codes [LOINC]), disease conditions (International Classification of Diseases [ICD]), medical services and procedures (Current Procedural Terminology [CPT]). Systemized Nomenclature of Medicine (SNOMED) is another vocabulary able to represent everything listed above and more.⁹² Some of these vocabularies are in very wide use. For example, use of NDC codes for medications is industry standard, and the barcode on every vial, jar, and package of prescription medication represents the NDC code for that specific brand and formulation of that drug.

Of course, vocabularies by themselves are only the beginning. Just as words need to be arranged into sentences and paragraphs to make sense, medical vocabularies need to be organized into specific data models for data to be aggregated and compared across institutions and entities. Examples of publicly available data models include those developed specifically for data aggregation for observational research (Observational Medical Outcomes Partnership [OMOP]; Informatics for Integrating Biology & the Bedside [i2b2]) and those to share data and for more general interoperability at both the individual patient level and in the aggregate (Health Level Seven International Version 2 [HL7v2]).⁹³ A testament to the maturity of these systems is that the NIH-sponsored National COVID Cohort Collaborative (N3C) was developed in short order and facilitated a great deal of hypothesis generation and hypothesis testing using large, multi-institutional data sets early in the pandemic.⁹⁴

While anesthesiology can certainly benefit from this maturity, there is no gold-standard approach to share a complete anesthesia record between institutions. A group at HL7 continues to develop a common data model for the anesthesia record, and once the standard is published it will need to be implemented by vendors to make a meaningful difference in interoperability and anesthesiology data sharing (Figure 10). Other critical future development areas for data standards and aggregation include development of techniques to deidentify and share image and video data (including radiology data); continued development of so-called model-to-data systems that allow testing of publicly transmitted models on privately held data without the need to share the data itself; development of standardized synthetic data sets generated with machine learning techniques that allow data sets to be shared that closely mimic, but are in fact not, PHI.⁹⁵

CONCLUSIONS

An unprecedented lengthening of the population’s lifespan has occurred. By 2034, adults over the age of 65 will outnumber children in the United States. Accelerated public health surveillance and insights into the determinants of disease at both the personal and population level could be instrumental to inform and shape public health policies and their effective implementation as our population continues to age.

A Possible Ideal Future State…

Ideally, over the next 10 years, popularization and implementation of telehealth and remote monitoring at scale will collect data with standardized definitions from a variety of sources across the health care continuum (Figure 11). Health care data is now merged with data that supports Urban Informatics—“a study of urban phenomena to address domain-specific urban challenges such as a pandemic response through a data science framework and computational techniques including sensing, data mining, information integration, modeling, analysis, and visualization” with essential context that includes people, place, and time. Data pipelines will deliver this vast amount of contextual data contemporaneously and automatically at the local, regional, and national levels into HIEs who all follow the same data standards. The tension and balance between patient privacy and the protection of the public’s health is resolved, and HIEs grant access to these data to protect the public’s health and use next-generation tokenization so that patients’ privacy, security, and confidentiality are not compromised.

Established Natural Language Processing (NLP), Machine Learning (ML), and Medical Artificial Intelligence (AI) algorithms developed and vetted by teams of practitioner informaticists validate, analyze, and interpret the HIE stored data provide insights into disease determinants at both the personal and population levels that are transformative for public health. Data visualizations assemble and concisely display this high volume of population health information and enable rapid analysis and interpretation with essential people, place, and time contexts to influence population-level severe chronic disease management. Implementation of precision public health initiatives guided by those same practitioner informaticists is effected through clinical decision support to practitioners at the point of care that identifies patients who could benefit from established protocolized interventions, ensure that the interventions occurred in a timely fashion. Finally, these findings and interventions are communicated transparently to the entire population.

Outcome data are then collected to support surveillance and determine the interventions’ effectiveness and use RWD and RWE to rapidly identify patient attributes and intervention protocol improvement opportunities to further improve patient and population health or mitigate a population health threat. The system would essentially adapt, evolve, and respond from its “experience” into something even stronger through its algorithms and practitioner informaticist oversight. The outcomes and improvement opportunities identified are also communicated transparently to the population with an update cadence that reflects the rate and significance of adaptation and evolution of what is known of the crisis, and what response is planned or in the process of implementation.

This example of a public health informatics infrastructure that could be ready for the next public health crisis is led by the practitioner informaticists who specialize in a variety of medical disciplines and are actively engaged in public policy and clinical informatics. A deep understanding of medicine and what useful data medical documentation possesses are the key to not only the development of this vital infrastructure but to its effectiveness to prevent or mitigate the next public health crisis.

A Possible Path Forward…

MU and Promoting Interoperability developed and implemented over the last decade was the beginning. These initiatives bolstered the EHRs nationally, but recent pandemics have exposed many weaknesses in the present data gathering and reporting structure of PHI and would likely benefit from a similar initiative to build out the public health infrastructure at all levels to aggregate the data for AI. Regardless of national incentivization, there are numerous opportunities and intriguing future possibilities for practitioner informaticist leadership.

Health Policy, Patient Privacy, and Public Health and Safety

A review of over 2700 hospitals highlighted difficulties with public health data and concluded there is a need to strengthen the public health data infrastructure.⁵⁶ Another review concluded that the current “hierarchical information supply chain model” public health model should be converted to a “peer-to-peer exchange with population health providers” and suggested innovations that included “a national patient identifier, population-level data exchange for immunization data, and computable electronic quality measures.”⁵⁸

Our COVID experience has driven a suggested overall framework that maps existing technologies onto the traditional public health disease outbreak stages of “prepare, respond, and recover.” It also incorporates stakeholders and sectors beyond governmental public health such as academic institutions, payers, providers, employers, etc.⁹⁶

The proposed Public Health 3.0 informatics tools must directly address public health data infrastructure and the United States’ aging population, the emergence of Telehealth, Mobile Health, remote monitoring and care and its associated data, and medical AI tool development to be fully effective. The enhanced tools could include early warning system enhancements for real-time, linked data on emerging health threats; early pandemic solution identification; skilled workforce development; and most importantly, common nationwide data standards for data access and exchange. This is an opportunity for practitioner informaticists who want to influence public policy.

Data Gathering

Manual keyboard data entry is required in many data-gathering efforts within systems and degrades data integrity. Data transmission and transfers may lack automation, and even with automation, data can be “lost in translation” due to the inconsistent application of data standards.

EHR reinforced discrete data creation and real-time data inputs through automated and standardized interfaces would support greater data integrity and support automated real-time reporting from EHRs and other sources (social media, etc).

Medical AI and the Last Mile Problem

Complex data from numerous sources must be defined consistently to facilitate the effective application of NLP, ML, and AI used to analyze gathered data and reach proposed conclusions. The right conclusion is not enough, however. It must be acted on, which requires delivery through “last mile” annunciation pathways to clearly present identified issues, their suspected causes, and the proposed right action to the right practitioner. The right action must be evidence-based and proven to accomplish the desired outcome. The right action then must be performed on the right patient, at the right time. Finally, outcomes are continuously monitored to guide and inform data collection and better-proposed conclusions in near-real-time.

The delivery of this clinical decision support will require reliable data source identification, integration of advanced data science techniques, and appropriate predictive modeling techniques that are free of algorithmic bias. Eventually, the real-time AI limitations⁹⁷ could be overcome by the development of better, or new, predictive modeling and projections.⁹⁸ Anesthesiologist informaticists that can identify reliable data sources, apply existing or develop new AI population informatics methodologies, deliver AI-proposed conclusions to the point of care, or measure outcomes could all have seats at the population health leadership table.

Communication

Messaging successes and failures has demonstrated some lessons in how to effectively deliver public health information to the population at large. In one study of message retransmission (eg, “retweeting”) on social media, the number of followers, message origin (eg, governmental authorities), and inclusion of video and audio increased retransmission. Mentions, replies, Uniform Resource Locator (URL) inclusion, and images use without text reduced retransmission.⁹⁹ Separately, another major weakness of the current system is the prevalence of mis- and disinformation¹⁰⁰ that goes beyond social media. Preprint servers that disseminated nonpeer-reviewed information proliferated during COVID. This generally has negative implications, but so far, the results have been mixed.^101,102

The message quality and clarity that is delivered is mostly dependent on how well the data is presented. Many “Ad hoc” visualization efforts occurred during the pandemic, and descriptions of dashboards constructed in a relatively short time frame are described in the literature.⁵⁹ Despite the most robust and thorough planning efforts, ad hoc visualization solutions to respond to unexpected events will be required.

There are novel data visualization techniques that have been used in other disciplines, such as Sankey diagrams (“alluvial representations of proportional flows [eg, matter, patients] as graphical tributaries”)¹⁰³ and “chloropleth.”¹⁰⁴ When EHR data are delivered in real-time, the data visualizations created to date are promising.⁹⁸

When the quality of data gathered is improved and approaches near-real-time, the advancement of real-time dashboards and application of novel data visualization techniques are another area where anesthesiologist informaticists could lead the way given our experience with the delivery of near-real-time data in the perioperative and intensive care settings.

Imagine what we could have accomplished during these latest pandemics if we had scaled monitoring, AI, and a fully interoperable HIE. How many lives could have been saved with RWE delivered to the point of care in near real-time? Anesthesiology has a rich history as a leader in safety and could serve and lead public health informatics to prepare us for the next, inevitable public health crisis (Figure 12). The significant gaps throughout our systems provide countless leadership opportunities to bring us to a possible ideal state and prepare us for the next public health crisis.

GLOSSARY