6 E.H. Shortliffe and m.s. blois 2 Fig. 1.1 Inputs to the clinical-care record. The t telephone calls or prescriptions, and data obtained directly paper record is created by a from patients). The record thus becomes a merged collec- cesses that informat tion of such data, generally organized in chronological regarding direct encounters health pro order and patients, laboratory results, reports of that makes the system both smoothly integrated information onto datasheets that were later with workflow and useful for essentially every transcribed into computer databases for statistical patient encounter. nalysis(Fig. 1. 4). The approach was labor- The arguments for automating clinical-care intensive, fraught with opportunities for error, records are summarized in Chaps. 2 and 12 and in and added to the high costs associated with ran the now classic Institute of Medicine's report on domized prospective research protocols computer-based patient records(CPRs)(Dick The use of EHRs has offered many advantages and Steen 1991(Revised 1997). One argument to those carrying out clinical research(see Chap hat warrants emphasis is the importance of the 26). Most obviously, it helps to eliminate the EHR in supporting clinical trials-experiments manual task of extracting data from charts or fill in which data from specific patient interactions ing out specialized datasheets. The data needed are pooled and analyzed in order to learn about for a study can often be derived directly from the the safety and efficacy of new treatments or tests EHR, thus making much of what is required for and to gain insight into disease processes that are research data collection simply a by-product of not otherwise well understood. Medical research- routine clinical record keeping(Fig. 1.5). Other ers were constrained in the past by clumsy meth- advantages accrue as well. For example, the ods for acquiring the data needed for clinical record environment can help to ensure compli trials, generally relying on manual capture of ance with a research protocol, pointing out to a
6 that makes the system both smoothly integrated with workfl ow and useful for essentially every patient encounter. The arguments for automating clinical-care records are summarized in Chaps. 2 and 12 and in the now classic Institute of Medicine’s report on computer - based patient records ( CPRs ) ( Dick and Steen 1991 (Revised 1997) ). One argument that warrants emphasis is the importance of the EHR in supporting clinical trials —experiments in which data from specifi c patient interactions are pooled and analyzed in order to learn about the safety and effi cacy of new treatments or tests and to gain insight into disease processes that are not otherwise well understood. Medical researchers were constrained in the past by clumsy methods for acquiring the data needed for clinical trials, generally relying on manual capture of information onto datasheets that were later transcribed into computer databases for statistical analysis (Fig. 1.4 ). The approach was laborintensive, fraught with opportunities for error, and added to the high costs associated with randomized prospective research protocols. The use of EHRs has offered many advantages to those carrying out clinical research (see Chap. 26). Most obviously, it helps to eliminate the manual task of extracting data from charts or fi lling out specialized datasheets. The data needed for a study can often be derived directly from the EHR, thus making much of what is required for research data collection simply a by-product of routine clinical record keeping (Fig. 1.5 ). Other advantages accrue as well. For example, the record environment can help to ensure compliance with a research protocol, pointing out to a Fig. 1.1 Inputs to the clinical-care record. The traditional paper record is created by a variety of organizational processes that capture varying types of information (notes regarding direct encounters between health professionals and patients, laboratory or radiologic results, reports of telephone calls or prescriptions, and data obtained directly from patients). The record thus becomes a merged collection of such data, generally organized in chronological order E.H. Shortliffe and M.S. Blois
1 Biomedical Informatics: The Science and the Pragmatics Fig. 1.2 Outputs from the clinical-care record. Once patient care. Numerous providers are typically involved in information is collected in the traditional paper chart, it a patients care, so the chart also serves as a means for y be provided to a wide variety of potential users of the communicating among them. The mechanisms for dis information that it contains. These users include health playing, analyzing, and sharing information from such professionals and the patients themselves but also a wide records results from a set of processes that often varies ariety of"secondary users"(represented here by the indi- substantially across several patient-care settings and viduals in business suits) who have valid reasons for institutions ccessing the record but who are not involved with direct clinician when a patient is eligible for a study or organizations, as well as individual provider when the protocol for a study calls for a specific groups, have invested heavily in guideline devel management plan given the currently available opment, often putting an emphasis on using clear data about that patient. We are also seeing the evidence from the literature, rather than expert development of novel authoring environments for opinion alone, as the basis for the advice. Despite clinical trial protocols that can help to ensure that the success in creating such evidence-based the data elements needed for the trial are compat- guidelines, there is a growing recognition that ible with the local EHRs conventions for repre- we need better methods for delivering the deci senting patient descriptors sion logic to the point of care. Guidelines that Another theme in the changing world of health appear in monographs or journal articles tend to are is the increasing investment in the creation sit on shelves, unavailable when the knowledge of standard order sets, clinical guidelines, and they contain would be most valuable to practitio- clinical pathways(see Chap. 22), generally in an ners. Computer-based tools for implementin effort to reduce practice variability and to develop such guidelines, and integrating them with the consensus approaches to recurring management EHR, present a means for making high-quality problems. Several government and professional advice available in the routine clinical setting
7 clinician when a patient is eligible for a study or when the protocol for a study calls for a specifi c management plan given the currently available data about that patient. We are also seeing the development of novel authoring environments for clinical trial protocols that can help to ensure that the data elements needed for the trial are compatible with the local EHR’s conventions for representing patient descriptors. Another theme in the changing world of health care is the increasing investment in the creation of standard order sets , clinical guidelines , and clinical pathways (see Chap. 22), generally in an effort to reduce practice variability and to develop consensus approaches to recurring management problems. Several government and professional organizations, as well as individual provider groups, have invested heavily in guideline development, often putting an emphasis on using clear evidence from the literature, rather than expert opinion alone, as the basis for the advice. Despite the success in creating such evidence - based guidelines , there is a growing recognition that we need better methods for delivering the decision logic to the point of care. Guidelines that appear in monographs or journal articles tend to sit on shelves, unavailable when the knowledge they contain would be most valuable to practitioners. Computer-based tools for implementing such guidelines, and integrating them with the EHR, present a means for making high-quality advice available in the routine clinical setting. Fig. 1.2 Outputs from the clinical-care record. Once information is collected in the traditional paper chart, it may be provided to a wide variety of potential users of the information that it contains. These users include health professionals and the patients themselves but also a wide variety of “secondary users” (represented here by the individuals in business suits) who have valid reasons for accessing the record but who are not involved with direct patient care. Numerous providers are typically involved in a patient’s care, so the chart also serves as a means for communicating among them. The mechanisms for displaying, analyzing, and sharing information from such records results from a set of processes that often varies substantially across several patient-care settings and institutions 1 Biomedical Informatics: The Science and the Pragmatics
8 E.H. Shortliffe and m.s. blois Fig 1.3 Complex proc nanded of the record. as that information to those who have valid reasons for shown in Figs 1. I and I g it. Paper-based documents are severely limited tion of a complex set of ng the diverse requirements for data collection both gather information n distribute rmation access that are implied by this diagram Many organizations are accordingly attempting one of the central topics in Chap. 10. Issues of decision-support tools with their direct data entry by clinicians are discussed in EHR systems, and there are highly visible efforts Chaps. 2 and 12 and throughout many other underway to provide computer-based diagnostic chapters as well. Chapter 13 examines the fourth decision support to practitioners. topic, focusing on recent trends in networked There are at least four major issues that have data integration, and offers solutions for the ways consistently constrained our efforts to build in which the EHR can be better joined with other effective EHRs: (1)the need for standards in the relevant information resources and clinical pro- area of clinical terminology: (2)concerns regard- cesses, especially within communities where ng data privacy, confidentiality, and security; (3) patients may have records with multiple provid- challenges in data entry by physicians; and (4) ers and health care systems(Yasnoff et al 2013 difficulties associated with the integration of record systems with other information resources in the health care setting. The first of these issues 1.1.3 Anticipating the Future of is discussed in detail in Chap. 7, and privacy is Electronic Health records Ihttp://www.forbes.com/sites/bruceupbin/2013/02/08/Oneofthefirstinstinctsofsoftwaredevel ibms-watson-gets-its-first-piece-of-business-in- opers is to create an electronic version of an healthcare/.(Accessed 4/21/13/) object or process from the physical world. Some
8 Many organizations are accordingly attempting to integrate decision-support tools with their EHR systems, and there are highly visible efforts underway to provide computer-based diagnostic decision support to practitioners. 1 There are at least four major issues that have consistently constrained our efforts to build effective EHRs: (1) the need for standards in the area of clinical terminology; (2) concerns regarding data privacy, confi dentiality, and security; (3) challenges in data entry by physicians; and (4) diffi culties associated with the integration of record systems with other information resources in the health care setting. The fi rst of these issues is discussed in detail in Chap. 7, and privacy is 1 http://www.forbes.com/sites/bruceupbin/2013/02/08/ ibms-watson-gets-its-first-piece-of-business-inhealthcare/ . (Accessed 4/21/13/). one of the central topics in Chap. 10. Issues of direct data entry by clinicians are discussed in Chaps. 2 and 12 and throughout many other chapters as well. Chapter 13 examines the fourth topic, focusing on recent trends in networked data integration, and offers solutions for the ways in which the EHR can be better joined with other relevant information resources and clinical processes, especially within communities where patients may have records with multiple providers and health care systems ( Yasnoff et al. 2013 ). 1.1.3 Anticipating the Future of Electronic Health Records One of the fi rst instincts of software developers is to create an electronic version of an object or process from the physical world. Some Fig. 1.3 Complex processes demanded of the record. As shown in Figs 1.1 and 1.2 , the clinical chart is the incarnation of a complex set of organizational processes, which both gather information to be shared and then distribute that information to those who have valid reasons for accessing it. Paper-based documents are severely limited in meeting the diverse requirements for data collection and information access that are implied by this diagram E.H. Shortliffe and M.S. Blois
1 Biomedical Informatics: The Science and the Pragmatics record Definition of data elements Data sheets .Definition of eligibility Process descriptions Stopping criteria Other details of the trial Fig. 1. 4 Traditional data collection for clinical trials. Alternatively, data managers have been hired to abstract Although modern clinical trials routinely use computer the relevant data from the chart. The trials are generally systems for data storage and analysis, the gathering of designed to define data elements that are required and the search data is still often a manual task. Physicians who methods for analysis, but it is common for the process of are for patients enrolled in trials, or their research assis. collecting those data in a structured format to be left tants, have traditionally been asked to fill out special data- manual processes at the point of patient care sheets for later transcription into computer databases. Electronic Health Clinical Data Record(EHR Repository Q Clinicaltrialdesign Clinical trial Definition of data elements database Definition of eligibility Process descriptions A Stopping criteria other details of the trial Results Fig 1.5 Role of electronic health records(EHRs)in sup- interaction of the ith the ehr permits two- porting clinical trials. with the introduction of EHR sys- way communica tems, the collection of much of the research data for ity and efficiency of the clinical trial. Physicians can be clinical trials can become a by-product of the routine care reminded when their patients are eligible for an experi of the patients. Research data may be analyzed directly mental protocol, and the computer system can also remind from the clinical data repository, or a secondary research the clinicians of the rules that are defined by the research database may be created by downloading information protocol, thereby increasing compliance with the experi- from the online patient records. The manual processes in mental plan Fig. 1. 4 are thereby largely eliminated. In addition, the
9 Medical record Computer database Data sheets Analyses Results Clinical trial design •Definition of data elements •Definition of eligibility •Process descriptions •Stopping criteria •Other details of the trial Fig. 1.4 Traditional data collection for clinical trials. Although modern clinical trials routinely use computer systems for data storage and analysis, the gathering of research data is still often a manual task. Physicians who care for patients enrolled in trials, or their research assistants, have traditionally been asked to fi ll out special datasheets for later transcription into computer databases. Alternatively, data managers have been hired to abstract the relevant data from the chart. The trials are generally designed to defi ne data elements that are required and the methods for analysis, but it is common for the process of collecting those data in a structured format to be left to manual processes at the point of patient care Clinical trial database Clinical Data Repository Electronic Health Record (EHR) Analyses Results Clinical trial design •Definition of data elements •Definition of eligibility •Process descriptions •Stopping criteria •Other details of the trial Fig. 1.5 Role of electronic health records (EHRs) in supporting clinical trials. With the introduction of EHR systems, the collection of much of the research data for clinical trials can become a by-product of the routine care of the patients. Research data may be analyzed directly from the clinical data repository, or a secondary research database may be created by downloading information from the online patient records. The manual processes in Fig. 1.4 are thereby largely eliminated. In addition, the interaction of the physician with the EHR permits twoway communication, which can greatly improve the quality and effi ciency of the clinical trial. Physicians can be reminded when their patients are eligible for an experimental protocol, and the computer system can also remind the clinicians of the rules that are defi ned by the research protocol, thereby increasing compliance with the experimental plan 1 Biomedical Informatics: The Science and the Pragmatics
E.H. Shortliffe and m.s. blois familiar notion provides the inspiration for a new efficiency of airplanes and air travel would not software product. Once the software version has have improved as they have A similar point can been developed, however, human ingenuity and be made about the importance of committing to creativity often lead to an evolution that extends the use of EHRs today, even though we know that the software version far beyond what was ini- they need to be much better in the future tially contemplated. The computer can thus facil- itate paradigm shifts in how we think about such Consider, for example, the remarkable differ. 1.2 Communications familiar concepts Technology and Health ence between todays office automation software Data Integration and the typewriter, which was the original inspi ration for the development of"word processors". An obvious opportunity for changing the role and hough the early word processors were functionality of clinical-care records in the digi- designed largely to allow users to avoid retyping tal age is the power and ubiquity of the Internet. papers each time a minor change was made to a The Internet began in 1968 as a U.S. research document, the document-management software activity funded by the Advanced Research of today bears little resemblance to a typewriter. Projects Agency(ARPA) of the Department of Consider all the powerful desktop-publishing Defense. Initially known as the ARPANet, the facilities, integration of figures, spelling correc- network began as a novel mechanism for allow tion, grammar aids, "publishing"on the Web, use ing a handful of defense-related mainframe com of color, etc. Similarly, todays spreadsheet pro- puters, located mostly at academic institutions or grams bear little resemblance to the tables of in the research facilities of military contractors, numbers that we once created on graph paper. To to share data files with each other and to provide take an example from the financial world, remote access to computing power at other loca- sider automatic teller machines(ATMs) and tions. The notion of electronic mail arose soon facilitation of todays worldwide banking in ways thereafter, and machine-to-machine electronic that were never contemplated when the industry mail exchanges quickly became a major compo- depended on human bank tellers nent of the networks traffic. As the technology It is accordingly logical to ask what the health matured, its value for nonmilitary research activi record will become after it has been effectively ties was recognized, and by 1973 the first medi implemented on computer systems and new cally related research computer had been added opportunities for its enhancement become increas- to the network (Shortliffe 1998a, 2000) ingly clear to us. It is clear that EHRs a decade During the 1980s, the technology began to be from now will be remarkably different from the developed in other parts of the world, and the antiquated paper folders that until recently domi- National Science Foundation took over the nated most of our health care environments. Note of running the principal high-speed backbone that the state of today's EHR is roughly compa- network in the United States. Hospitals, mostly rable to the status of commercial aviation in the academic centers, began to be connected to what 1930s. By that time air travel had progressed sub- had by then become known as the Internet, and in stantially from the days of the Wright Brothers, a major policy move it was decided to allow com and air travel was becoming common. But 1930s mercial organizations to join the network as well air travel seems archaic by modern standards, and By April 1995, the Internet in the United States it is logical to assume that todays EHRs, albeit had become a fully commercialized operation,no much better than both paper records and the early longer depending on the U.S. government to sup- computer-based systems of the 1960s and 1970s, port even the major backbone connections will be greatly improved and further modern- Today, the Internet is ubiquitous, accessible ized in the decades ahead. If people had failed to through mobile wireless devices, and has pro- se the early airplanes for travel, the quality and vided the invisible but mandatory infrastructure
10 familiar notion provides the inspiration for a new software product. Once the software version has been developed, however, human ingenuity and creativity often lead to an evolution that extends the software version far beyond what was initially contemplated. The computer can thus facilitate paradigm shifts in how we think about such familiar concepts. Consider, for example, the remarkable difference between today’s offi ce automation software and the typewriter, which was the original inspiration for the development of “word processors”. Although the early word processors were designed largely to allow users to avoid retyping papers each time a minor change was made to a document, the document-management software of today bears little resemblance to a typewriter. Consider all the powerful desktop-publishing facilities, integration of fi gures, spelling correction, grammar aids, “publishing” on the Web, use of color, etc. Similarly, today’s spreadsheet programs bear little resemblance to the tables of numbers that we once created on graph paper. To take an example from the fi nancial world, consider automatic teller machines (ATMs) and their facilitation of today’s worldwide banking in ways that were never contemplated when the industry depended on human bank tellers. It is accordingly logical to ask what the health record will become after it has been effectively implemented on computer systems and new opportunities for its enhancement become increasingly clear to us. It is clear that EHRs a decade from now will be remarkably different from the antiquated paper folders that until recently dominated most of our health care environments. Note that the state of today’s EHR is roughly comparable to the status of commercial aviation in the 1930s. By that time air travel had progressed substantially from the days of the Wright Brothers, and air travel was becoming common. But 1930s air travel seems archaic by modern standards, and it is logical to assume that today’s EHRs, albeit much better than both paper records and the early computer-based systems of the 1960s and 1970s, will be greatly improved and further modernized in the decades ahead. If people had failed to use the early airplanes for travel, the quality and effi ciency of airplanes and air travel would not have improved as they have. A similar point can be made about the importance of committing to the use of EHRs today, even though we know that they need to be much better in the future. 1.2 Communications Technology and Health Data Integration An obvious opportunity for changing the role and functionality of clinical-care records in the digital age is the power and ubiquity of the Internet. The Internet began in 1968 as a U.S. research activity funded by the Advanced Research Projects Agency (ARPA) of the Department of Defense. Initially known as the ARPANET , the network began as a novel mechanism for allowing a handful of defense-related mainframe computers, located mostly at academic institutions or in the research facilities of military contractors, to share data fi les with each other and to provide remote access to computing power at other locations. The notion of electronic mail arose soon thereafter, and machine-to-machine electronic mail exchanges quickly became a major component of the network’s traffi c. As the technology matured, its value for nonmilitary research activities was recognized, and by 1973 the fi rst medically related research computer had been added to the network (Shortliffe 1998a , 2000 ). During the 1980s, the technology began to be developed in other parts of the world, and the National Science Foundation took over the task of running the principal high-speed backbone network in the United States. Hospitals, mostly academic centers, began to be connected to what had by then become known as the Internet, and in a major policy move it was decided to allow commercial organizations to join the network as well. By April 1995, the Internet in the United States had become a fully commercialized operation, no longer depending on the U.S. government to support even the major backbone connections. Today, the Internet is ubiquitous, accessible through mobile wireless devices, and has provided the invisible but mandatory infrastructure E.H. Shortliffe and M.S. Blois