Which data standard is likely to be used by electronic health records for genetic information?

Open-source EMR systems are becoming very popular in less developed countries of the world. In the USA, however, many commercial EMR systems are vying for dominance in healthcare organizations that have the financial flexibility to afford them. This chapter contains discussions and summaries of a number of open-source EMR systems in use today. Each healthcare organization must evaluate the pros and cons of adoption of open-source EMR systems, and judge whether or not features of specific EMR packages suit the needs of the organization. The chapter will provide the “grist” for the decision “mill” for choosing the EMR package that is suited best to the organization.

SMART and FHIR

Fully interoperable electronic medical records systems are far from reality. Many HIE approaches achieve some degree standardized transmission and receipt of data sets, but true semantic interoperability exists only in limited settings and EHR components are certainly not substitutable. In the United States, some standards specified by Meaningful Use regulations seem excessively complex, incompletely specified, or inconsistent in their implementations. As a result, many “standard” implementations have not yet realized their intended degree of interoperability. In part as a reaction to these complexities, a growing number of vendors and developers are creating “apps” using an updated version of SMART (Substitutable Medical Apps, Reusable Technology) that exploits the data models and application programming interface specified by less complex, openly licensed HL7 standard called Fast Health Interoperability Resources (FHIR) (see Chapter 16: An Apps-Based Information Economy in Healthcare) (Mandel et al., 2016).

4.2.2 Electronic Medical Records

Although the primary purpose of EMR systems is to store data for use in clinical practice, they are clearly valuable repositories for research use. However, the indexing and retrieval capabilities of these systems are highly tailored to the workflow of clinical practice, on the hospital wards and in the outpatient clinics. There are no general search facilities. Nevertheless, EMR systems are being used to support research. The usual method for this is to create a separate data warehouse, in which data are extracted (copied) from the EMR and put into a suitable form, usually a separate relational database. Such data warehouses are then searchable in all the ways mentioned in Chapter 1, insofar as the data fit the relational model. However, a growing proportion of the content of EMR systems is in the form of unstructured text. Even more is in the form of PDF files that are the result of scanning, or output from some other system. The unstructured text is amenable to NLP techniques as described. The PDF files are at present essentially unreachable by automated methods.

In addition, since much of the data contained in EMR systems has a temporal aspect, another big challenge is to represent time-oriented data and query about time. Examples are series of vital signs measurements during an inpatient stay. Each should have a time stamp. Some way to describe trends, events, and other structures is needed. This is an active area of research and experimentation.

Performance and storage scalability

The requirement of scalability for EMR systems is considered restrictive as most of the current systems are based on relational databases which are not designed to scale, which calls for a need of newer scalable database systems to maximize output at the expense of lower storage requirement (Jin et al., 2011). Based on the objective tests (performance and storage size) conducted in this study, it was noticeable that NoSQL databases such as MongoDB and Neo4j performed better with a lower disk space requirement compared to MS SQL Server.

To scale is to either increase hardware capacity or increase the amount of data stored per capacity. For an example based on results in, 25.5 MB was required to store 99 records in MS SQL Server, whereas 99 records occupied only 0.50 MB and 18.87 MB in MongoDB and Neo4j respectively. Therefore, with an effective similar storage size of 25.5 MB, MongoDB and Neo4j can theoretically store up to 5049 and 135 records, a 51 and 1.35 time increased in the amount of data stored per same capacity respectively due to newer storage and compression technology.

On the contrary, NoSQL database systems are based on shared-nothing approach where servers have their own dedicated resources such as RAM, processor or storage, allowing scaling horizontally (Cattell, 2011).

Family History Tools and Electronic Health Records

If family history tools were integrated into electronic medical record (EMR) systems, they could offer clinicians the advantages of standardizing the collection and organization of family history data. Electronic clinical decision support could also provide familial risk assessment and pedigree analysis, standardized guidance for referral for genetics consultation and testing, and evidence-based recommendations for interventions specific to the familial risk.

Additionally, family history tools integrated within personal health record (PHR) systems used by consumers have the potential to improve self-management of disease and disease risk factors through built-in messaging about health promotion and disease prevention activities tailored to an individual's familial risk and personal characteristics. Such PHR products could further improve the accuracy of family history reports by capabilities that allow health information (including results of genetic testing) to be transferred directly to family members rather than relying on clinicians to obtain and review records and transmit information. This direct communication among family members would improve familial risk assessment and pedigree analysis. The privacy concerns that are often raised by consumers regarding familial or genetic information could also be addressed if these PHRs offer methods for giving consumers control over (1) who may receive their information, (2) what specific types of information these recipients may get, and (3) how the recipients are limited in further sharing of this information.

Family history applications developed for EMR and PHR products must have common data requirements and technical standards to ensure optimal exchange of family history data and ultimately use of family history information for disease management and health promotion activities, including the ordering and interpretation of genetic tests. At present, several different standards development organizations are working to create a framework for representing and exchanging the contents of EHRs (including the family health history). The two most prominent of these are Health Level 7's Clinical Document Architecture and the American Society of Testing and Materials International Continuity of Care Record. Two in this case is not better than one, as the differing standards and data architectures may prove to be substantial obstacles to efficient data exchange. Fortunately, it appears that both organizations will work collaboratively to define a common standard (Ferranti et al., 2006).

Population Approach

Although family history is known to be a risk factor for many chronic diseases, its use in preventive medicine and public health has been de-emphasized compared with modifiable risk factors such as smoking and diet. However, recent studies show that a large fraction of the population is likely to have a family history of one or more common diseases. For example, a population-based study of family history of cardiovascular disease in Utah showed that 72% of early coronary heart disease (diagnosed before age 55 years) in the population occurred in 14% of families and 86% of early stroke occurred in 11% of families (Hunt et al., 2003). In a recent analysis of the National Health and Nutrition Examination Survey 1999–2002, 48% of the study population reported having at least one relative with diabetes (Hariri et al., 2006). And in Michigan, 7% of the population reported having an immediate family member diagnosed with colorectal cancer according to the 2005 Behavioral Risk Factor Survey (Personal communication, Deb Duquette, Michigan Department of Community Health, 2006). These data suggest that implementation of family history based strategies to assess disease risk, influence early disease detection, and encourage lifestyle changes could lead to overall population health benefits. An advantage of family history based approaches to prevention is that they do not focus exclusively on an individual's risk factors but can work within a framework of biologic and cultural relationships to affect risk factor reduction.

The Next 4 Years – by-2018 Predictions

•

More than 90% of US hospitals and larger medical practices will be using EMR systems with advanced analytics and predictive analytics.

More than 70% of US doctors will be working in organizations that charge fees based on other than fee-for-service models; ACOs (accountable care organizations) and HMOs will become the dominant model for health care in the USA. Integrated delivery organizations will use advanced Health IT systems with predictive analytics.

New types of professional models for health care will emerge, such as Wal-Mart delivery of drop-in medical care; we will see rapid growth of comprehensive health plans with shared risk, and “medical tourism” with patients traveling to high quality overseas centers of excellence at lower costs.

Globally, millions upon millions of patients will be remotely monitored (telehealth).

Predictive analytics will combine with health IT systems, digital and mobile, and with comprehensive genome services studies on all high-risk patients and those with chronic diseases to create fully personalized medicine; personalized drug therapy will improve outcomes and reduce patient safety risks.

Contents of Electronic Medical Record Databases

EMR databases may contain data for patients from one site or many, or those representing multiple sites; those sites may be using the same EMR system for entering the patient data or using different EMR software. Because these databases represent different collections of sites (i.e., clinical practices or medical facilities) and of EMR software, their contents can differ widely. Many of these databases bring together data from a collection of sites of the same type; for example, there are EMR databases from general practitioners only and others from oncology clinics only. Others provide data from integrated health systems and include a combination of outpatient primary care, outpatient specialty, and inpatient settings. The latter type of databases can allow a complete view of all of the medical care a patient receives within that health system (Fig. 3.2.1).

Information on active diagnoses for patients is an essential part of all EMR databases used for drug safety research. Diagnoses may be entered using a standard coding system, such as ICD-10 (WHO, 2016), or can be entered as free text. The date attached to a diagnosis may represent the date of a visit during which the condition was given consideration by the physician or the original date of diagnosis; some databases provide both of these dates, allowing a view of duration of disease and patterns of follow-up care for the disease. Similarly, for any database that might be considered for PASS, some levels of detail are needed on prescription treatments received. Although some EMR databases contain linked pharmacy dispensing data, most include prescriptions only. The database might note, for example, the name of the medicinal product, date prescribed, and dose prescribed; repeat prescriptions would be noted, but the dates on which the patient filled the prescription and any allowed refills would not be available. Supply times allowed in each prescription may or may not be indicated in the data.

One of the hallmarks of most EMR databases is the availability of laboratory results data. Many of these databases contain laboratory test names, either in free text or using a coding system such as Logical Observation Identifiers Names and Codes , along with the results of the test, date of the test, and sometimes the normal range specific to the laboratory that processed the test (LOINC, 2017). The completeness of the laboratory results data depends, in part, on the data entry mechanism used. When the laboratory reports data electronically into the system, completeness is excellent for all laboratory tests measured in that practice. However, when manual entry is required, the results of the tests may not be available consistently. Nonstandard tests that are processed by specialty laboratories are particularly subject to this limitation in many databases; in some cases, the specialty laboratory results appear infrequently if at all.

Although the use of EMR in clinical practice includes an electronic recording of progress notes and other open, free-text information, this component of the record is typically not available in database extracts that can be purchased for analyses by parties other than the data owner. Such open fields may contain patient names and other identifiers; sharing the data would thus violate patient privacy. Some EMR data owners are able to conduct targeted searches of the notes to identify information of interest. The information is placed into separate fields and contains only the data elements of interest. For example, a search of progress notes for any mention of symptoms of psychosis can be used to create new variables indicting the name of the symptom and date recorded. Finding the information within the free-text data is a complex process that requires sophisticated natural language processing to ensure, for example, that a note stating that a patient did not have a given event (e.g., condition or treatment) is not incorrectly coded as the patient having had the event (Xu et al., 2011; Henriksson et al., 2015).

Site-specific EMR databases that reflect, for example, primary care settings only may include some limited information for other settings. Referrals to other specialties, summary information from hospitalizations, and notes of medications that were prescribed elsewhere can appear in some databases. In general, the information that is available on diagnoses and treatments received outside of the EMR system should be used with caution, as it may have a high likelihood of missing or erroneous data. Fig. 3.2.2 presents the relative degree of availability of various types of information in different classes of EMR databases compared with the data available in claims databases (see Chapter 3.1 for detailed discussion about administrative claims databases). Because of the wide variation in EMR databases, however, some specific databases within a given type may differ considerably with respect to availability of one or more data types. A key requirement, prior to embarking on any PASS using an EMR database, is to assess carefully the type and extent of data available in the selected database that is relevant to identifying the study population, predictors, and outcomes.

The Electronic Medical Records and Genomics (eMERGE) Network

The Electronic Medical Records and Genomics (eMERGE) Network is a national consortium organized by the National Human Genome Research Institute to combine DNA biorepositories information with electronic medical record (EMR) systems for large-scale, high-throughput genetic research. There are several academic medical centers across the US that participate in this network. The eMERGE model does not require active recruitment for the study or gathering of samples because it uses cases and controls from various EMR systems, for which genetic samples have already been collected as specimens, thus saving time and money. This network uses these data not only for genetic research, but also to help develop some other frameworks and guidelines related to genomic medicine in the areas of ethics, legal issues, privacy, and community engagement (http://emerge.mc.vanderbilt.edu/; Katsanis and Katsanis, 2013).

Be Aware of All Potential Readers

Appropriate radiology reporting requires no apologies. Other health care providers may gain access to radiology reports through requests of medical documents or, in the future, through EMR systems. Some advocate that radiologists provide readings and reports directly to patients to facilitate their involvement and sharing of responsibility in health care decision making.131,132 This is often done to report normal results so as to decrease patient anxiety and to answer questions posed directly by a patient to a radiologist.62 Patients, family members, and legal representatives often have access to radiology reports. In fact, the federal Health Insurance Portability and Accountability Act (HIPAA) of 1996 requires a patient be able to see and obtain copies of their medical records. Do not write anything in a radiology report that should not be read by some other person who is legally involved with the case.109

Summary

With the advent of Jaffee v. Redmond and the HIPAA regulations, documentation of psychotherapy has achieved an unprecedented level of protection. However, the rush toward nationwide, interoperable electronic medical records systems raises important concerns about how these protections will be maintained. Most therapists prudently maintain a general clinical/medical record of treatment that is factual, legible, and complete insofar as the objective clinical status of the patient is concerned. Psychotherapy itself may be documented in abstract terms in the general clinical record without details that are personal to the patient. As a safer option, therapists may record more specific information about the process and content of psychotherapy in protected psychotherapy notes kept in a separate part of the patient's identifiable clinical record. These notes have a considerably higher level of protection from disclosure because specific, delimited patient authorization is required for disclosure. The therapist may also elect to keep highly sensitive or personal information in personal working notes entirely outside the identifiable and permanent clinical record, but important cautions apply to this practice.