24 The Future of Computer Applications in Biomedicine

The Future of Computer Applications in Biomedicine

L AWRENCE M. F AGAN AND E DWARD H. S HORTLIFFE

After reading this chapter, you should know the answers to these questions:

●

What are possible future directions for biomedical informatics?

●

What are the forces that are driving these changes?

In this book, we have summarized the current state of biomedical informatics in a vari- ety of application areas and have reflected on the development of the field during the past 50 years. To provide a background for our discussions, we opened the book with a glimpse into the future—a vision of medical practice when individual physicians rou- tinely and conveniently use computers and electronic health records to help with infor- mation management, communication, and clinical decision making. In this chapter, we again look forward, this time concentrating on likely trends in biomedical applications of computers, on current avenues of research, and on the issues that will determine along which paths biomedical informatics will develop.

24.1 Progress in Biomedical Computing

We begin by looking back at the changes in biomedical computing since the first edition of this book was published in 1990. Then we look ahead to the not-too-distant future—

presenting a few scenarios that we can extrapolate from the current trends in the field.

These scenarios provide perspective on the ways that computers may pervade clinical practice and the biological science laboratory. A key aspect of the clinical scenarios is the extent to which, unlike most specialized medical paraphernalia of today, medical computing applications are integrated into routine medical practice rather than used on an occasional basis. In much the same way, computers are becoming a crucial part of the analysis of data in the research laboratory, especially in the areas of genomics and proteomics, where the amount of incoming data is very large (see Chapter 22). The real- ization of a highly integrated environment depends on the solution of technological challenges, such as integrating information from multiple data sources and making the integrated information accessible to professionals when, where, and in the form that it is needed. Integration of medical and biological information also encompasses social issues, such as defining the appropriate role of computers in the workplace, resolving questions of legal liability and ethics related to biomedical computing, and assessing the effects of computer-based technology on health care costs. The chance that our

829

hypothetical scenarios will become reality thus depends on the resolution of a number of technological and social issues that will be debated during the coming years.

24.1.1 Looking Back to 1990

In the first edition of our book, the closing chapter included two future scenarios of medical care and discussed emerging topics such as the Unified Medical Language System (UMLS), integrated academic information management systems (IAIMS), and the medical information bus (MIB). Today the UMLS is employed in information retrieval systems as a tool for converting textual medical information into standardized terms taken from coding schemes and terminologies such as MeSH, SNOMED, and ICD and to help translate from one vocabulary to another (McCray and Miller, 1998).

IAIMS sites are now scattered around the country with many different models being implemented. The MIB has been approved as the IEEE 1073 family of standards for medical device interconnection (Stead, 1997b) and has been incorporated into multiple instruments at the bedside (see Chapter 17).

The scenarios we discussed in the first edition included computer-based support dur- ing both cardiac bypass surgery and long-term care of a patient with a chronic disease.

Although the information-support capabilities have changed considerably in the decade since the first version of this chapter was written, it is possible that the practice of med- icine has changed just as much. For example, less invasive alternatives to open-chest bypass surgery have become more common. Stricter criteria for admission to the hospi- tal and shorter lengths of stay once hospitalized mean that sicker patients are routinely cared for in outpatient settings. In such situations, the need for computer-based track- ing of a patient’s medical status is increased. This need has led to experiments such as the use of wireless pen-based computers by home health care nurses for logging patient conditions and Internet-based disease management interactions between clinicians and patients in their homes. Major attention to data protection and patient data confiden- tiality has significantly altered the technological solutions to such data management and access tasks (see Chapter 10).

Significant advances have been made in raw computing power (e.g., hardware and software for the manipulation of three-dimensional images); interconnectivity (e.g., high- speed network backbones and wireless connections to palm-sized handheld computing devices); the ability to store very large amounts of data (e.g., the terabyte data storage device shown in Figure 24.1); and the development of infrastructure—particularly in the area of communication standards (e.g., health level 7 (HL7) and object broker architec- tures). On the other hand, the anticipated level of seamless integration between applica- tions, highly interconnected medical databases with embedded decision-support tools, and ubiquitous computing support have remained elusive but are an increasing focus of policy as well as technical emphasis.

24.1.2 Looking to the Future

During testimony before the U.S. House of Representatives Committee on Science in

1997 concerning the future role of the Internet, one of the authors (EHS) laid out a set

of long-term goals for medical informatics. Like the scenarios we depicted nearly a decade ago, these goals depend on the occurrence of both technical and social changes for their fulfillment. If the assumptions identified in Section 24.1.3 prove valid, medical practice in the future will incorporate aspects of the following scenarios:

●

Low-cost, high-quality telemedicine: The telemedicine experiments of the mid-1990s

were dependent on specialized equipment and expensive special-use communications

lines. This has evolved such that the Internet is a common vehicle for linking medical

experts with other clinicians and patients at a distance (National Research Council,

2000; Shortliffe, 2000). In the future, the Internet will be able to support clear video

images routinely, with high-fidelity audio links to support listening to the heart and

Figure 24.1. Multi-terabyte mass storage in a tape robot facility. (Source: Reprinted by permis-

sion from StorageTek, Louisville, CO. 1998.)

lungs, and common computing platforms at both ends of the links to make telemed- icine a cost-effective form of medical practice. Patients will avoid unnecessary travel from rural settings to major medical centers, primary care clinicians will have expert consultation delivered to them in their offices in a highly personalized fashion, and patients will accomplish in single office visits what now often requires multiple visits and major inconvenience. There is reason to believe that such applications will become commonplace soon, with several successful demonstration projects under way to demonstrate cost-effectiveness and a positive benefit for patients (see Chapter 14).

●

Remote consultation: Quick and easy electronic access between clinical providers to dis- cuss patient cases will improve access to expert patient care and enhance patient satis- faction. For example, an attending physician, residents, and medical students in a community clinic who treat a patient with an unusual skin lesion will obtain immediate teleconsultation with a dermatologist at a regional medical center. The remote medical team will learn from the dermatologist, the expert will receive clear, diagnostic-quality images of the lesion, and the patient will promptly receive a specialist’s assessment. All too often today, patients, when referred to major centers, experience significant delays or fail to keep their appointments due to travel problems. Instead of sending patients to the experts, we will improve their care by using the Internet to bring the experts to them. Current demonstration projects in a handful of locations have shown the feasi- bility of such remote access to expertise; it remains to make the applications common- place, well integrated with routine care, and generally accepted as reimbursable clinical activities.

●

Integrated health records: We envision the day when citizens no longer will have mul- tiple records of their health care encounters scattered throughout the offices of numerous physicians and the medical record rooms of multiple hospitals. Instead, their records will be linked electronically over the Internet so that each person has a single “virtual health record”—the distributed, but unified, summary of all the health care they have received in their lives. Furthermore, this record will be secure, treated with respect and confidentiality, and released to providers only with the patient’s per- mission or during times of medical emergency according to strictly defined and enforced criteria (National Research Council, 1997). Important steps have been taken recently to make this scenario more likely to happen. The development of a National Health Information Infrastructure and adoption of nomenclature standards such as SNOMED-CT, and privacy rules such as Health Insurance Portability and Accountability Act (HIPAA) are contributing toward this goal. In 2004, President Bush announced a federal goal to implement ubiquitous electronic health records for all citizens within a decade. In addition, the U.S. armed services are actively pursing electronic dog-tags that will contain abstracts of the electronic medical record of each participant in the military.

●

Computer-based learning: Soon, medical students on their orthopedics rotation,

preparing to observe their first arthroscopic knee surgery, will be able to go the

school’s electronic learning center and use the Internet to access and manipulate a

three-dimensional “virtual reality” model of the knee on a computer at the National

Institutes of Health. They will use new immersive technologies to “enter” the model

knee, to look from side to side to view and learn the anatomic structures and their spa-

tial relationships, and to manipulate the model with a simulated arthroscope, thus get- ting a surgeon’s-eye view of the procedure before experiencing the real thing. Using the experimental Next-Generation Internet, remote access to medical dissections is becoming available in a limited way. In the near future, many hard-to-learn proce- dures, such as female pelvic examination, will be routinely taught to medical students on a computer simulation/mannikin with immediate feedback rather than with patients or living models.

●

Patient and provider education: Health science schools are starting to provide distance- learning experiences via the Internet for postgraduate education, refresher courses, and home study by health science students. Eventually, clinicians will be able to pre- scribe specially selected video educational programs for patients that will be delivered to home television sets by a direct Internet connection. Our hospitals and clinics will use video servers over the Internet not only to deliver such materials to patients but also to provide continuing medical and nursing education to their staffs. Providers are beginning to make patient-oriented versions of the electronic medical records avail- able online along with relevant online information tailored to the patient’s problems (Cimino et al., 2002).

●

Disease management: High-speed Internet access via Digital Subscriber Line (DSL), cable, or satellite is now being offered to most families in the United States and more than 50 million households now have broadband connectivity. Soon, clinicians will move beyond the simple use of telephones for managing patient problems at a dis- tance to using their visual senses as well via two-way video links. The infirm will receive “home visits” via video links, thus avoiding unnecessary office or emergency room visits, and care managers will have important new tools for monitoring patients that emphasize prevention rather than crisis management. Early experiments show remarkable enthusiasm by patients when familiar physicians and nurses provide such videoconferencing interactions in the home (see Chapter 14).

Over these last 5 years since these goals were elucidated, we have made significant progress toward meeting these long-term objectives for computer-assisted learning, provider and patient communication, and medical care utilizing high-speed networks and fast, commodity computers. Similar changes have been taking place in the collec- tion, interpretation, and dissemination of biological information. For example with the advent of very fast, multiprocessor supercomputers, researchers can begin to model bio- logical processes such as the folding of a protein or the binding of a drug to a receptor site. Some of this modeling work can be done using the high-speed Internet and distant computers, such as the biologically focused supercomputers in San Diego, California, and Pittsburgh, Pennsylvania. One scientist working in this area has suggested some of the major challenges and opportunities facing the field of computational biology:

●

A computational model of physiology: Can we create a simulation of the human body and estimate of the effects of medications on the diseased and nondiseased portions

1

R.B. Altman in a presentation entitled “Final thoughts: Further opportunities in bioinfomatics and compu-

tational biology”, presented at a conference on Bioinformatics Methods and Techniques, Stanford, CA. June

23–25, 2003.

of the body with sufficient fidelity to avoid most animal and early human testing of drugs? Having this simulation model would bring a tremendous benefit by reducing the number of years of testing that occurs for most drugs. However, the complexity of the task is daunting. Much of the pathophysiology needed to build this model is unknown, and even the parts that are known would create such a complex set of rela- tionships that the computers models may be intractable given near-term computa- tional capabilities. Furthermore, the genetic variation between individuals that is being studied in pharmacogenetics experiments greatly increases the complexity of the modeling process.

●

Design of new compounds for medical and industrial use: Can we design a protein or nucleic acid to have a specified function? The determination that a particular drug can be used to treat a medical condition has traditionally been done by testing a large col- lection of substances in the laboratory to see if any show in vitro activity. This step is then followed by extensive animal testing. Now that we have a better understanding of protein structure, and a clearer model of how to modify the disease process, can the drug creation process be switched to build biological custom materials to reverse or deter a pathological process?

●

Engineering new biological pathways: Can we devise methods for designing and imple- menting new metabolic capabilities for treating disease? The biological metabolic pathways of various species are being mapped out quite rapidly. It is interesting to observe the variation from species to species in pathways that perform similar meta- bolic functions for the animals. This suggests the possibility of building new meta- bolic pathways in areas such as inborn metabolic diseases. Some diseases, such as sickle cell anemia, have a single flaw that must be overcome. In other genetic diseases multiple elements of the pathways may be missing, and there will be a need to con- struct an alternative pathway that takes into account the particular manifestations of the disease.

●

Data mining for new knowledge: Can we ask computer programs to examine data (in the context of our models) and create new knowledge? As we create large databases of measurements taken during clinical care, the question arises about finding new pat- terns in those data. Exactly what effects does a drug have on various laboratory tests and measurements? If we have enough data collected across different patients but in similar situations, will we have enough statistical power to recognize unknown rela- tionships? A large number of statistical approaches are being employed to perform a structured analysis of large data sets to learn new relationships.

These biological challenges are still likely to be unanswered as we approach the 10th edi- tion of this textbook. Because they require the development of considerable biological knowledge, computational techniques, and new methodologies for analysis, they likely represent distant goals of biomedical computation.

24.1.3 Assumptions Underlying the Scenarios

To help you to evaluate these scenarios of the future, we must make explicit the assump-

tions on which these speculations are based. In particular, we assume that health care

workers and life scientists will work increasingly with computers in their daily lives and that improvements in computer technology will continue, independent of technological advances in biomedical science. Furthermore, we assume that concerns about health care cost containment and the threat of malpractice litigation will continue to be unre- solved issues for the near term.

The technological development of medical computing depends in large part on advances in general computing capabilities. Except in the area of medical imaging, little computer technology is first developed for medical applications and then applied to the rest of industry. This is especially true now that a few general-purpose microprocessors and operating systems have become standard for all personal computers. Specialized computer chips will continue to be created for computationally intensive medical appli- cations, such as signal processing. In these image-processing applications, rotation, fil- tering, enhancement, and reconstruction algorithms must handle more data than can be processed with the standard microprocessors; thus, a market exists for specialized machines.

It is difficult to predict whether the development of new general-purpose computer products will continue to follow an evolutionary trend or will undergo a paradigm shift—defined by Kuhn in his book on the nature of scientific discovery as a complete change in perspective, such as occurred with the revelation that the Earth is not flat (Kuhn, 1962). Computer processing has gone through some major shifts in direction over the last 40 years: from single-user batch processing, to timesharing on a central resource, and then back to single-user processing, this time on local machines with access to specialized machines through a network. As more and more people try to access key Internet sites, we have moved back to a version of the timesharing model of 30 years ago.

The human–machine interaction style has changed dramatically, with graphical inter- faces for novices almost completely replacing command-line interfaces. Pen, speech, and three-dimensional interfaces have been built but have not been widely deployed, except in the case of pen-based datebook applications and spoken systems for very defined tasks such as airline reservations or banking. For example, Figure 24.2 shows one approach for using three-dimensional representations for literature retrieval. This appli- cation built at the Palo Alto Research Center (PARC) uses the three-dimensional spa- tial information to show complex trees in much greater detail than in most two-dimensional layouts. There also has been a significant change from electrical to optical methods of network transmission (fiberoptic cables) as well as a steady progres- sion from analog to digital recording of information, best illustrated by the switch from film to digital images in radiology, the widespread introduction of high-speed fiber net- works, and the development of satellite communication networks. For the purposes of this chapter, we shall assume the continued progression of current trends rather than a significant paradigm shift. An unanticipated discovery that is just around the corner could, of course, quickly invalidate the assumptions.

Another subtle, but pervasive, assumption underlies this book, much of which is writ-

ten by researchers in biomedical informatics. We tend to believe that more technology,

thoughtfully introduced, is usually better and that computers can enhance almost any

aspect of clinical practice and biomedical research—especially information access, the

diagnostic and therapeutic components of the decision-making process, and the analy- sis of huge amounts of biological data. For example, paper documents are still the mainstay of medical records in many medical settings. Still, we strive to eliminate the paper-based components and assume that a well-designed interface (e.g., one that allows handwritten or continuous-speech input) applied to a sufficiently fast computer can significantly improve the overall process of recording and retrieving clinical data.

Although this assumption has yet to be formally verified, there is a groundswell of development and investment activity based on successful experiments that encourage the belief that electronic records will positively transform the way in which we provide patient care and monitor health.

People frequently criticize medical professionals for being technocrats—for encour-

aging an increase in mechanization and electronic gadgetry that tends to alienate both

workers and patients. Such increases fuel the concern that modern medicine is becom-

ing increasingly impersonal and sterile. How do we meld the automated environments

proposed in this chapter’s introductory scenarios with our wistful memories of kindly

family doctors making house calls and attending their patients’ weddings, christenings,

bar mitzvahs, and the like? The reality, of course, is that the trend away from such tra-

ditional images predated the introduction of computers in medicine and has resulted

more from modern pressures on health care financing and the need for subspecializa-

Figure 24.2. Three-dimensional representation of the MeSH tree using the Information

Visualizer Toolkit created at Xerox PARC. A portion of a citation about breast cancer is shown

in the foreground. (Source: Reprinted from Hearst/Karadi, SIGIR’97, courtesy of ACM.)

tion to deal with an increasingly complex subject area. The role for computers and other information technologies results as much from these pressures as it does from a blind faith that all technology is good, useful, and worth the associated costs. Many argue, in fact, that the prudent use of computing technology will introduce the kinds of efficien- cies needed if clinicians are to return to an era in which more relaxed time spent with patients is feasible.

Once medical computing applications have been shown to be effective, the technolo- gies will need to be evaluated carefully and consistently before their routine adoption.

We need to know that the benefits exceed the costs, both financial and sociological. The debate about where and how computers should be used is even more complex in devel- oping countries, where advanced technology might partially compensate for shortages in medical expertise but where scarce health care resources might be more effectively employed to provide sanitation, antibiotics, and basic medical supplies. Nonetheless, the scenarios above were painted with the assumptions that there will be an increased appli- cation of computers in all aspects of medicine and that the key difference between the future and today will be that computers will become ubiquitous and that they will have a high degree of interconnection and an increased ability to interoperate.

24.2 Integration of Computer-Based Technologies

Most of the individual capabilities described in the preceding scenarios exist today in pro- totype form. What does not exist is an environment that brings together a large variety of computer-based support tools. The removal of barriers to integration requires both tech- nological advances, such as the development of standards for data sharing and commu- nication (see Chapter 7), and a better understanding of sociological issues, such as when computer use may be inappropriate or how the need for coordinated planning can over- come logistical barriers to connecting heterogeneous resources in a seamless fashion.

We can begin to assess the degree of connectivity in a medical center by asking sim- ple questions. Can the laboratory computer communicate results to the computer that provides decision support, without a person having to reenter the data? Do the pro- grams that provide decision support use the same terms to describe symptoms as do those that professionals use to perform electronic searches of citation databases? Do physicians and other health personnel use computers to get information without think- ing about the fact that they are using a computer system, just as they pick up a medical chart and use it without first thinking about the format of the paper documents?

Computer systems must be integrated into the medical setting in three ways. First,

applications must fit the existing information flow in the settings where they are to be

used. If the machine sits in a corner of the clinic, out of the normal traffic flow, and if

there is another way to accomplish the specific task, then the computer system is likely

to be ignored. Likewise, programs that arbitrarily constrain physicians to unnatural

procedures for entering and accessing information are less likely to be used. User inter-

faces should be flexible and intuitive; just as the fields of a paper medical form can be

completed in an arbitrary order, data-entry programs should allow users to enter infor-

mation in any order.

Surgeons attempting so-called telepresence surgery over the Internet, bringing special- ized expertise to an operating room possibly hundreds of miles away, will be unable to assist in the procedure if the movements they make with hand devices at one end are not instantly reflected in what they see happening with the actual instruments at the other end of the link. How do we ensure interoperability across the many networks that now span our coun- try (National Research Council, 2000)? Can we guarantee adequate response time for the telesurgery application not only on the major backbone networks but also on the last seg- ments of wire, cable, or wireless network that come into offices and other remote settings?

Second, computer systems should provide common access to all computer-based resources, so a user cannot tell where one program ends and another starts. In this book, we have described such diverse applications as computed tomographic scanning and bibliographic searching. Many of these systems have been developed independ- ently, and most are completely incompatible. In the future, the radiologists’ picture archiving and communications system (PACS) workstation (see Chapter 18) should deliver more than just images—for example, a radiologist may wish also to search eas- ily for references on unusual presentations of a specific disease process and to include these references in a paper being composed with a text editor. Ideally, users should not have to switch between computers, to stop one program and to start another, or even to use different sets of commands to obtain all the information they need. That the desired information resources may exist on multiple machines in different parts of the medical center or the country should be invisible to the user.

Third, the user interface must be both consistent across applications and easy to use, which may require multiple interface modalities, such as pointing, flexible spoken natural-language interfaces, and text input. Both at the user interface and internally, programs should use a common terminology to refer to frequently used concepts, such as a diagnosis, a symptom, or a laboratory test value.

We are seeing increasing amounts of medical information packed into smaller, more powerful computers, such as hand-sized personal digital assistants (PDAs). The config- uration of computers is starting to change. Figure 24.3 shows a computer system that is worn attached to the body. Using spoken input or a keypad mounted on the arm along with a heads-up display, the computer is inherently as mobile as the person using the system. Although this device may seem far in the future for medical care, a less sophis- ticated version of this equipment is used everyday in the rental car business to check-in returning cars and print receipts. Although the rental car business is far more structured than medical practice, the example shows that this type of technological change can be successfully integrated into the workplace.

Figure 24.4 is a fanciful figure from Wired Magazine that hints at the future effects of

nanotechnology. Technology is not quite at the Fantastic Voyage level where miniature

robots flow through the body repairing problems, but the field of microelectromechan-

ical systems (MEMS) has created methods to build very small sensors and miniature

mechanical devices. In the figure, the illustrator imagines miniature devices being used

as an intervention to remove “smart dust” household sensors that might have been acci-

dentally inhaled. Certainly, the ability to build sensors and treatment devices at such a

small scale will influence future medical care, especially in the management of chronic

diseases such as diabetes.

24.3 Future Role of Computers in Health Care and Biomedicine

As we discussed in the previous section, fulfillment of the opening scenarios in this chapter will require significant technological changes. Equally important are the orga- nizational and attitudinal changes that will be necessary to implement the new tech- nologies as they emerge. Health professionals, health institutions, medical system developers, and society as a whole must carefully consider the appropriate role of com- puters in medicine and assess the potential benefits of computers in terms of improved access to information, enhanced communication, increased efficiency of health care delivery, and higher quality of medical care.

Although the potential benefits of using computers are many, there are also potential costs, only some of which are monetary. For example, computer-based medical record systems will never exactly replicate the flexibility of current paper-based systems. This flexibility includes the ability to create progress notes about patients, using any words in any order and in any format, with or without diagrams, to record the information.

Figure 24.3. A wearable computer, including monacle display, voice input, belt-mounted central processing unit, and hand-mounted keypad. (Source: Photograph by Lawrence M. Fagan.

Xybernaut, Mobile Assistant, and MA IV are registered trademarks of Xybernaut Corporation.)

Computer-based systems limit flexibility in return for increased legibility and access to the information and for the ability to use the information for other purposes, such as clinical research studies that use multiple patient databases. We see the same pattern in the use of automated bank-teller machines. There are only a few ways to complete a cash withdrawal transaction through a sequence of button pushes, but there are a large number of ways in which we can make this request of a human teller. Automated tellers are available at 3:00

.

. in the morning, however; human tellers are not.

The idea that computer approaches would require the additional structuring of

medical records was perceived by Lawrence Weed more than 30 years ago (Weed,

1969). Weed noted that, for medical records to be useful, they had to be indexed

such that important information could be extracted. In particular, he proposed that

the medical record be organized according to the patient’s current problems—the

problem-oriented medical record (POMR). Variations of the POMR have become a

standard feature of medical record keeping, regardless of whether computers are

used. In the problem-oriented medical information system (PROMIS), a computer-

based implementation of the POMR, such rigid and time-consuming indexing of

Figure 24.4. Amusing illustration of miniature robotic devices used as a treatment of accidental

inhalation of “smart dust” robotic sensors. (Source: Illustration by Chuck Henderson. Reprinted

by permission from Wired Magazine. This illustration appeared in the April, 2004 issue.)

patient problems was required that clinicians ultimately proved reluctant to use the system. Standardization provides benefits but exacts costs in terms of decreased flexibility; it makes information more accessible but restricts freedom to pursue alternate means to accomplish the same result. It is unlikely that new computer innovations will ever eliminate this trade-off.

Similarly, the use of computers requires trade-offs with respect to confidentiality of medical information. Legitimate users can more conveniently access computer-based records in well-designed systems. Without sufficient security and policy measures, how- ever, unauthorized users may threaten the confidentiality and integrity of databases.

Fortunately, with adequate attention to security issues, modern methods backed by effec- tive security policies can ensure that patient data are kept with greater confidentiality in computer systems than they are in the paper charts of hospital wards. The growing vol- ume of clinical data stored electronically, widespread remote access capabilities, and the trend toward secondary and tertiary uses of clinical information result in the need for ongoing attention to concerns of security and confidentiality (see Chapter 10).

Earlier fears that computers could replace physicians have not been borne out, and computers are likely to remain decision-support tools rather than substitute decision makers (Shortliffe, 1989). It is more likely that computers will be used increasingly to monitor the quality of health care delivered and to help evaluate physician performance.

Greater automation will therefore change the nature of medical practice in nontrivial ways. The challenge for system developers and users will be to identify the solution that provides the optimal balance between flexibility and standardization.

Computers in the future are likely to have even a greater influence on the practice of biology in the research laboratory. Exponential increases in the availability of experi- mental data necessitate computational support. Examples include cross-species investi- gation of DNA sequences, analysis of microarray data, and simulation of cellular function. Just as clinicians need to adjust their practice to accommodate computational support, biologists need to provide a computational infrastructure for the collection, analysis, and dissemination of laboratory data. An exceptional illustration of the power of computers is in the field of pharmacogenetics, where clinical and biological data are combined to determine individual responses to drug therapy. This requires the develop- ment of terminologies to support these disparate data sources, a complex database schema to store the information, and intelligent searching function to highlight the key relationships. It is likely that these types of biological and clinical applications will be a focus of research and development in the coming decades.

24.4 Forces Affecting the Future of Medical Computing

In this book, we have identified several important factors that affect the current and future

role of computers in medicine. These factors include advances in biotechnology and com-

puter hardware and software, changes in the background of health professionals, changes

in the medicolegal climate, and changing strategies for health care reimbursement. The rel-

ative strengths of these forces will determine how likely it is that the scenarios we proposed

will take place and how quickly we can expect such changes to occur.

24.4.1 Changes in Computers and Biomedical Technology

Modern computers are smaller, less expensive, and more powerful than were their pred- ecessors. Although microchip designers are approaching the physical limits on how close together electronic elements can be placed, these trends will continue. By the time a new microprocessor chip or memory chip is adapted for use in new computer systems, manufacturers are creating samples of the next generation of chips. The most important ramification of current trends in microcomputer technology is that it is now possible to include a microprocessor and memory in most pieces of medical equipment.

The ability to connect multiple devices over high-speed networks has enabled a dra- matic change in the way medical computer systems are designed. It is now practical to develop systems in which multiple data storage devices are accessed by complex comput- ers that sort and abstract the numerous patient data that are generated. This design con- trasts with the traditional approach in which health professionals would need to gather information from many different devices or locations in the medical center to obtain a complete picture of a patient’s status. Relatively low-cost computer workstations or net- work-based terminals now allow information to be manipulated at each patient’s bedside, as well as at other work areas in medical centers and ambulatory clinics.

High-speed networks and private intranets allow physicians in private offices to con- nect to computers in the hospitals where they admit patients or to repositories of data from multiple care settings and systems.

We have much more data to transmit over the network, such as those from digitized radiology images, online reports, and computer-based charting of the patient’s condition.

In the future, even faster networks and larger storage devices will be necessary to manage the overwhelming volume of data that will be created by all-digital clinical data systems.

24.4.2 Changes in the Background of Health Professionals and Biologists

Computers will continue to be made faster and less expensive and will have more fea- tures; however, sufficient computing power now exists for most applications. Thus, the limitations on the pervasiveness of computers in medicine do not hinge as crucially on the development of new hardware as they did in the past. The availability of relatively inexpensive and powerful computers is changing health care workers’ familiarity with machines by exposing these people to computers in all aspects of their daily lives. This increased familiarity in turn increases acceptance of computers in the workplace, another crucial determinant of how computers will fare in the next 15 years.

Since the early 1970s, people have increasingly interacted with computers in their

daily lives to perform financial transactions, to make travel arrangements, and even to

purchase groceries. In many situations, people do not use the computer themselves but

rather talk to an intermediary, such as an airline reservation agent or bank clerk. Large

computer systems are so deeply integrated into many business practices that it is not

uncommon to hear, “I can’t help you—the computer is down.” This switch to computer-

based record keeping for most financial transactions is so pervasive that you would be

surprised and concerned if you were to receive a monthly bank statement that was writ-

ten out in longhand. Within the last 10 years, many mediated computer-based transac- tions have been replaced by direct contact between a consumer and a computer system.

We can withdraw cash from automated bank tellers, request a trip routing from a com- puter at the rental car stand, or obtain an account balance by a touch-tone telephone.

As described in almost every chapter in this book, the acceptance of the Internet model of interaction is cutting out intermediary steps and allowing users to access large online databases directly from home or work.

Young health care workers today have been exposed to computers throughout their edu- cation. Many college courses assign projects that must be carried out using the computer.

It is difficult to determine how the current hodgepodge of computers in the medical setting will bias these users. It may be that, when the integrated system becomes available, health care professionals will not use the system fully because of previous negative experiences in less sophisticated environments. On the other hand, familiarity with computers and with their operation may prepare users to accept well-designed, easy-to-use systems.

24.4.3 Legal Considerations

The number of malpractice lawsuits and the sizes of the settlements have increased in recent years. Today, the specter of potential legal action hangs over every medical inter- action. Computers can either exacerbate or alleviate this situation. The computer-based diagnostic system may provide a reminder of a rare but life-threatening disease that might have been overlooked in the differential diagnosis. On the other hand, decision- support systems might generate warnings that, if ignored by health care workers, could be used as evidence against those workers in a court action (see Chapter 10).

During the last few years, there has been a national-level focus on the cause of med- ical errors during the diagnostic and treatment process. The Institute of Medicine (IOM) has performed several key studies in this area (Kohn, 1999; Committee on Quality of Health Care in America, 2001; Aspden, 2003). These studies showed a series of issues with the medical process that can lead to medical errors. One of the results of these studies has been new requirements to track errors and near misses.

The second IOM report suggested that the number of errors would be reduced by the widespread introduction of information technology. The 2003 IOM report laid out possible data standards that would help to increase patient safety. In 2004, the gov- ernment increased its efforts to promote a National Health Information Infrastructure that will provide mechanisms to interlink medical records at diverse institutions, and thus decrease the possibilities of errors or repeated tests because of increased access to clinical information.

24.4.4 Health Care Financing

Some people assume that the continuing evolution of computer hardware and software

is the most important force influencing the development of medical computing. Social

issues such as health care financing and the legal aspects of medicine, however, proba-

bly outweigh the technological factors. Perhaps the strongest force at work today is

the pressure to control health care costs. Health care financing influences all choices

regarding the acquisition and maintenance of high-technology equipment and information systems.

Current schemes for health care financing have been designed to slow the rate of growth of health care costs. These policies translate into pressure to reduce costs in every aspect of medical diagnosis and treatment, such as to substitute ambulatory care for hospitalizations, to shorten hospital stays, to select less expensive surgical proce- dures, and to order fewer laboratory tests. As incentives for making optimal decisions in these areas increase, there is a greater need for computers that can collect, store, interpret, and present data during the decision-making process.

Order-entry systems routinely screen test orders against criteria for test ordering and question or cancel tests that do not meet the criteria. A more sophisticated clinical decision-support system might serve as an adjunct to an order management system, assessing which tests are most appropriate to order for specific patients who require expensive workups (e.g., custom-tailored evaluation of thyroid function). Such a sys- tem might also evaluate drug orders and suggest less expensive substitutes that are equally effective while checking for drug–drug interactions, and so on.

It is now common that computers are used by health plans to enforce a particular style of care through concurrent or post hoc review of the medical decision-making process, including decisions about length of stay and tests performed. Thus, one force promoting the use of computers by clinicians is the knowledge that computers are often used by insurers to review these clinicians’ decisions after the fact. The clinicians may prefer to know what the computer system will advise beforehand so that they can be pre- pared to justify intentional deviations from the norm. Increasingly, health care admin- istrators and medical practice directors are requiring use of clinical systems by clinicians to understand and manage clinical interventions and outcomes in response to decreas- ing reimbursement for services and managed care contracting arrangements.

In the United States, an entire new and thriving industry has evolved in response to efforts to manage health care costs more effectively: pharmacy benefits managers (PBMs). Highly dependent on computer technology to manage their business functions, the PBMs manage the prescription benefits for insured individuals, working with retail pharmacies but also providing mail-order options and encouraging appropriate conver- sion to generic equivalents that can greatly reduce the cost of care for insurers (and for the employers who are the PBM’s clients). Note that the PBMs, like managed care organizations, must distinguish between their customers (generally employers) and their members (employees and their families who receive their medical insurance through the employee’s work). PBMs are only one example of new industries that have evolved in part because of the unusual nature of health care financing in the United States when compared with other parts of the world (see Chapter 23).

Although the use of computer technology can help health professionals to cope with

the growing complexity of medical practice, it also contributes to the increasing cost of

health care. Studies are beginning to show the benefits of outpatient computerized

patient order entry (Johnston, 2003). Some computers are so embedded in the clinical

environment, especially in complex environments such as operating rooms or intensive

care units, that it is hard to show a decrease in patient morbidity or mortality due to the

computers alone. Large numbers of devices are already being used in these settings, and

patients are often treated for multiple concurrent problems. Even if patient-monitoring equipment could help health care workers to recognize potentially dangerous situations earlier than they otherwise would, showing that the computer system affected the out- come would be difficult. At the other extreme are devices that can be shown to make a difference, but whose cost is very high, such as computed tomography and magnetic res- onance imaging systems. Because they replace invasive techniques (which have a signif- icant potential for causing harm) or provide information that is available from no other source, there has been little debate about the utility of these new modalities. Instead, the high cost of this equipment has focused attention on how best to distribute these new resources. Demonstrating the return on investment in computer technology is a major challenge to widespread implementation of health care computer systems in light of their often staggering capital cost and the expense of staff resources required to support the many clinical users and to implement, integrate, manage, and maintain increasingly complex distributed systems and networks. No matter how computer systems are used in the future, we will need to evaluate the influence of the computer application on health care financing and to assess the new technology in light of alternative uses of resources.

24.5 Looking Back: What Have We Learned?

An introductory book can only scratch the surface of a field as varied and complex as

biomedical informatics. In each chapter, we have examined technical questions about

how a system works (or ought to work); we must also view each area in light of the

health care trends and the social and fiscal issues that shape the ways in which clinical

care is delivered now and in the future. In this chapter, we have emphasized the rich

social and technological context in which biomedical informatics moves ahead both as

a scientific discipline and as a set of methodologies, devices, and complex systems that

serve health care workers and, through them, their patients. One of the most important

changes is the increase in the use of computers to help with the analysis of biological

data. As more clinically important genetic data come online, new application areas such

as pharmacogenetics are influencing clinical care. This type of new application that spans

clinical and biological data is clearly the start of a new trend in informatics. Glimpses of

the future can be at once both exciting and frightening—exciting when we see how emerg-

ing technologies can address the frequently cited problems that confound current health

care practices but frightening when we realize that methodologies must be applied wisely

and with sensitivity if patients are to receive the humane and cost-sensitive health care

that they have every right to expect. The question is not whether computer technologies

will play a pervasive role in the health care environment of the future, but how we can

ensure that future systems are designed and implemented effectively to optimize tech-

nology’s role as a stimulus and support for the health care system and for individual

practitioners. The outcomes of the process will depend as much on health care planners,

practitioners, and policymakers as they will on the efforts of system developers and bio-

medical informatics professionals. It is to all such individuals that this book has been

dedicated.

Suggested Readings

Altman R.B. (2001). Challenges for intelligent systems in biology. IEEE Intelligent Systems, 16(2):

2001, 14–18.

In this summary article, R.B. Altman lays out some major future challenges for biocomputation.

http: //www.citl.org/research/ ACPOE_Executive_Preview.pdf (Web site accessed July 9, 2004).

This Web site provides an executive summary of the cost/ benefits of Ambulatory Computerized Provider Order Entry. See Johnston (2003) for the full report.

Committee on Quality of Health Care in America, Institute of Medicine, Crossing the Quality Chasm: A New Health System for the 21st Century. Washington, D.C.: National Academies Press.

This report discussed structural changes to the practice of medicine to address some of the issues discussed in the IOM report, To Err is Human.

Stead W.W., et al. (Eds.) (1998). Focus on an agenda for biomedical informatics. Journal of the American Medical Informatics Association, 5(5):395–420. Special Issue.

The 1998 Scientific Symposium of the American College of Medical Informatics (ACMI) was devoted to developing visions for the future of health care and biomedicine and a strategic agenda for health and biomedical informatics in support of those visions. The first five articles contained in this special issue illustrate these findings and continue to be timely.

Questions for Discussion

1. Select an area of biomedicine with which you are familiar. Based on what you have learned in this book, propose a scenario for that area that takes place 20 years in the future. Be sure to think about how issues of system integration, networking, and changes in workflow will affect the evolution of computers in the setting you describe.

2. Imagine that you are a patient visiting a health care facility at which the physicians have made a major commitment to computer-based tools. How would you react to the following situations?

a. Before you are ushered into the examining room, the nurse takes your blood pres- sure and pulse in a work area and then enters the information into a computer ter- minal located in the nursing station adjacent to the waiting room.

b. While the physician interviews you, he or she occasionally types information into a computer workstation that is facing the physician; you cannot see the screen.

c. While the physician interviews you, he or she occasionally uses a mouse-pointing device to enter information into a computer workstation located such that, when facing the physician, you cannot see the screen.

d. While the physician interviews you, he or she occasionally uses a mouse-pointing device to enter information into a computer workstation that you both can see.

While doing so, the physician explains the data being reviewed and entered.

e. While the physician interviews you, he or she enters information into a clipboard-

sized computer terminal that responds to finger touch and requires no keyboard

typing.

f. While the physician interviews you, he or she occasionally stops to dictate a phrase. A speech-understanding interface processes what is being said and stores the infor mation in a medical record system.

g. There is no computer in the examining room, but you notice that between visits, the physician uses a workstation in the office to review and enter patient data.

Now imagine that you are the physician in each situation. How would you react in each case? What do your answers to these questions tell you about the potential effect of computers on patient–physician rapport? What insight have you gained regarding how interactive technologies could affect the patient–physician encounter? Did you have different reactions to scenarios c and d? Do you believe that most people would respond to these two situations as you did?

3. You are the medical director of a 30-physician multispecialty group practice. The practice is physician-owned and managed and maintains a tight affiliation with a nearby academic medical center. You are considering implementing an ambulatory medical record system to support your practice operations. Discuss at least eight sig- nificant challenges you will face, considering technology, user, legal, and financial fac- tors. How will you address each issue?

4. Defend or refute the following proposition: “Knowledge-based clinical systems will be widely used and generally accepted by clinician users within the next 5 years.”

5. You are asked to design a pharmacogenetic system that can help to understand which patients will respond to particular medicines. You need to design a database of studies that relate descriptions of the medical conditions, medications, and genetic sequences.

Which terminologies would you choose for each of these data types? What are some key

elements that would need to be included in the database schemas? What problems do

you anticipate having in the design of the data structures for this task?