Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
III New Ideas for Safely and Monitoring
Information Technology and Blood Safely ]. Michael Fi~zmaurice Assume there is a newly discovered virus in donor blood that is detectable by a test, and it is associated with a small number of people initially. You want to trace those who received the infected blood, and you may also want to trace those donors who were potentially exposed via their reported behavior. Maybe you can find that out through answers to their blood and interview questions. Do you have the information system to do it? What would it take to develop that capability? I want to talk about what a computer-based patient record is, what a clinical decision support system is, what evidence there is that they work, what barriers stand in the way of the development of a computer-based patient record system, and then what is the applicability to a blood information system. A computer-based patient record is just a collection of health data in electronic form and is part of the health information system of a physician or of a hospital. It could also be part of a health information system in a managed care organization or an insurance company. In electronic form that information is legible and available. When you ask for it, with the proper authority, you can get it, and if somebody else asks for it, they can get it at the same time. It is communicable, and you can search it. It is also a powerful tool for organizing patient care data, improving patient care, and strengthening communication among health care providers. That may be one of its strongest points: retrieving medical knowledge that is applicable to that particular patient at the time you make a decision. The operating hypothesis is that computer-based patient record systems can improve both physician performance and patient outcomes of care. The primary role for this collection of electronic data is to support the delivery of care to a particular patient. It brings the information to the physician, promotes communication, documents care, and records the reasoning behind the choices that are made. The secondary role of the computer-based patient record is to build a clinical data repository. All the data do not have 69
70 BLOOD AND BLOOD PRODUCTS: SAFETY AND RISK to be linked together and located in one place, as long as you can find what you want and pull it in. My agency, the Agency for Health Care Policy and Research (AHCPR), was created by Congress to determine what works in the community's practice of medicine. We know what works in the ideal setting of a clinical trial. In the community, patients behave differently and skill levels are not all the same. What are the differences? Why do we find so much variation across the country in procedures and in patient outcomes? These are important questions. The computer-based patient record is valuable for improving population-based care because you can query a large number of records and learn a great deal about the community factors that lead to those patient conditions and outcomes of care. A clinical decision support system, utilizing computer-based patient records, is simply computer software that aids decision making by providing diagnostic suggestions, treatment suggestions, testing prompts, drug alerts for potential drug-drug or drug-food interactions, therapeutic protocols, and practice guidelines. AHCPR promotes the development of practice guidelines. We would love to be able to drop them into computerized patient record systems to find out what difference they make in the process of care and patient outcomes, and even to research and test patient care pathways. A clinical decision support system requires a knowledge server, and that is simply a link between the patient and the information necessary for that patient's care. Normally, a physician is that link, but a physician using a mechanical instrument like a computer and a computerized record can do so much more and doesn't have to worry about forgetting. What does a good clinical decision support system require? It requires a body of data about the patient that a physician considers important. It requires knowledge sources; if this, then that, and then that. How do I test for this; how do I test for that? Suppose I suspect a given diagnosis? In what order should I do the tests; do I do an X-ray, then an MRI (magnetic resonance image), then a CAT (computer-assisted tomography) scan, and then a PET (positron emission tomography) scan? It requires a knowledge server that is the link between the computer-based patient record and the information source. You can describe the components of that knowledge server as medical logic modules, sentences that say, "If this is a female patient over 50 years old, and you don't find a Pap smear result in her record, then suggest that she get one. If you don't find a mammography result in her record, suggest that she get one. If the patient is 65 years of age, has asthma, and there is no evidence of an immunization for the most recent flu virus, suggest that the patient get one." It also needs a common nomenclature so that physicians talk the same language and can query the
NEW IDEAS FOR SAFETY AND MONITORING 71 computerized knowledge base and can query the scientific literature using the same language, or else you need a good translator for that language. These decision support systems improve patient care processes. In some cases they have measurably improved patient outcomes, but most of the studies that have looked at such systems are about improving patient care processes. For example, providing preventive care infonnation to physicians and their patients improves compliance with immunizations. However, if you take the computer assist away, then immunization goes back down to previous levels. These systems also support diagnosis of high-risk patients, and of course, you often treat high-risk patients differently than you treat low-risk patients. The systems also help determine the sometimes toxic drug dose for obtaining the desired therapeutic levels. With these systems we can avoid having to look up in tables the many different patient factors and doing all of the extensive research necessary for a particular patient. If you haven't seen a patient like that for awhile, the computer can help you remember drug and test ordering. In a clinical trial supported by AHCPR, the researcher had one set of physicians with regular computer screens for drug and test ordering and another set of physicians with special computer screens for drug and test ordering that were geared to question drug selection.54 Both groups of physicians had screens and both groups did test ordering. Those with the special screens designed to question the choice of drug or test on the basis of the suspected diagnosis and medical knowledge found in the literature and to put cost savings suggestions up front wound up saving money and reducing lengths of stay. However, those physicians spent about 5.5 more minutes per patient per 10-hour shift. Thus, you have technology that improves the process of care, but costs the physician time. The physician does not get paid any more for the extra time. There must be some way to get the benefits down to the decision maker for taking the time to enter the information and query the system, thereby improving the care of the patient and the process of care. Other studies, also funded by AHCPR and its predecessor, the National Center for Health Services Research, showed improved blood use through computer-based screening of orders, and that a computer-based patient interview elicited more HIV-related risk factors in the health histories of blood donors than one conducted by a staff member. These donors considered being queried by a computer to be more private, and they were more honest with their answers. We need to look at improving the sensitivity not only of the lab screening but of donor screening as well. 54Tierney, WM, ME Miller, JM Overhage, CJ McDonald (1993). Physician inpatient order writing on microcomputer workstations: Effects on resource utilization. Journal of the Almerican Medical 'association, 26963): 379-383.
72 BLOOD AND BLOOD PRODUCTS: SAFETY AND RISK Clinical decision support systems also reduce malpractice threats. Massachusetts physicians have lower emergency room malpractice premiums if they use a particular system. The reason is not because the system results in better decisions, but because it produces better documentation. More court cases are lost because of a lack of documentation than because it is proven that someone made a bad decision. If we are going to use these clinical decision support systems, what are some of the questions we need to answer? Most of the studies that I described were single-site studies in academic medical centers. We need more multiple- site tests of the same system. We need to know how much user perfonnance is enhanced. We need to know how much patient outcomes are improved. In 10 randomized trials that were studied, only three of them measured patient outcomes. The others dealt with the process of care. You might assume that improving the process leads to improved patient outcomes, but that can be a big assumption not necessarily a wrong assumption, but a big one. We need to know the costs of achieving these gains. Specifically, knowledge is needed about the costs of equipment, training, time, and energy spent convincing other peers to use a computer-based patient record system, as well as the costs of maintaining both a paper-based system and a computer-based system for a time so that people feel comfortable with the new system. We also need to investigate whether system-wide health care costs are contained, or we are just swapping one set of costs for another with no net gain. We need to support an accessible medical knowledge base, and medicine is doing a good job of building that base. We also need to determine how strong a match of medical terms is needed, because we don't all talk the same language. How much do our words have to mean the same thing before we can communicate medical information upon which others are willing to act? How do we integrate clinical decision support systems into the environment and provide these systems with features that physicians will want, whether that involves communicating by typing, writing, speaking, or pushing a button. The mechanism itself may be on your desktop, or maybe its a clipboard, or perhaps its hand held, or even at the bedside of the patient. Whatever it is, it must be flexible enough to improve the provider-patient relationship. There is very little written in the scientific literature about how you best go about the implementation of these systems. What has led to success and what has led to failure? We know that physicians are reluctant to enter data; there is an uncertain effect on physician productivity and income. The confidentiality and privacy of physician's text notes are important to them, and there is also a variation in use of clinical decision support systems within a given hospital department by department. Those who are used to working with high-tech equipment are more willing to work with a computer-based patient record. Furthermore, there is insufficient information in the scientific
NEW IDEAS FOR SAFETY AND MONITORING 73 literature about the factors that lead to the successful implementation of a clinical decision support system in hospitals and physicians' offices. One thing is certain: you will know they have arrived when managed care organizations begin using them extensively. There are many extended uses of computer-based patient record data beyond the care of the specific patient. Among these are their use for research, quality assurance, patient care treatment paths, treatment strategies assessment and medical technologies assessment. Computer-based patient record data can provide infonnation to a wide range of users for making many different choices, depending upon how they are aggregated and analyzed. This patient care data should be · uniformly defined, · linked accurately · collected together into databases, and · held confidentially. This is an ambitious vision. We don't as yet have widespread computer-based patient records. Some patient care data are normally computerized, such as lab records and radiology test reporting. Unfortunately, what happens is that the data are printed out of the auto analyzer onto paper and the paper is sent to the patient's floor, where two or three holes are punched in it and it is slapped into a paper medical record. We lose the advantage of computerization. Today many decisions are based on data of inferior quality. Many decisions are made on the basis of claims data, not just Medicare claims data, which are the best in the world, but also private insurance company claims data. Decisions are being made on the basis of those large claims databases without sufficient clinical information. Clinical, medical record patient care data are becoming more valuable for decision making. For example, Kaiser Permanente has planned over the next 5 years to spend millions of dollars developing and implementing a computer-based patient record system. The Mayo Foundation is also developing a computer-based patient record system. To support the growing private sector use of computer-based patient records, AHCPR and the National Library of Medicine are funding eight test beds to test the commonality of nomenclature and how easy it is to exchange clinical information across different computer-based patient record systems. What do you need to know if you are going to get involved in computer-based patient records? First, you need standards. You need definitions for signs, symptoms, and conditions of patients. That is, you need codes for the diagnoses of patients and procedures performed on patients, and they have to be used for more than just billing. You also need to define file contents. What data do you want to see to make a particular decision about
74 BLOOD AND BLOOD PRODUCTS: SAFETY AND RISK whether to admit a patient to the hospital or to order another series of drug tests or order an X-ray? You need to define those data and be able to pull them up electronically for that decision. Second, you need confidentiality and privacy protection for the patient care data. You need to define what is individually identifiable data, balance the benefits against the potential harm, and then specify what is a fair treatment of those data. In addition, someone has to be responsible for data quality assurance. When the data aren't accurate, somebody's head not the patient's has to roll. Electronically stored records are essential, but the current environment is not especially receptive to their use. There are variations across states in health care privacy laws, regulations, and practices. Some states have "quill pen" laws, which means you have to sign in pen and ink for the medical record to be admissible as evidence in court. This rules out electronic signatures. Third, standard unique identifiers for patients, health care providers, and payers must be developed. These are essential to obtain economies of scale in information technology. Fourth, malpractice concerns must be addressed. Are you giving medical advice across borders when you are dealing with drug information? Do you need to be licensed in each state in which you are exchanging information? Who has legal jurisdiction in case there is an exchange of erroneous information and the patient is hanned? Fifth, the security and integrity of your system must be ensured. You will have to deal with purposeful violations of privacy as well as accidental violations. You also have to deal with the accuracy of medical knowledge in clinical decision support systems and the accuracy of data transmissions. Who is to blame if important data get lost in the phone wire or an electronic switch or a faulty file server? Who is legally responsible for bad patient outcomes due to a flaw in a transmitted image or misreported medical knowledge? We need better benefit/cost methodologies, but whose benefits and whose costs should be the focus of study? We have to assess the business risks, as well as the business benefits, because, after all, this is a business investment . . ( .eclslon. Once you have computerized patient records, can you pull the patient care data together into a regional health repository? How is it governed? Who owns the data? Who gets to use the data? Who sets the rules? Do the owners set it? Does the regulatory agency set it? Do you want all these data packed together in a centralized database or distributed? If distributed, every time you send a question out to be answered, do the data suppliers get to decide whether they want to answer your question or withhold the data?
NEW IDEAS FOR SAFETY AND MONIT0~2ING 75 What about the use of computerized patient records to register blood information? Let us assume that the purpose is to reduce the incidence of blood-borne disease by identifying and excluding bad blood using a computerized database. A second purpose could be to trace back to the source of the contaminated blood. Let us assume that you already have a computerized system in place. Then you identify the hurdles and how to overcome them. We can assume that blood donors and recipients voluntarily report necessary personal information. What happens if they don't? Do you deny a transfusion to a patient who refuses to provide you with information? We can assume the patient consents to blood donation, collection of personal characteristics, linkage with other information, and disclosure, but to get that informed consent you may have to define exactly what disclosures you plan. The use of the database has to be governed. Who does it? Government? The private sector? At a national level? At a local level? Again, you have to identify the hurdles and who is responsible for addressing them. You have to have a disclosure policy. There is no federal privacy law that generally governs health data, except those related to AIDS and some other communicable diseases. By and large there is no federal law that protects the privacy of your own health data. You have to rely on your state and your health care provider. Who determines the user authorization and that an inquiry is really from that authorized user? You need a system of inquiry, and you have to decide how you are going to respond to the inquiries: by computer, by telephone, in writing, or in person? We have covered the database structure before when we talked about computer-based patient records, but do you have a database in each blood bank? Do you have a database nationally? You have to decide that. You need standards for this database, and you need to define the database content. What are you going to use to identify the person: Social Security number? Suppose the patient refuses? You say, "Well, then I will get the name, date of birth, sex, and mother's maiden name. That will give me 95 to 98 percent accuracy." But is that good enough? You have to have some individual or some institution monitoring the data quality: confidentiality and privacy, what laws apply, who bears the legal responsibility for this, how secure the storage system is, and what protections is built into your system? Then we get into system liability. Who is responsible? The caregiver is responsible for his or her own decisions, but when decisions are shared among several providers in separate states, how is responsibility determined? Who is responsible for the system liability? If you have information in the decision support system about contraindications, who is responsible for the integrity and accuracy of that information?
76 BLOOD AND BLOOD PRODUCTS: SAFETY AND RISK Finally? how do we overcome these barriers? Many people believe there should be a national privacy law that would set out appropriate uses of personally identifiable health data and specify the conditions under which they can be used. What are the penalties if somebody misuses these data? How much do they have to pay? How long do they have to spend in jail? Should there be federal relief from liability due to system failure? Somebody ought to be liable for it. You might say, "Well, we can take care of that with federal no-fault compensation." Yes, but it comes out of the pockets of the taxpayers and creates undesirable incentives. Some things that could be useful to get a system off the ground may not be wise in the long term. There should be informed consent forms for both the donor and the patient. There should be a review of existing secured systems to know what kind of model to adopt for a secure blood bank registry system. And there should be a development of Finding sources. Who benefits from this? If the public benefits, should there be a public payment for this? If it is solely a private benefit, then do you add the cost to the charge per unit of blood? Do major health insurers like the Health Care Financing Administration agree that a computerized system of blood records is needed, and will these insurers pay the additional charge? All of this should be pilot and market tested. You cannot just rush into it. You need to pilot test it in a couple of places, work out the kinks, look at the benefits, look at the costs, and then move ahead. What is at stake is the potential to increase the value and security of our national blood and blood product system by using applications of health information technology. A greater knowledge of the barriers and hurdles will lead to better public and private solutions.
Strategies for Dealing with Potentially Infected Recipients Ernest R. Simon The basic premise underlying strategies for dealing with potentially infected recipients is that recipients may have a need, and certainly have the right, to know if they may have been infected. Clearly, with the right to know goes the right not to know. There is substantial evidence that, when asked, many recipients elect not to know if they may have been infected. The second basic premise is that look-back is ineffective, inefficient, and certainly costly, more so for hospitals and physicians than for blood centers. Only a tiny portion of the time, effort, and cost of current look-back programs for human immunodef~ciency virus (HIV) and human T-lymphotropic virus (HTLV) I/II yield productive results. With hepatitis C virus (HCV), the problems are magnified. The issue is not look-back versus no look-back, but look-back versus something better than look-back. Look-back provides a cosmetic approach; I propose a substantive solution. There are two situations, those dealing with future recipients (which I will discuss) and those dealing with prior recipients. For future recipients, a transfusion episode should be considered to consist of three distinct components: pretransfusion, the transfusion itself, and posttransfusion actions. Several weeks before the anticipated need, the prospective transfusion recipient should be provided with information regarding the benefits of transfusion, as well as risks and alternatives. In addition, the rationale for testing the potential recipient for infectious disease markers both prior to the transfusion and after the transfusion should be clearly explained. The patient should be told of possible personal benefits that testing might provide. Conditions may be revealed that modify the type of care that the patient should receive and may even help with diagnosis. It is important that the testing in no way interferes with the appropriate care that the patient may get on the basis of the results of the test. In addition, testing may possibly help identifiable partners of the patient avoid infection if the patient practices appropriate preventive behavior. 77
78 BLOOD AND BLOOD PRODUCTS: SAFETY AND RISK A second rationale deals with the possible benefit to hospital staff if they know of the presence of certain potential infectious agents in the patient. Despite the practice of universal precautions, such knowledge may provide additional safety to the hospital staff. And finally, pretransfusion testing may provide a baseline to assuage liability concerns. It is clear that such viral marker testing is not necessarily limited to potential transfusion recipients. One can and should consider testing all surgical patients, or for that matter all patients, because the likelihood of the presence of infectious agents in patient populations is at least two orders of magnitude higher than it is in the volunteer donor population. Therefore, the yield of positive results from testing hospitalized patients is very much higher than is the yield from testing donors. Such viral marker testing is feasible for elective surgery. In the case of emergency surgery, a pretransfusioh sample could be collected and held for subsequent testing as necessary. The third phase of the transfusion episode takes place 6 to 12 months after the transfusion and completes the transfusion episode. At this time, the physician is reminded to contact the patient and/or the patient is reminded to contact the physician. The testing is repeated. If reactive results are obtained, confirmatory testing is done and the patient is counseled appropriately. This approach is analogous to other follow-ups, for example, in surgery, cancer cell therapy, and so forth. For hepatitis B virus (HBV), HIV, HCV, and HTLV I/II, current testing may benefit the patient, a third party, or both. It is clear that both are benefited with HIV and HBV testing, but with HCV it is unclear whether a third party is benefited in addition to the patient. For HTLV I/II the benefits to the patient or third party are questionable. With current screening of the blood supply and pretransfusion testing of the patient, the incremental additional yield of positive results at 6 to 12 months posttransfusion would be expected to be very low. The advantages over look-back are significant. Because posttransfusion follow-up is independent of the donor, it casts a wider net. It does not depend on a repeat donor who is now positive for the marker. It includes recipients of blood from one-time donors and repeat donors who have not returned or who have since been eliminated from the donor pool by surrogate tests performed prior to anti-HCV testing and donors who were subsequently disqualified without donating. Furthermore, it reduces or eliminates unproductive administrative complexities including a tortuous and sometimes flawed records trail and interventions and follow-ups by blood centers, transfusion services, hospital records departments, and multiple physicians. It targets surviving recipients and avoids tracing deceased patients. It embraces future transmissible agents with ease and shortens the interval between putative transmission and
NEW IDEAS FOR SAFETY AND MONITORING 79 detection. Finally, by acknowledging that zero risk for transfusion is not achievable now, it should be a powerful incentive to decrease the inappropriate use of blood components. Implementation is simple. Informed choice, by which the patient is informed of possible unexpected outcomes, should precede the transfusion episode. On discharge the message can be reinforced and specific information regarding mechanisms for follow-up can be provided. The responsibility for follow-up could rest with the patient as it should, or it could be expanded by including the physician. If the patient is transfused and discharged alive, the hospital information system could trigger automatic reminders to the attending physician and/or the patient after ~ to 12 months. An acceptable approach is essential. The specifics are not. By linking it to appropriate community and physician education with the active involvement of the U.S. Public Health Service, state and local health departments, medical and blood service organizations, blood centers, hospitals, manufacturers, and in particular the national media, routine posttransfi~sion follow-up becomes an extension of the HIV program recommended by the Presidential AIDS Commission and the American Hospital Association to its member hospitals. The message must emphasize that transfusion accounts for a small fraction of these diseases and that the focus on transfusion recipients is merely a part of our overall health strategy. In summary, this approach to future recipients does not mean look-back versus no look-back, but look-back versus something better than look-back. The testing approach advocated is substantive. Look-back is cosmetic. A substantial yield of pretransfusion testing may be expected. The additional incremental yield of posttransfusion testing is expected to be very low given the sensitivity of currently used tests.