Emerging Needs and Opportunities for Human Factors Research (1995)

Roberta L. Klatzky, M.M. Ayoub

In considering human factors research needs related to health care, we are particularly concerned with new research topics that are likely to reflect societal and technological developments in the coming decade. We also consider more long-standing problems that previous research has failed to address. In keeping with the general goals of this volume, our goal is to highlight problems in which human capabilities and processes play a critical role and thus to identify areas in which human factors researchers can make a significant contribution. Health care consumes a large and growing proportion of the U.S. gross national product; new developments have consequences for the quality and duration of many lives. Even a small change in the effectiveness of workers, medical devices, or care-giving environments can translate into a large impact on costs and human comfort.

Health care has not been a traditional focus for human factors research. This was noted two decades ago in a report on an international NATO symposium (Pickett and Triggs, 1974), whose goal was to provide examples of how human factors techniques could be applied to medical issues. In an introductory overview to that report, Rappaport cited a variety of contributions that human factors specialists could make to the health care field (see also Rappaport, 1970). He foresaw the increased use of computers in medicine and suggested that human factors specialists might interact directly with manufacturers of medical equipment to improve designs.

These seminal efforts notwithstanding, human factors researchers have not been greatly involved in the area of health care. For example, despite the long-standing concern of human factors specialists with sources of error, a symposium on human error in medicine held at the 1991 meeting of the Human Factors Society (Bogner, 1991) appears to have been the first of its kind (Van Cott, 1991). The dearth of attention to the potential contributions of human factors research is unfortunate; the field has much to offer with respect to designing medical devices, human-machine interfaces, and medical environments such as operating rooms and laboratories. Another area of application is training—not only of medical personnel but also of patients and home caregivers who use medical devices. Further, because of rapid technological change, the skills of medical personnel and of others working with medical devices may require frequent updating.

While we emphasize new research needs in this chapter, we should not forget that the health-care arena offers ample opportunities to apply already well-known principles of human factors. Medical errors point to the need for such application; there is a huge number of contexts in which error can and does occur.

Operating rooms are one example. There the medical practitioner might be compared to a member of a flight crew, as many of the same human factors issues apply. There is a necessity for teamwork, often with critical timing constraints (Helmreich and Schaefer, 1994; Regnier, 1993). Good communication among team members is imperative. There can be peaks of high stress, which are likely to coincide with points when timing is particularly important. The workplace is complex, with multiple instruments giving independent readings that must be integrated by the user. And of course, errors resulting from these sources may have dire consequences. Bogner (1991) has cited the substantial number of potentially preventable incidents in which anesthesia has resulted in brain damage or death.

The administration of drugs is another area in which errors frequently occur. Many medical errors within the home can be traced to inadequate instruction in the use of medicines or devices or to the complexities of labeling. Although steps to reduce such errors could be taken on the basis of existing human factors principles, it is also important to evaluate the target context carefully; this may call for further research (Cook, 1991).

The plan of the chapter is as follows. We first discuss some general technological and societal trends that are likely to have an impact on medical practices in the future and consider the human factors implications of these trends. We then describe, in more detail, several broad areas that are likely to raise specific issues for human factors research: enhanced understanding of risk factors and its impact on programs for behavioral change and disease prevention, advances in medical information technology, technological advances in medical instrumentation, and ergonomic issues that

arise in health care procedures. Clearly, we cannot hope to introduce every potential area of application for human factors research in medicine. The sample we do present is intended to demonstrate the importance of increased involvement in the medical field by human factors specialists. A recurring theme in this discussion is how technological advances in medical care create new needs for human factors research. Although new health care needs may arise independently from technology—for example, through demographic or social trends or through new understanding of requirements for device design—responses to these needs increasingly reflect today's highly computerized environment.

A number of trends in medicine indicate ample opportunities for contributions from human factors research in the coming decade. Some of these trends are related to the development of new health care needs, whereas others are related to new tools, particularly tools made possible by advances in electronics.

As the relatively numerous post-World War II generation ages, it will inevitably place greater demands on the medical system, and the concomitant increase in demand for health care by that aging population is likely to raise a number of issues that human factors research can address. (Human factors research needs for the aging have recently been reviewed in a publication of the Committee on Human Factors—Czaja, 1990.)

Whereas the human factors industry has been extensively concerned with the design of workplace environments, there is a need for research on the design of appropriate domestic environments for the aged population. One goal should be to minimize the risk of accidents, which involve the elderly disproportionately (Czaja, 1990). A broader question is how environmental design might promote good health, both physical and psychological. The answer should take account of the effects of communal living, which seems to be increasingly likely in the face of high housing and health costs. The goal of constructing a living space that encourages social interaction while maintaining individual privacy is a challenging one. Another problem is how dwelling designs might encourage exercise in an otherwise sedentary population without imposing undue physical stress.

Aging also introduces problems of compliance with health recommendations, for example, difficulty in adhering to an adequate diet or remembering to take pills. These problems may be significantly ameliorated by simple external reminder systems that aid what has been called prospective

memory. The human factors area has contributed to research on sensory, biomechanical, and psychomotor effects of aging (see Small, 1987). There are also cognitive changes that can fruitfully be studied from a human factors perspective with the goal of improving the health practices of an aging population.

It is also important to note that the current population of middle-aged adults is different from its predecessors: it is healthier, better informed, and more interested in health care. Many patients today seek to play an active role in prevention and medical treatment. The trend toward increasing patient participation in medical decision making raises many important questions: How can we inform patients about risk factors, preventive measures, and treatment options? How can we aid their decisions? We will consider these issues in more detail below.

An event of substantial societal impact is the emergence of highly infectious and/or untreatable diseases. The prospect of disease transmission in the workplace indicates an important problem for the human factors area, and medical personnel are particularly vulnerable. Interactions between patients and health care workers are coming under increasing scrutiny because of highly publicized cases of disease transmission. Clearly, there is a call for measures to safeguard the welfare of both groups, and human factors research should be fully involved in the development of such safeguards.

The increasing cost of health care is another trend, one outcome of which is decreased hospitalization and, concomitantly, increased home care. More and more frequently, home care incorporates medical devices that may be quite complex to maintain and operate. Human factors concerns should include the design of these devices, the nature of instructions, how individuals are trained to use them, and how stress might alter user competence.

For a substantial number of individuals in the United States, neither home care nor hospitalization is an alternative. Homeless people combine a high propensity for medical needs with a low ability to pay for treatment. They may be shut out of the medical system, and when those who find medical help are given recommendations for treatment, they may be unable to purchase needed drugs or conform to medical regimens. What human factors research can contribute is less clear in this area than in others such as workplace or instrument design. Yet it is in keeping with the goals of this volume that we mention the need to create avenues for diagnosis and treatment among those least able to seek help.

Technologically advanced tools for health care include not only instruments for treatment of disease and trauma, but also systems for communication among medical researchers, practitioners, and supervisory agencies. Computer technology has created such diverse phenomena as new methods for medical imaging, large-scale databases related to patients and treatments, on-line bibliographies, automated decision aids, and computer-controlled surgical devices. As was noted above, new medical devices also make home treatment possible for patients who would once have required care in a health facility. New technology for assisting and retraining the disabled is discussed in Chapter 3.

As is generally the case, technological advances have produced new tools faster than research can be done on human factors related to their use. We need to understand how to make new techniques and devices as effective and accessible as possible.

In both the United States and abroad, individuals are increasingly seeking control of health care practices. This reflects a growing knowledge of risk factors and greater interest in taking preventive steps. The computer has had significant impact on people's knowledge of health risks, a development that is in keeping with our general theme of research needs based on technological advances.

One role of the computer in medicine has been to provide data about patient populations, giving rise to assessments of risk factors and prognostic factors in disease. For example, a very large database has been compiled by the Surveillance, Epidemiology, and End Results (SEER) program of the National Cancer Institute. Combined with data from the National Center for Health Statistics, which provides mortality rates, the SEER data provide direct assessment of cancer incidence and mortality over time. These data have been used to study a variety of issues related to etiological factors in cancer. Another example of a significant data source is the Breast Cancer Detection Demonstration Project (BCDDP), with over a quarter of a million participants. Data from the BCDDP have led to a technique for absolute risk assessment in breast cancer (Gail et al., 1989). Numerous databases are available from epidemiological surveys and experimental studies in the United States and abroad.

It is increasingly clear that behavioral, demographic, and familial characteristics

are predictors of the risk of incurring disease. Among the most significant risk factors that have been identified are behavioral characteristics, including whether an individual smokes and how much dietary fat he or she consumes. For example, it has been estimated that approximately 75 percent of deaths from cancer are attributable to lifestyle (Williams, 1991). Cessation of smoking could reduce lung cancer deaths by over 75 percent; modification of diet could reduce cancer deaths by over 25 percent (Newell and Vogel, 1988).

Health risks arise not only from personal behaviors and attributes; risks due to substances found in the workplace and natural environment are being identified by both experimental and epidemiological techniques (see Swanson, 1988). On-the-job exposure to radiation is a health risk. Vibration may cause damage to the circulatory and nervous systems and to bones and joints. Increasingly, occupational stress is being considered a risk factor (Levi, 1990; Smith, 1987).

Some newly encountered risk factors pose formidable challenges. These include the risk of AIDS transmission to health care workers from needle sticks or handling of bodily fluids. In communities with particularly high rates of HIV infection, it has been estimated that dental workers may have risk levels equivalent to those of homosexuals (Morris and Turgut, 1990). In one survey of medical residents in training during 1989, more than half those surveyed had HIV patients under their care, and 9 percent reported having been exposed to HIV from a needle stick (Hayward and Shapiro, 1991). Hepatitis B virus is an even greater occupational risk for the medical community, whose members have a risk of infection up to 10 times that of the public (Hadler, 1990).

An approach known as ''health risk appraisal" began some two decades ago in an effort to make use of risk information (for a general review, see Health Services Research, 1987). The general idea behind the approach is that risk factors can be measured for a person (objectively or by self-report), and an estimate of that person's risk of disease can then be derived. A health professional would then communicate this personalized estimate to the patient, along with information about what factors are contributing to risk and what possible lifestyle changes he or she can adopt in response. If the person decides to make the suggested changes, the risk factors should be altered and risk should accordingly be reduced.

Each of the basic assumptions underlying health risk appraisal has been criticized. The adequacy of risk-factor assessment has been questioned, relying as it does on the patient's ability to retrieve personal data and his or her inclination to report them accurately. The models used to derive risk

have also been criticized for making statistical assumptions that may be simplified or inaccurate. In addition, there is still not a great deal of evidence for the effectiveness of risk assessment in inducing risk-reduction behaviors.

These criticisms have not, however, prevented the public from being intensely interested in risk factors and disease prevention. Public interest in reducing cholesterol, for example, has considerably altered dietary habits in the United States. National agencies have used known risk factors such as familial incidence and age to adjust recommended regimens of screening tests for diseases. It seems clear that risk assessment (population and personalized), along with attendant recommendations for behavioral change, will increasingly be part of the American health scene.

Once people are informed of risk factors, preventive practices may take various forms, including controlling toxic substances, designing environments that reduce local exposure, and promoting risk-reducing behaviors. Of these interventions, design of the physical workplace has been a traditional focus of human factors and will continue to be important in reducing exposure to risk factors and occupational hazards. A less obvious contribution to be made by human factors is in helping to change the behaviors of individuals. In particular, an important vehicle for behavioral change is the institution of educational and motivational programs for groups.

Group prevention programs have taken place in educational, community, and work environments. For example, the Pawtucket Heart Health Program attempted to treat obesity in an entire community (Lasater et al., 1991; Carleton et al., 1991). Components of the program included a monthly citywide "weigh in," labeling of items in grocery stores and on menus, blood-cholesterol screening, a student cook-off, and church programs, all facilitated by enlisting a large number of community volunteers. Substantial reductions in serum cholesterol were observed in the initial pilot study. Another example is the "Know Your Body" project, funded by the National Heart, Lung, and Blood Institute and the National Cancer Institute. This attempted to modify risk-inducing behaviors in schoolchildren and was found to be successful in reducing the rate of cigarette smoking several years later (Walter, 1989). Community screening for breast cancer is another success story (Paryani et al., 1990). The movement toward risk-prevention and screening programs in the workplace is growing (Breslow et al., 1990).

One scenario for the creation of a community or workplace prevention program is to begin with controlled trials that demonstrate the effectiveness of a preventive behavior, followed by extensive dissemination of results to enhance public awareness. This has been done in the case of the link

between reduced blood cholesterol and decreased incidence of heart disease. Following one well-publicized prevention study, there was a substantial increase in public acceptance of the beneficial effects of cholesterol reduction (Schucker et al., 1987). This undoubtedly created a receptive audience for community interventions like the Pawtucket program.

There is a clear need for further human factors research on how information about risk and its reduction should be communicated and how these communications are interpreted. Risk communication is complex, as a recent National Research Council publication attests (National Research Council, 1989). Even when the effects of a risk factor can be reasonably estimated, the nature of the risk may be communicated in different ways. Should we be told that a risk is 5 percent for a year or (equivalently) 0.4 percent for a month? Is it more meaningful to know the risk of death or the expected reduction in years of life expectancy? Does it help to have a standard of comparison, for example, to know that the risk of HIV to surgeons performing 25 operations on infected patients is about the same as that of death on a Louisiana oil rig over the course of a year (Orient, 1990)? Once communicated, risk statistics are subject to cognitive reasoning processes that may introduce other sources of error or distortion. For example, whether a risk is framed in terms of the probability of positive or negative outcomes will affect an individual's response (Tversky and Kahneman, 1981). Another problem is that the small probabilities associated with many medical risks may be difficult for people to interpret. Perhaps that is why having personal knowledge about someone who contracts a disease has such a potent effect. For example, the national incidence of tests for breast and colon cancer increased dramatically after Nancy and Ronald Reagan incurred these diseases.

Becker and Janz (1987) have provided a theoretical analysis of health risk appraisal that points to some other relevant issues. One is whether personalized risk estimates are more effective than population estimates, and if so, why. Another is how to motivate risk-reducing behavior, for example, by promoting belief in the alterability of risk factors or by increasing the awareness of risky behavior. The introduction of warnings on cigarette packages is an example of public policy intended to increase risk awareness.

The design of group programs to increase positive health practices is another area in which human factors research can make important and widespread contributions. Programs aimed at reducing smoking, decreasing fat intake and increasing fiber, reducing recreational sun exposure, and inducing

compliance with screening programs for such diseases as breast and colon cancer are among those that promise to have very beneficial results (Fink, 1987). Research to increase the effectiveness of such programs would be of considerable societal value.

The computer can not only store data but also deliver data directly to medical practitioners and patients. The general area concerned with information encoding, representation, and communication in medicine is called medical informatics. Chapter 7 discusses medical informatics in the broader context of information systems. In this section we consider a number of medical contexts in which information delivery is critical and the implications of these contexts for human factors research needs.

Several databases are directly available to medical workers and, in some cases, patients as well. Examples are MEDLINE, which provides titles and abstracts of recently published articles in the medical literature and was developed by the National Library of Medicine, and CANCERLIT, a similar bibliographic system with a narrower range of citations. MEDLINE is available on CD-ROM from several companies and can be accessed by modem.

The National Cancer Institute has developed Physician Data Query (PDQ), a computer database for information about advances in cancer treatment and clinical trials (see Hubbard et al., 1987). Protocols of currently active clinical trials are entered into PDQ and are indexed by disease and eligibility criteria. In addition, PDQ provides state-of-the-art information about disease characteristics and prognosis. There has been considerable effort to make PDQ available to the medical community. The system can be accessed at thousands of medical libraries and centers, and several commercial vendors have been licensed to distribute PDQ via networks and CD-ROM.

Information delivery can be extremely helpful in medical decision making. Computer programs variously known as knowledge-based systems, decision aids, and expert systems are meant to facilitate the many medical decisions that must be made by both medical workers and patients. Whether these programs succeed depends in part on how well they are tailored to the humans who use them.

Automated decision systems emerged at least three decades prior to the 1990s (Ledley and Lusted, 1961). These systems generally have two components—a knowledge base of facts, often in the form of rules, and procedures for using the facts, sometimes called the inference engine. When applied to a specific problem, the system also requires data about its characteristics. Expert systems are so called because the underlying knowledge is derived from experts. The term decision aid suggests a program that provides some of the information that is useful in decision making, often quantified or structured data, but that leaves other elements to the user. In practice, an expert system generally functions as a decision aid in that it contributes to decision making rather than governing it.

One of the best-known expert systems is MYCIN (Shortliffe, 1976), which is designed to identify the source of a bacterial infection and to recommend treatment. MYCIN has been augmented by instructional programs that train potential users. Recently the National Library of Medicine has developed AI/RHEUM, a diagnostic system for rheumatologic diseases (Kingsland et al., 1986). This program features an extensive help system that incorporates video images as well as text.

The range of application of knowledge-based systems is now very wide, including, for example, diagnosis, prescription, causal interpretation of data, medical consulting, and selection of adjuvant therapies. When one considers that each of these areas can be appled to many specific diseases or problem areas, it is clear that the number of potential programs is immense. Commercially available shells, which specify a format for the knowledge base and provide an inference engine, are intended to simplify the development of new systems.

Computerized decision aids are attractive in part because of evidence that human decision making is far from ideal. People have been characterized as subject to heuristics and biases that introduce error (for review in a medical context, see Dowie and Elstein, 1988). People's decisions differ depending on the way problems are presented (framing effects), how much has been expended on a solution in the past (sunk costs), what sample solutions are provided (anchoring and adjustment), and the extent to which they can retrieve past solutions from memory (availability). Medical decisions are complex; the computer can facilitate decision making by providing as much relevant data as possible, along with a content-free objective decision process.

Past human factors research in several areas has intersected with needs in medical informatics. Relevant research topics include modeling of decision making and design of usable interfaces for expert systems. We suggest

here, then, not so much new substantive topics as specific applications that present new research needs.

A traditional goal of human factors research has been to improve the accessibility of tools. This goal is critical to the successful application of systems for information delivery, as Chapter 7 discusses at length. A need for increased accessibility of medical information systems is clear in that the explosion of information available to medical practitioners has not necessarily resulted in widespread use. The National Cancer Institute recently conducted a formal evaluation of the PDQ system (Czaja et al., 1989). Despite intensive publicity about the system and efforts to make it maximally available to health care workers, fewer than 50 percent of physicians in cancer specialties were aware of it, and only about half these physicians reported using it. Use by community physicians who were not in oncological specialties was termed "very low." The greatest use was by employees of the Cancer Information Service, who provide information to the public.

The low level of PDQ use was attributed in part to physicians' discomfort with computers. The primary use of computers in a clinical practice tends to be for administration and word processing rather than for access to information or for automated decision facilitation. There are also problems with negative attitudes about computers. For example, only 5 percent of non-oncologists in the PDQ review indicated that computerized data retrieval was very important.

It seems clear that there is a yawning gap between the availability of technological tools—which provide data about disease risk, diagnosis, prognosis, and treatment as well as automated aids for decision making—and the use of these tools. Human factors research can play an important role in bridging this gap. Some of the problems that must be considered are the following:

• The user interface. It is by now a given that the success of a computer program designed for general use depends on its "user friendliness." Because the designers of PDQ were aware that medical practitioners might be unfamiliar with computers, PDQ software offers menus as well as direct commands; some instructions are also given on-line. The evaluation suggests, however, that these measures are insufficient to promote widespread use.

• Medical training. Medical schools do not currently include extensive training in medical informatics in their curricula, although efforts to increase such training are under way (Ball and Douglas, 1990). The National Library of Medicine has created an elective medical informatics course for advanced medical students, which is intended to create a "seed crop" of researchers in this area (National Institutes of Health, 1991). Given the demands of the medical curriculum, any such training must be highly efficient;

it is unlikely to receive a substantial portion of overall education time. Postgraduate training in the use of specific programs—conducted, for example, in connection with professional meetings—may be a way to increase computer fluency.

• Medical schools also do not currently provide extensive training in epidemiology and statistics (Klatzky et al., 1994); again, it is unclear that such training can be justified in the general curriculum. Yet the evaluation of retrieved information is critical, and this may depend on an understanding of the statistical measures and models that were used to derive it. Klatzky et al. suggested that one approach to this problem is to provide on-line introductions to basic statistical concepts as part of a database retrieval program or decision aid.

• Attitudes about technological aids. Negative attitudes about the usefulness of computerized data retrieval were cited as a potential barrier to the use of PDQ by practicing physicians. More generally, discomfort with decision aids such as computer-guided diagnoses or treatment plans may preclude their use. Efforts to bring technological aids to the practicing clinician should include attitude change as a goal. It should be made clear that computer programs can greatly assist the physician in acquiring relevant data and putting them to use.

It should not be forgotten that patients, too, play a role in decision making. There has been relatively little effort to provide data to patients or to assist them in making decisions about treatment. Research is needed to ascertain the extent to which patients wish to play a decision-making role as well as to determine what resources might aid them. A straightforward approach is to construct patient-oriented databases of general information about diseases and appropriate references. The question then becomes how to maximize the accessibility of the information.

Patients often make decisions under considerable stress. Research needs related to the effects of stress on cognition are reviewed in Chapter 10, but it is important to note here that there has been little research on the stress induced by diagnosis of disease. In some diseases, for example, breast cancer, a range of treatment options are available. Typically, patients make decisions about treatment shortly after diagnosis, when stress is likely to be greatest. Meyerowitz (1980) suggests that the psychological impact of breast cancer diagnosis and treatment may be as threatening as the disease itself. Human factors researchers could address such questions as the effects of stress on decision making in this context and the potential role of automated data retrieval and decision aids. The prevalence of breast cancer suggests that it could provide a useful paradigm for research on stress among medical patients.

The continued presence of the familiar blood-pressure cuff notwithstanding, the computer has virtually revolutionized medical instrumentation. Bioinstrumentation is a term that has been applied to the combination of electronic and biological technology (Wise, 1990). Bioinstruments include, for example, biomedical imaging systems that take advantage of the computer for gathering, enhancing, and displaying data; electrochemical sensors that incorporate whole living cells in their construction; artificial sensing devices such as tactile sensors based on force-sensitive polymers; and transducers such as cochlear implants.

As is the case with any technological change, advances in medical instrumentation lead to problems in mastering the new technology and create needs for human factors research. The pace of advance means that researchers must "enter the loop" quickly. This has frequently not been the case. Here we briefly describe examples of new developments in instrumentation and consider related human factors needs.

Medical imaging is a century old; the X-ray was discovered in 1895. The fundamental requirements for an imaging system remain the same. What is needed is some means of acquiring information about the imaged site and some means of displaying that information. However, technological advances have greatly expanded the ways to meet both of these requirements and have added other capabilities, particularly new means of analyzing, enhancing, and interpreting the data.

A variety of techniques for acquiring information about a targeted site are in widespread use today. Two-dimensional images like that provided by the X-ray are the basis for computed tomography, in which a contiguous sequence of cross-sectional images is used to synthesize a three-dimensional representation. Magnetic resonance imaging relies on the activity of atoms when placed in a magnetic field to construct a two-dimensional image. Positron emission tomography uses the distribution of radioactivity to produce two-dimensional slices that can then be used to construct a three-dimensional model. Still other methods are thermography, which measures the distribution of temperature, and ultrasound.

The resulting data are often subjected to a number of pre-processing algorithms that, for example, detect edges, compute regions, reduce noise, and extract and enhance critical features. The data are then displayed in various ways. Multiple slices from different planes and at different depths are often displayed simultaneously. A three-dimensional surface description can be displayed as a high-quality two-dimensional projection onto a

viewing screen. There is also the possibility of fully three-dimensional display techniques such as holograms or a display through stereoscopic viewers. Color can be used to differentiate regions, especially when there is no direct counterpart between acquired data and a visual feature (e.g., oxygen consumption or temperature).

New techniques are increasingly expanding the ways medical images are displayed as well as how they are intepreted with the aid of knowledge-based systems. The potential usefulness of automated interpretation is substantial, given the density of the imaged information and the complexity of the mapping between diagnostically relevant attributes of the body and visually apparent aspects of the image. Swets et al. (1991) have described a two-part interactive decision aid that first prompts the user to rate a given image according to a series of scales and then uses the obtained scale values to estimate the probability of malignancy, which is then communicated to the user. An analysis of breast-cancer diagnoses with and without the decision aid revealed a clear advantage for using the aid, an advantage that increased with the difficulty of the diagnosis.

The use of film-based radiological images is expected to decrease, owing to the advent of digital environments (Arenson et al., 1990). A rapidly developing tool is the picture archiving and communication system (PACS), a digital workstation that will allow physicians to call up and display stored images. The 1990 meeting on medical imaging of the International Society for Optical Engineering had 15 paper sessions and a poster session devoted to PACS (Dwyer and Jost, 1990). Among its many positive aspects, PACS offers the possibility for on-line image enhancement and interpretation aids; however, a negative aspect is the potential degradation in image quality due to digital display.

Techniques derived from perception and decision science have been widely used to evaluate the relative merits of imaging methods, as well as to assess the effectiveness of enhancement techniques and the degradation under systems like PACS. In particular, an analysis using the area under the relative operating characteristic (ROC; in the context of signal detection theory, this is also called the receiver operating characteristic) has proven highly effective. The ROC plots the proportion of true positive responses as a function of the proportion of false positives in a diagnostic test. The ROC measure is independent of the decision criterion and the frequency of diagnosed events (Metz, 1989; Swets, 1988). For example, Swets et al. (1979), using this measure, found computed tomography to be more accurate than radionuclide scanning in diagnosing brain lesions.

Much attention has already been given to human factors considerations in the design of medical imaging workstations (O'Malley and Ricca, 1990). A key to physician acceptance of these systems seems to be perceived savings in time and effort. In one survey study of factors influencing the

acceptance of PACS systems by radiologists (Saarinen et al., 1989), speed of information delivery was found to be the primary concern. Thus, a particularly positive aspect was the time that might be saved in traveling to and from the site where radiological files were stored. Physicians also favored a system that would eliminate the need for multiple trips to retrieve films that had been misplaced or temporarily checked out by others. On the negative side, physicians were concerned with the possibility of significant downtime for the system. Somewhat surprisingly, interest in PACS was not affected by the age of the physician and was only slightly affected by prior experience with computers.

A biosensor is a device that places a transducer in close contact with molecules of a biological substance in order to produce an output that is correlated with concentration of the substance. The general sequence by which the sensor operates is that the measured substance interacts with a "biological mediator" (e.g., one or more enzymes, antibodies, or bacteria), leading to an output (e.g., electrochemical, optical, or calorimetric) that can be transformed into an electrical signal (Mascini et al., 1990). Reviews of biosensor technology can be found in Claremont (1987), Higgins and Lowe (1987), and Wilkins (1989). Schultz (1991) traces the modern biosensor to two antecedents: information technology, particularly the development of miniaturized components, and molecular biology, which identifies biomolecules that can serve as the mediators for recognizing the target substance.

The field of biosensor technology is now burgeoning at both basic and applied levels. Attractive features of this technique include its speed, the need for low-volume samples, and the possibility of in situ measurement and long-term implantation. The potential variety of biosensors is very large; sensors have been developed to measure such substances as urea, cholesterol, proteins, alcohol, and penicillin. These devices have widespread application beyond health care; for example, they are used in the food industry to measure contaminants.

One of the first areas of health care application of biosensors, and probably the most developed, is sensing of glucose concentrations (e.g., Ross et al., 1990; Schaffar and Wolfbeis, 1990). This has obvious important benefits to diabetics. For decades, diabetics have done home glucose monitoring by using visual assessment of color changes on reagent strips dipped in urine or blood. These methods have a high potential for error, and, particularly with urine testing, there is considerable temporal lag between a change in blood glucose and the ability to detect it. Subjective visual monitoring and urine tests for diabetes have given way to blood glucose monitors, pocket-sized electronic devices that assess the level of

glucose from a blood sample by sensing results of its interaction with an enzyme. By performing periodic measurements and adjusting insulin injections or diet accordingly, diabetic patients can maintain acceptable glucose levels with minimal fluctuation.

Biosensors not only make possible more direct and accurate evaluation but also provide the possibility of long-term implantation. An implanted glucose sensor is critical to the development of an artificial pancreas that would monitor glucose levels and adjust insulin injections automatically; most of the components for such a system now exist. Expert-system technology can also be combined with glucose sensing to regulate glucose levels (Lougheed et al., 1987).

As was mentioned above, a clear trend in health care is to discharge patients after shorter hospitalizations, necessitating increased home care. Often, this is made possible by sending the patient home with a medical device. For example, ventilators may be placed in the home for patients with respiratory problems (Bach et al., 1992; Thompson and Richmond, 1990). Patients may receive medication at home with infusion pumps, electronically controlled devices used to regulate the flow of drugs along lines implanted in the body, for example subcutaneous catheters or venous access ports. These devices are used to supply antibiotics, drugs to reduce pain, and chemotherapy (New et al., 1991; Reville and Almadrones, 1989; Storey et al., 1990). Some patients receive ''total parenteral nutrition," their entire daily calorie requirement, intravenously at home (Bisset et al., 1992; Viall, 1990). Home dialysis machines now enable kidney patients to undergo dialysis even while they sleep (Delano and Friedman, 1990; Health Devices, 1991). Although these devices have considerable benefits, not only economically but also in terms of patient welfare and satisfaction, they also put responsibility on home caregivers and patients alike.

The contributions that human factors researchers can make to the development and use of medical imaging seem enormous. New techniques for acquiring, displaying, and evaluating images call for increased interaction between the human user and the computer. User options for image enhancement and display adjustment increase the complexity of these systems. The same points apply as those that emerged in the evaluation of the PDQ system. The effectiveness of these systems will substantially depend on the extent to which the interface renders technology accessible, given the attitudes and training of physicians and technical staff.

Human factors researchers can also play a role in developing and evaluating systems for image enhancement and knowledge-based interpretation. Each new modality creates anew such questions as, which imaged features are most useful for diagnosis? Currently, the discovery of relevant features may be a piecemeal process taking several years; systematization of this process is badly needed (Swets, personal communication).

Advances in medical instruments to be used in hospitals and homes also create a substantial need for human factors research. Of course, any new instrumentation calls for training of operators and appropriate design of machine interfaces. The increased use of medical devices at home carries with it additional problems.

One is the adequacy of training. Written instructions and warnings should obviously be very clear, and hands-on training is likely to be necessary for all but the simplest devices. Lack of education and language skills by home users may hamper training efforts, as may advanced age. Another consideration is that stress from an incident such as machine malfunction could undermine the effects of training. Although the effects of long-term stress imposed on home caregivers and patients have received some attention (e.g., Smith et al., 1991; Wegener and Aday, 1989), the cognitive effects of acute stress merit more study. Clearly, patients and caregivers should have access to support personnel outside the home. Often, pharmacists are the ones responsible for advising patients about a device when they themselves have inadequate training (Kwan and Anderson, 1991).

Training is but one concern. Another is that the home environment may not be well suited for the device in, for example, the type and reliability of its power supply, the proximity of water and electrical connections, and the space available for the equipment and appendages like oxygen tanks. Ambient temperature may affect device function, and the home may not be adequately equipped to control it. Home use often demands that patients follow a regular schedule, indicating a need to provide records of past administration and to signal the next application. As was mentioned previously, the elderly may in particular need reminder systems to trigger self-care.

Glucose measurement devices have been closely evaluated from a human factors perspective (e.g., Kelly et al., 1990; McDonald, 1984; Moss and Delawter, 1986) and hence provide a test bed with which to evaluate home care more generally. Kelly et al. found a number of problems with even such a relatively simple device. In evaluating the instructional materials provided with blood glucose monitors, most diabetes educators judged them inadequate and believed that additional instruction was necessary. User error was identified in nearly three-fourths of reported problems. Patients often failed to clean the device or to calibrate it properly, possibly in order to reduce the time spent in operating it, and errors occurred at virtually

every stage of operation. In efforts to cut costs, some patients split test strips so they could be used twice, with potentially adverse effects on sensing accuracy.

The case of the blood glucose monitor serves to emphasize the need for human factors research related to home use of medical devices. Whereas blood glucose monitors are simple devices and have been designed for patient use, other devices that are sent to the home may be far more complex and may have been designed for hospital use only (e.g., for ventilators, see Health Devices, 1988). These devices are even more likely to have inadequate written documentation for home caregivers and it is likely that any in-hospital training prior to discharge will be devised on an ad hoc basis.

In this section we discuss some of the biomechanical problems encountered in health care, the magnitude of these problems, and the research topics they raise. Traditional concerns of human factors research, including design of the workplace, design and implementation of devices, and performance modeling, have substantial potential for application in medical contexts.

In any work system, including the health care delivery system, stresses are imposed on the body. These stresses can be (a) mechanical, involving the musculoskeletal system; (b) physiological, involving the cardiopulmonary system; or (c) psychological. Physically demanding tasks often produce mechanical stress for the musculoskeletal system, particularly the spine, resulting in low back injury. In health care delivery systems, handling of materials—particularly the handling of patients—results in high spinal stresses.

Back injuries account for approximately one of every five injuries and illnesses in the workplace (Bureau of Labor Statistics, 1982). Nursing aides and licensed practical nurses ranked fifth and ninth, respectively, in compensation claims for back injuries (Klein et al., 1984). Available statistics indicate that nursing personnel are as likely to suffer from a compensable back injury as are workers in occupations with heavy load-handling tasks (Jensen, 1987). Back injuries are the result of large stresses imposed on the spine. High levels of biomechanical stress imposed on the musculoskeletal system, particularly the spine, have been reported in tasks performed by nursing personnel (Gagnon et al., 1986; Stubbs et al., 1983; Torma-Krajewski, 1986). According to Lloyd et al. (1987), efficient and safe patient transfer practices should be based on sound biomechanical considerations.

Nursing personnel—and nursing aides in particular—often lift, move,

and transfer patients whose weights range from 37 kg to over 100 kg. These weights are higher than the capacity of most females (National Institute for Occupational Safety and Health, 1981). Gagnon et al. (1986) reported that the compressive forces on the L5/S1 disc ranged from 5.74 to 7.95 kilonewtons (kN) for a single person handling a 72 kg mannequin. These values and the estimated compressive forces for handling a patient using two persons are approximately 4.44 kN; the values are significantly higher than the recommended National Institute for Occupational Safety and Health safe compression limits of 3.4 kN.

The management of chronic pain and disability after an injury has always been considered the exclusive domain of medical professionals. Ergonomic involvement in the pre-injury (prevention) and post-injury (rehabilitation and return-to-work) stages has been shown to contribute significantly to the successful control and management of overall prevention and rehabilitation and the avoidance of disability recurrence (Khalil et al., 1985, 1988; Rosomoff, 1987; Rosomoff et al., 1981). Important issues in the rehabilitation of individuals with certain disabilities are measurement of their functional performance, accurate and objective examination and assessment of their capabilities and limitations, and quantitative descriptions of them in terms of human performance profiles (HPP) (Abdel-Moty, 1991).

In the evaluation of HPP, specific areas of interest are the following: (a) physical characteristics of the patient such as strength, flexibility, endurance, and posture; (b) functional capacity of the patient in performing certain activities such as lifting and walking; and (c) work-related capacities such as the ability to perform specific job tasks under prescribed conditions. In evaluating patients with chronic pain or certain physical disabilities, it is very important to use objective measurement of specific abilities (Khalil et al., 1990). The HPP can then be compared with that of healthy persons of equivalent age, sex, and work category in order to determine functional capacity.

To enhance biomechanical research in health care delivery systems—particularly for back disorders—biomechanical modeling research can be invaluable. As representations of the real system, models can be useful in examining the behavior of the system under consideration (in this case, the human body) without exposing the body to a variety of hazardous conditions. Ergonomic models are discussed in another publication of the Committee on Human Factors (Kroemer et al., 1988).

In order to reduce and control back injuries in health care delivery systems, a major research effort is needed to evaluate handling of patients to reduce stress on the spine. The research must focus on several areas: (a) the training of personnel in methods of handling patients to reduce spinal stresses; (b) the evaluation of facility designs—which include beds, wheelchairs, and other devices to facilitate patient handling—for all systems, whether they use manual methods, assistive devices, or both; and (c) the evaluation of existing design and the ultimate redesign of assistive devices used in patient handling to reduce spinal stresses. The research should also consider several other variables related to spinal stress. These are (a) the time it takes to perform the handling activity, (b) patient comfort and safety, and (c) patient characteristics (Garg et al., 1991). Furthermore, it is important that the research pay careful attention to the differences between laboratory environments and the actual work site; laboratory results must be carefully evaluated (Garg et al., 1991) prior to field application.

The evaluation of the functional capacity of patients in need of physical rehabilitation is another area in which human factors research can play a role. Currently several techniques are being used to assess functional capacity. Some of these are quite subjective and rely heavily on data obtained through observation; others are more objective and gather data through quantitative measurement of such factors as mobility and strength. Even when the more objective measures are employed, there is a need to integrate and translate these measurements into functional capacity values. Therefore, research in the development and use of objective measures for the evaluation of functional capacity is sorely needed. This will make it possible to determine more accurately the level of individual abilities, which in turn can help determine the optimal program of rehabilitation (Abdel-Moty and Khalil, 1988; Abdel-Moty et al., 1989, 1990).

Research in biomechanical modeling is also needed. The current biomechanical models are quite basic at best and cannot adequately deal with the complicated structures of the musculoskeletal system. Three-dimensional biomechanical models that include, for instance, the effects of soft tissue and the individual muscle tension generated in performing typical patient handling could provide important insights into handling methods and design of work areas and equipment to minimize the stresses on the spine. In addition, good biomechanical models can be helpful in developing manual handling methods that minimize stresses on the body, particularly the spine.

The primary aim of this chapter has been to call attention to the substantial contributions that could be made by human factors researchers in the area of medical care. This will be at least an initial step toward redressing the underutilization of human factors researchers in medical contexts. The topics that we have described include some traditional concerns but also point to a broad spectrum of relatively novel research needs that arise as a result of our changing societal and technological environment. Many of these problems will require that human factors personnel work with experts in other disciplines. We have identified a large number of research opportunities in the hope that their diversity will attract a larger number of researchers to the area. We suspect that many of these problems, unfortunately, will not be dealt with in the next decade.

Particularly important research goals noted in this chapter are the following:

• to identify and eliminate sources of error arising from the medical workplace and medical devices;

• to design health-promoting environments for the aged;

• to identify health risks in the workplace;

• to determine techniques for effective risk communication;

• to develop health risk reduction and illness prevention programs for groups;

• to design user-appropriate interfaces for new medical devices, such as decision aids, imaging systems, and biosensors;

• to identify ways to facilitate access to health information by medical personnel;

• to identify barriers to effective use of medical devices in the home and to redesign devices and home environments so as to promote effective home care;

• to identify and eliminate sources of biomechanical stress on the musculoskeletal system from health care practices;

• to develop techniques for measuring functional capacity of candidates for rehabilitation; and

• to develop biomechanical models to assist in determining potential health hazards.

As we indicated at the outset, the research needs that are reviewed here are far from exhaustive, and new technological developments are likely to expand the list. Additions will only reinforce the important role that human factors can play in fitting health care practices to those in need of care.

We would like to acknowledge the help of John Swets and Sue Bogner with this chapter.

Abdel-Moty, E. 1991 Ergonomics issues in low back pain: intervention strategies. Proceedings of the Human Factors Society 35th Annual Meeting. Santa Monica, Calif.: Human Factors Society.

Abdel-Moty, E., and T.M. Khalil 1988 Ergonomic considerations for the reduction of physical task demands of low back pain patients. Pp. 959-967 in F. Aghazadeh, ed., Trends in Ergonomics/Human Factors, Vol. IV. Amsterdam, Netherlands: North-Holland, Elsevier Science Publishing.

Abdel-Moty, E., T.M. Khalil, S.S. Asfour, M. Howard, R.S. Rosomoff, and H.L. Rosomoff 1989 Effects of pain on psychomotor abilities. Pp. 465-471 in A. Mital, ed., Advances in Industrial Ergonomics and Safety. New York: Taylor and Francis.

Abdel-Moty, E., T. Khalil, S. Asfour, M. Goldberg, R. Rosomoff, and H. Rosomoff 1990 On the relationship between age and responsiveness to rehabilitation. Pp. 49-56 in B. Das, ed., Advances in Industrial Ergonomics and Safety , Vol II. New York: Taylor and Francis.

Arenson, R.L., D.P. Chakraborty, S.B. Seshadri, and H.L. Kundel 1990 The digital imaging workstation. Radiology 176:303-315.

Bach, J.R., P. Intintola, A.S. Alba, and I.E. Holland 1992 The ventilator-assisted individual: cost analysis of institutionalization vs. rehabilitation and in-home management. Chest 101:26-30.

Ball, M.J., and J.V. Douglas 1990 Informatics programs in the United States and abroad. MD Computing 7:172-175.

Becker, M.H., and N.K. Janz 1987 Behavioral science perspectives on health hazard/health risk appraisal. Health Services Research 22(4):537-551.

Bisset, W.M., P. Stapleford, S. Long, A. Chamberlain, B. Sokel, and P.J. Milla 1992 Home parenteral nutrition in chronic intestinal failure. Archives of Disease in Childhood 67:109-114.

Bogner, S. 1991 Human factors and medicine. P. 682 in Proceedings of the Human Factors Society 35th Annual Meeting. Santa Monica, Calif.: Human Factors Society.

Breslow, L., J. Fielding, A.A. Herrman, and C.S. Wilbur 1990 Worksite health promotion: its evolution and the Johnson & Johnson experience. Preventive Medicine 19:13-21.

Bureau of Labor Statistics 1982 Back Injuries Associated With Lifting. Work Injury Report, Bulletin #2144:1. Washington, D.C.: U.S. Department of Labor.

Carleton, R.A., L. Sennett, K.M. Gans, S. Levin, C. Lefebvre, and T.M. Lasater 1991 The Pawtucket Heart Health Program: influencing adolescent eating patterns. Annals of the New York Academy of Sciences 623:322-326.

Claremont, D.J. 1987 Biosensors: clinical requirements and scientific promise. Journal of Medical Engineering and Technology 11:51-56.

Cook, R.I. 1991 How to do that voodoo that you do so well: medical human factors in the explicit context of use . P. 684 in Proceedings of the Human Factors Society 35th Annual Meeting. Santa Monica, Calif.: Human Factors Society.

Czaja, S.J., ed. 1990 Human Factors Research Needs for an Aging Population. Panel on Human Factors Research Issues for an Aging Population, Committee on Human Factors, National Research Council. Washington, D.C.: National Academy Press.

Czaja, R., C. Manfredi, D. Shaw, and G. Nyden 1989 Evaluation of the PDQ System: Overall Executive Summary. Report to the National Cancer Institute under Contract No. N01-CN-55459. Survey Research Laboratory, University of Illinois, May.

Delano, B.G., and E.A. Friedman 1990 Correlates of decade-long technique survival on home hemodialysis. ASAIO Transactions 36:337-339.

Dowie, J., and A. Elstein 1988 Professional Judgment: A Reader in Clinical Decision Making. Cambridge, England: Cambridge University Press.

Dwyer, S.J., III, and R.G. Jost, eds. 1990 Medical Imaging IV: PACS System Design and Evaluation. Proceedings of SPIE, the International Society for Optical Engineering in cooperation with the American Association of Physicists in Medicine, Vol. 1234, parts 1 and 2. Bellingham, Wash.: SPIE.

Fink, D.J. 1987 Preventive strategies for cancer in women. Cancer 60:1934-1941.

Gagnon, M., C. Sicard, and J.P. Sirois 1986 Evaluation of forces on the lumbo-sacral joint and assessment of work and energy transfers in nursing aides lifting patients. Ergonomics 29:407-421.

Gail, M.H., L.A. Brinton, D.P. Byar, D.K. Corle, S.B. Green, C. Schairer, and J.J. Mulvihill 1989 Projecting individualized probabilities of developing breast cancer for white females who are being examined annually. Journal of the National Cancer Institute 81:1879-1886.

Garg, A., B. Owen, D. Beller, and J. Banaag 1991 A biomechanical and ergonomic evaluation of patient transferring tasks: bed to wheelchair and wheelchair to bed. Ergonomics 34:289-312.

Hadler, S.C. 1990 Hepatitis B virus infection and health care workers. Vaccine 8 (March Supplement): S24-S28; discussion:S41-S43.

Hayward, R.A., and M.F. Shapiro 1991 A national study of AIDS and residency training: experiences, concerns, and consequences. Annals of Internal Medicine 114:23-32.

Health Devices 1988 Portable volume ventilators. Health Devices 17(4):107-131. 1991 Hemodialysis machines. Health Devices 20(6):187-232.

Health Services Research 1987 October issue 22(4).

Helmreich, R.L., and H.G. Schaefer 1994 Team performance in the operating room. Pp. 225-254 in M.S. Bogner, ed., Human Error in Medicine. Hillsdale, N.J.: Erlbaum.

Higgins, I.J., and C.R. Lowe 1987 Introduction to the principles and applications of biosensors. Philosophical Transactions of the Royal Society of London 316:3-11.

Hubbard, S., J.E. Henney, and V.T. DeVita, Jr. 1987 A computer data base for information on cancer treatment. New England Journal of Medicine 316:315-318.

Jensen, R. 1987 Disabling back injuries among nursing personnel: research needs and justifications. Research in Nursing and Health 10:29-38.

Kelly, R.T., J.R. Callan, T.A. Kozlowski, and E. Menngola 1990 Human Factors in Self-Monitoring of Blood Glucose. Task 4 Final Report. FDA/CDRH-90/60. Springfield, Va.: NTIS.

Khalil, T.M., S.S. Asfour, E. Abdel-Moty, R.S. Rosomoff, and H.L. Rosomoff 1985 New horizons for ergonomics research in low back pain. Pp. 591-598 in R.E. Eberts and C.G. Eberts, eds., Trends in Ergonomics/Human Factors. Amsterdam, Netherlands: North-Holland, Elsevier Science Publishing. 1988 Quantitative assessment of outcome of a low back pain rehabilitation program. Abstracts of the International Conference on the Study of the Lumbar Spine. Miami, Fla. April 13-15.

Khalil, T.M., E. Abdel-Moty, and T.M. Asfour 1990 Ergonomics in the management of occupational injuries. Pp. 41-53 in B.M. Pulat and D.C. Alexander, eds., Industrial Ergonomics: Case Studies. Norcross, Ga.: Industrial Engineering and Management Press.

Kingsland, L.C., III, D.A.B. Lindberg, and G.C. Sharp 1986 Anatomy of a knowledge-based system. MD Computing 3:18-26.

Klatzky, R.L., J. Geiwitz, and S.C. Fischer 1994 Using statistics in clinical practice: a gap between training and application. Pp. 123-140 in S. Bogner, ed., Human Error in Medicine . Hillsdale, N.J.: Erlbaum.

Klein, B.P., R.C. Jensen, and L.M. Sanderson 1984 Assessment of workers' compensation claims for back strains/sprains. Journal of Occupational Medicine 26:443-448.

Kroemer, K.H.E., S.H. Snook, S.K. Meadows, and S. Deutsch, eds. 1988 Ergonomic Models of Anthropometry, Human Biomechanics and Operator-Equipment Interfaces. Committee on Human Factors, National Research Council. Washington, D.C.: National Academy Press.

Kwan, J.W., and R.W. Anderson 1991 Pharmacists' knowledge of infusion devices. American Journal of Hospital Pharmacy 48:10 Suppl 1, S52-S53.

Lasater, T.M., L.L. Sennett, R.C. Lefebvre, K.L. DeHart, G. Peterson, and R.A. Carleton 1991 Community-based approach to weight loss: the Pawtucket "weigh-in." Addictive Behaviors 16:175-181.

Ledley, R.S., and L.B. Lusted 1961 Medical diagnosis and modern decision making. Pp. 117-157 in Proceedings of Symposia in Applied Mathematics, Vol. 14. Providence, R.I.: American Mathematical Society.

Levi, L. 1990 Occupational stress: spice of life or kiss of death? American Psychologist 45:1142-1145.

Lloyd, P., C. Tarling, J.D.G. Troup, and B. Wright 1987 The Handling of Patients: A Guide for Nurses, 2nd ed. London, England: The Royal College of Nursing.

Lougheed, W.D., A. Schiffrin, and A.M. Albisser 1987 Stabilizing blood glucose with a novel medical expert system. Biosensors 3:381-389.

Mascini, M., D. Moscone, and G. Palleschi 1990 Biosensor applications of continuous monitoring in clinical chemistry. Pp. 1429-1460 in D.L. Wise, ed., Bioinstrumentation: Research, Developments and Applications. Stoneham, Mass.: Butterworth.

McDonald, W.I. 1984 Quality control of home monitoring of blood glucose concentrations. British Medical Journal 288:1915.

Metz, C.E. 1989 Some practical issues of experimental design and data analysis in radiological ROC studies. Investigative Radiology 24:234-245.

Meyerowitz, B.E. 1980 Psychosocial correlates of breast cancer and its treatments. Psychological Bulletin 87:108-131.

Morris, R.E., and E. Turgut 1990 Human immunodeficiency virus: quantifying the risk of transmission of HIV to dental health care workers. Community Dentistry and Oral Epidemiology 18:294-298.

Moss, J.P., and D.E. Delawter 1986 Self-monitoring of blood glucose. American Family Physician 33:225-228.

National Institute for Occupational Safety and Health 1981 Work Practices Guide for Manual Lifting. NIOSH Technical Report No. 81-122. Washington, D.C.: National Institute for Occupational Safety and Health.

National Institutes of Health 1991 Clinical Electives Program for Medical and Dental Students 1992-1993 . NIH Publication No. 91-499. Bethesda, Md.: National Institutes of Health.

National Research Council 1989 Improving Risk Communication. Committee on Risk Perception and Communication. Washington, D.C.: National Academy Press.

New, P.B., G.F. Swanson, R.G. Bulich, and G.C. Taplin 1991 Ambulatory antibiotic infusion devices: extending the spectrum of outpatient therapies. American Journal of Medicine 91:455-461.

Newell, G.R., and V.G. Vogel 1988 Personal risk factors: what do they mean? Cancer 62:1695-1701.

O'Malley, K.G., and K.G. Ricca 1990 Optimization of a PACS display workstation for diagnostic reading. Pp. 940-946 in Medical Imaging IV: PACS System Design and Evaluation . Proceedings of SPIE, the International Society for Optical Engineering in cooperation with the American Association of Physicists in Medicine, Vol. 1234. Bellingham, Wash.: SPIE.

Orient, J.M. 1990 Assessing the risk of occupational acquisition of the human immunodeficiency virus: implications for hospital policy. Southern Medical Journal 83(10):1121-1127.

Paryani, S.B., T.A. Marsland, P. Faucher, E. Fontanelli, M. Freeman, H. Johnston, W. Morrow, P. Prabhu, M. Stearman, and F. Vines 1990 Breast cancer screening project in northeast Florida. Journal of the Florida Medical Association 77:29-31.

Pickett, R.M., and T.J. Triggs, eds. 1974 Human Factors in Health Care. Lexington, Mass.: D.C. Heath.

Rappaport, M. 1970 Human factors applications in medicine. Human Factors 12:25-35.

Regnier, S.J. 1993 Symposium underscores value of OR teamwork. American College of Surgeons Bulletin 78:73-81.

Reville, B., and L. Almadrones 1989 Continuous infusion chemotherapy in the ambulatory setting: the nurse's role in patient selection and education. Oncology Nursing Forum 16:529-535.

Rosomoff, H.L. 1987 Comprehensive pain center approach to the treatment of low back pain. Pp. 78-85 in Low Back Pain: Report of a Workshop. Rehabilitation Research and Training Center, Department of Orthopaedics and Rehabilitation. Charlottesville, Va.: University of Virginia.

Rosomoff, H.L., C. Green, M. Silbert, and R. Steele 1981 Pain and low back rehabilitation program at the University of Miami School of Medicine. In K.Y. Lorenzo, ed., New Approaches to Treatment of Chronic Pain. NIDA Research Monograph 36. Washington, D.C.: U.S. Department of Health and Human Services.

Ross, D., L. Heinemann, and E.A. Chantelau 1990 Short-term evaluation of an electro-chemical system (ExacTech) for blood glucose monitoring. Diabetes Research and Clinical Practice 10:281-285.

Saarinen, A.O., G.L. Youngs, and J.W. Loop 1989 The Attitude of Referring Physicians Towards PACS: A Pre-Installation Assessment. Report to the MITRE Corp. DIN/PACS Evaluation Project Contract N55-200. Department of Radiology, University of Washington DIN/PACS Evaluation Project, November 30.

Schaffar, B.P., and O.S. Wolfbeis 1990 A fast responding fibre optic glucose biosensor based on an oxygen optrode. Biosensors and Bioelectronics 5:137-148.

Schucker, B., K. Bailey, J.T. Heimbach, M.E. Mattson, J.T. Wittes, C.M. Haines, D.J. Gordon, J.A. Cutler, V.S. Keating, and R.S. Goor 1987 Change in public perspective on cholesterol and heart disease: results from two national surveys. Journal of the American Medical Association 258:3517-3531.

Schultz, J.S. 1991 Biosensors. Scientific American 265(2): 64-69.

Shortliffe, E.H. 1976 Computer-Based Medical Consultations: MYCIN. New York: Elsevier.

Small, A.M. 1987 Design for older people. Pp. 495-504 in G. Salvendy, ed., Handbook of Human Factors. New York: Wiley.

Smith, C.E., C.K. Giefer, and L. Bieker 1991 Technological dependency: a preliminary model and pilot of home total parenteral nutrition. Journal of Community Health Nursing 8:245-254.

Smith, M.J. 1987 Occupational stress. Pp. 844-860 in G. Salvendy, ed., Handbook of Human Factors. New York: Wiley.

Storey, P., H.J. Hill, Jr., R.H. St. Louis, and E.E. Tarver 1990 Subcutaneous infusions for control of cancer symptoms. Journal of Pain and Symptom Management 5:33-41.

Stubbs, D.A., P.W. Buckle, M.P. Hudson, and P.M. Rivers 1983 Backpain in the nursing profession, II: the effectiveness of training. Ergonomics 26:767-779.

Swanson, G.M. 1988 Cancer prevention in the workplace and natural environment: a review of etiology, research design, and methods of risk reduction. Cancer 62:1725-1746.

Swets, J.A. 1988 Measuring the accuracy of diagnostic systems. Science 240:1285-1293.

Swets, J.A., R.M. Pickett, S.F. Whitehead, D.J. Getty, J.B. Schnur, J.B. Swets, and B.A. Freeman 1979 Assessment of diagnostic technologies. Science 205:753-759.

Swets, J.A., D.J. Getty, R.M. Pickett, C.J. D'Orsi, S.E. Seltzer, and B.J. McNeil 1991 Enhancing and evaluating diagnostic accuracy. Medical Decision Making 11:9-18.

Thompson, C.L., and M. Richmond 1990 Teaching home care for ventilator-dependent patients: the patients' perception. Heart and Lung 19:79-83.

Torma-Krajewski, J. 1986 Analysis of Lifting Tasks in the Health Care Industry. Paper presented at the University of Washington Symposium on Occupational Hazards to Health Care Workers. Seattle.

Tversky, A., and D. Kahneman 1981 The framing of decisions and the psychology of choice. Science 211:453-458.

Van Cott, H.P. 1991 Human Error in Medical Devices. Paper presented at the Symposium on Human Factors and Medicine, 35th annual meeting of the Human Factors Society. San Francisco, September.

Viall, C.D. 1990 Daily access of implanted venous ports: implications for patient education. Journal of Intravenous Nursing 13:294-296.

Walter, H.J. 1989 Primary prevention of chronic disease among children: the school-based ''know your body" intervention trials. Health Education Quarterly 16:201-214.

Wegener, D.H., and L.A. Aday 1989 Home care for ventilator-assisted children: predicting family stress. Pediatric Nursing 15:371-376.

Wilkins, E.S. 1989 Towards implantable glucose sensors: a review. Journal of Biomedical Engineering 11:354-361.

Williams, G.M. 1991 Causes and prevention of cancer. Statistical Bulletin of Metropolitan Insurance Companies 72:6-10.

Wise, D.L., ed. 1990 Bioinstrumentation: Research, Developments and Applications. Stoneham, Mass.: Butterworth.