The adoption of electronic health records (EHRs) is transforming health care [1-4]. This digital infrastructure allows health services to store a patient’s complete medical history and collect additional observations and results in real time. Having this information in a standardized, readily accessible format provides a foundation for clinical tools to analyze these data and provide clinicians with the information to make evidence-based decisions at the point of care [1,2,4].

EHRs are enabling health care to move from protocol-based medicine to precision medicine [5,6] and helping bring about the next generation of evidence-based practice. Critical to this transformation are the clinical decision support systems (CDSSs). CDSSs are electronic systems that use the information in an EHR to support the treatment of a specific disease or group of related diseases [7]. Using a patient’s data in the EHR, a CDSS processes this information in real time and presents the results to clinicians, often with the context provided by the relevant clinical guidelines [7]. The clinician is then able to filter these outputs through the lens of their clinical experience, and the nuance of the scenario, to provide an individual with a precise intervention based on their unique physiology, medical history, and current situation (Figure 1).

CDSSs are usually carefully designed by groups of experts, undergo rigorous testing, and operate within strict governance structures. As a result, CDSSs have been shown to reduce medication errors and adverse clinical events [8]. By using the information in EHRs, CDSSs allow health care systems to move past models of practice designed for paper-based systems and enable new models of care that are better able to meet the quadruple aim of health care [9,10].

One exciting model of care, enabled by EHRs and CDSSs, is learning health care systems (LHSs). An LHS uses the data collected in routine clinical practice as evidence to determine the efficacy of an intervention. These learnings can then be used to inform clinicians treating patients with the same condition. An LHS shows how using the data routinely captured by an EHR in routine practice can be used to provide value to patients, clinicians, and the broader health care system [1,2,4]; however, for many health care systems, it is an aspirational goal (Figure 1).

Outside of LHSs, EHRs and CDSSs have the potential to facilitate a new paradigm in care—precision medicine [11,12]. Precision medicine refers to a tailored approach to care, guided by an individual’s medical history, environment, and genetic makeup [13,14]. The structured information in an EHR and the tools to contextualize and present this information to clinicians at the point of care have been used to benefit patients across a range of different areas of health [15,16]. While the capacity for digital health systems to capture and return information surrounding the patient’s medical history is well established, genomic data are not routinely incorporated into CDSSs alongside traditional clinical data sources.

Genomic data are widely accepted to be a foundational component of precision medicine [13,14]. Identifying the molecular cause of a patient’s condition can lead to tailored interventions [17], a better understanding of a patient’s prognosis [18], and can help individuals make informed decisions in family planning [19]. The information in an individual’s DNA is routinely being used to provide precision clinical care across a range of different areas (Table 1). A prime example of the potential of genomic information is oncology, where genomic testing is used to identify the range of mutations acquired by an individual’s tumor, leading to tailored therapeutic interventions [20]. The management of infectious disease is another area that shows the potential of genomics in personalized medicine, as genome sequencing can be used to diagnose specific pathogen as well as determine the strain of the infectious agent as well as its antibiotic-resistance profile [21]. The information in an individual’s DNA can have tremendous potential for many different areas of precision health care. However, for many clinicians in different areas of medicine, this information is only accessible by ordering a genomic test.

Population studies have revealed that each individual’s genome contains millions of different genetic variants [32]. The sheer number of variants means that it is unrealistic for a single specialist to keep track of the clinical significance of each of these variants across the range of diseases they examine. While genomic analyses would appear to be a prime candidate for the development of specialized CDSSs to support the use of genomic practice across a range of different areas of health (Table 1), CDSSs that routinely incorporate genetic information are rare [33,34]. There are likely many causes to this deficit; however, a significant factor to this can be attributed to the availability of interoperable genomic data within EHR. As a result, when many clinicians order genomic tests, the data are analyzed once, and the results are stored as a static PDF, locking the information away from future analyses.

Significant progress has been made in the development of systems to facilitate the use of genomic data in EHRs, such as clinical-grade genomic standards, file formats, and terminologies like Logical Observation Identifiers Names and Codes and Systematized Nomenclature of Medicine—Clinical Terms [35-38]. However, the adoption of these advances by EHR providers has been sluggish. As a result, EHRs are still struggling to store genomic data in a way that allows this information to be used by CDSSs. Without the capacity to access genomic data, clinicians are removed from an essential data source and will struggle to realize the full potential of precision medicine [12].

The reluctance to integrate genomic data into EHRs is likely due to a number of reasons. Some may suggest that the cause of this hesitation reflects the sheer volume and complexity of genomic data as well as the substantial amount of computer processing power and expertise required for genome analysis [39]. However, given the capacity of a VCF (variant call format) or VRS (variation representation) file to summarize the variants in a patient’s genome in a relatively potable format, the hesitancy to adopt these standards could be attributed to the complex ethical or social or legal questions surrounding genomics [12,40].

Despite these challenges, there are 2 questions that must be addressed to build a foundation to integrate genomic data into an EHR and enable genomics-empowered precision medicine: determining the right data to store and determining the right structure of these data. These questions are unlikely to have simple answers, as the answers will reflect the specific clinical questions being asked. While it is tempting to compare the virtues of exome and genome sequencing, discuss the impact of emerging technologies, or highlight the potential to bring other types of “omics” data into the EHR, these conversations are out of scope for this viewpoint. To us, it is clear that clinicians, scientists, and administrators must answer these questions together to ensure that genomic data can provide value across a range of different areas of precision medicine in their unique health service.

When determining how genomic data will be stored in an EHR, these conversations must address a unique attribute of genomic data—its (largely) static and unchanging nature. This attribute is typically brought up in discussions of secondary uses of genome data within the health care system [41]. However, a separate area of tremendous importance surrounds our evolving understanding of the clinical significance of a patient’s genomic data [42], as our changing understanding of the clinical relevance of a patient’s genetic data opens up new potential models of care.

The unchanging nature of a patient’s DNA and a rapidly changing understanding of the importance of that data mean that if a patient did not receive a molecular diagnosis after genomic testing, reanalyzing the same information at a later date with the context of new discoveries and new techniques can produce new molecular diagnoses [43-45]. While discovery and changing understandings are not unique to genomics, in contrast to other fields, the rate and volume at which new genomic information is accumulating is so extraordinary that reinterpreting existing genomic data with the context provided by new discoveries is known to increase diagnostic yields [42].

Special considerations will be needed to harness the levels of change associated with genomic data when designing genomics-enabled EHRs and CDSSs. Moreover, they highlight the need for these digital solutions to alert laboratories and clinicians when clinically important information has changed and robust systems in place for clinicians and laboratories to be empowered to use this information (Textbox 1).

While there is still work to be done, the eventual widespread adoption of genomic-enabled EHRs will facilitate the move from a traditional, prescriptive approach to medicine to personalized models of care. However, this will require a change in the way we approach genomic testing.

Currently, genomic tests resemble a “pull-based” approach. In this approach, only the genes of interest are analyzed, and the additional information needed to contextualize a patient’s genetic variants is “pulled” from the literature or analysis resources once. While there is a movement away from this philosophy, the singular, request nature of this approach prevents patients and clinicians from benefiting from our rapidly evolving understanding of genetic variants.

An alternative approach would be to perform genome sequencing once and store this information with the view that it will be used across the range of interactions an individual would have with the health system throughout their lifetime (Table 2). This will be facilitated by storing the data in structured, secure, interoperable formats, with the assumption that these data will be aligned to newer reference genomes, analyzed with different variant callers, and compared to constantly evolving virtual gene panels. While the raw genomic data might not need to be directly accessible in the EHR, reliable access to genome data will support every future interaction with a precision medicine–enabled health care system.

In this model, a CDSS could be designed around a “push” model. In the event of an inconclusive test, changes in the amount of information associated with the condition can be automatically monitored, and when it passes a threshold, the EHR can alert both the patient and the clinician to the potential for reanalysis. Patients who receive a molecular diagnosis from genomic testing could still benefit from continued monitoring by a CDSS. For example, the CDSS could highlight novel treatment interventions based on new information, such as new, targeted pharmacogenomic recommendations and potential clinical trial opportunities.

Key to this approach is the accessibility of genomic data for CDSSs. To give CDSSs access in a safe and transparent manner, there are significant challenges to overcome. Some of these challenges will be addressed from a bioinformatics perspective; however, others will require a clinical or health informatics solution, and some others still will require a policy or multidisciplinary approach.