1. Introduction

This course is about medical computing. By the time you complete it, you should understand the information needs of medicine and the ways in which computers can support those needs.

a. The setting

When we think of medical computing, normally it is hospital information systems that come to mind. These are the often large, centralized systems that support the wide variety of informational operations at the hospital. Equally important, however, are smaller-scale settings for medical computing, including community hospitals, convalescent hospitals, skilled nursing facilities, clinics, doctor's offices, pharmacies, and-increasingly in the future-the patient's home.

The enormity of changes that are already taking place in medical care and its financing, and the anticipated changes that are sure to come in the future ensure that the setting of clinical computing will continue to change dramatically. For example, although "in principle" arguments for the creation of computer-based patient records have been made for three decades, it seems that only the very down to earth desire to control the spiraling costs of health care has finally motivated many institutions (and, most importantly, those who pay for care) to foster the creation of medical record systems. Similar pressures for cost-effectiveness of care now motivate serious efforts toward "outcomes analyses" (figuring out what works, and how well), even though the scientific underpinnings of medicine have supported the same conclusions for centuries.

Medical computing is not something of value in and of itself. It derives its value from the good it delivers to the patient, to care givers, and to society. Many of the motivating factors that drive future developments in medical computing will, therefore, derive from forces in the broader field of medical care. Nevertheless, the challenge of doing medical computing well is to support these external needs in the best possible way. Further, revolutionary changes in technology can and do change people's fundamental conception of the needs of a field. A dramatically improved medical computing infrastructure that empowers patients to play a significant role in making decisions about and in carrying out their own care, for example, can change our notions of how health care is to be delivered and organized.

b. Modes of care

Development of medical computing takes place against a backdrop of enormous changes in society's conception of medicine. Medical care began with the cure model: A person is normally well, but on occasion he falls ill, consults a doctor, and (he hopes) recovers to a state of wellness. This model probably may go back to ancient societies; the shaman was mostly consulted when something was wrong. Historically, in fact, this approach, which involved mostly avoiding doctors, was a successful strategy. The abilities of practitioners were very limited, and hospitals were (and still can be) very dangerous places.

As our abilities to treat more serious diseases have improved, and as people have come to live longer lives, it is becoming common not to cure disease but to manage it. The middle-aged patient with coronary artery disease does not expect to be returned to his twenty year old state of health. Instead, many patients are in for long, chronic courses of therapeutic management, whose goal is to maintain their current state of health and to manage ongoing complications. As a result, many patients expect to have frequent and ongoing contacts with the healthcare system. Under this recurring care model, any individual episode of care relies significantly on what has happened to the patient before, thus placing greater demands on clinical record systems.

Preventing a disease is, of course, far better than either curing or managing it. This is true not only for the patient but also for insurance companies or governments, for whom the costs of prevention are almost invariably far smaller than the costs of care. As we come to understand the relationships between our lifelong behavior and the risks of disease, we realize that one of the most important factors in health care can be the degree to which the patient takes responsibility for his or her own behavior and care. This further expands demands on health information systems, which will now have an important role in educating individuals, even before they are patients.

The advent of detailed genetic testing for both existing and potential genetic problems is about to open a whole new era in health care. We are on the threshold of a period in which relatively easily testable characteristics of our own individual genetic endowments will determine our medical destiny, at least with a discomforting degree of certainty. The effect of this change on health care is still murky, but one can imagine that prevention will become a much more patient-specific matter. Thus, rather than the doctor advising a diet high in fiber and low in fat, a detailed genetic knowledge of a specific patient's susceptibilities to disease may suggest a range of prophylactic interventions, including change in dietary habits, locale, lifestyle, exposures to environmental factors known to be particularly dangerous to this patient, and even anticipatory medical or surgical interventions.

Medicine traditionally recognizes three important roles for the clinician:

diagnosis: determining what, if anything, is wrong

prognosis: predicting what will happen

therapy: intervening in the natural process of disease to cure or ameliorate it.

Much of today's clinical expertise concerns diagnosis and therapy. Indeed, diagnosis is considered subservient to therapy, because it is therapy that actually affects the patient's pathophysiologic state. Prognosis has traditionally been more important, because in an era when medicine could do less for people, at least predicting the ravages of disease was a useful guide to what the patient could expect. Of course, this need is still present but has been overshadowed in the rush to "treat the disease, not the patient" that is the hallmark of contemporary medical care. "Genetics as medical destiny" will probably serve to restore the importance of prognosis to the clinical skills set.

c. Motivations for health information systems

Billing

The traditional driving force behind the creation of computing systems for health care has been billing. This observation can lead to a cynical analysis of the motivations of health care institutions. It is clear that financial concerns are paramount in most of our society's established institutions, and the desire to get bills paid has been a powerful force driving the development of medical computing systems. Like any other business, accurate record keeping is essential to being able to allocate charges for services rendered and to justify to payors the charges being billed. Therefore, most large hospitals recognized quite early that the billable aspects of care must be carefully recorded, and this led to the development of the earliest hospital information systems. Not surprisingly, these systems had extensive records about the patient's insurance coverage, surgeries that had been performed, laboratory tests ordered, drugs given, etc., but relatively little about the patient's medical history, diseases in their family, or, indeed, even the outcomes of tests performed or medical therapy administered. Because only the actions taken were billable, not their outcomes, billing systems were interested mostly in recording billable events, not whether or not they succeeded.

Two of the early exceptions to this dismal picture were clinical laboratory systems and pharmacy systems. In the laboratory, where most of the data were already generated under control of a computer, it was obvious to many that throwing these already-digitized data away was a poor approach. Early uses of hospital-wide computer systems made rapid access to lab data possible, because the results did not need to await printing and distribution to the paper chart, and this feature was popular with physicians. Even at sites without accessible computer terminals, there were systems as early as the start of the 1970's where laboratory results could be accessed by automated telephone systems []. Pharmacy systems computerized early probably for quality control reasons. Printing labels for medications was easier with computerized systems, and once the data had been digitally captured there was no rational reason not to retain them. Further, early experiments with alerting and decision support systems in the pharmacy area suggested that capturing drug orders could help reduce potentially very serious drug-drug interactions and allergic reactions.

Third-Party Payments

Oddly from today's vantage point, the growth in health care expenditures through the early 1980's was hardly countered by any effective forces or policies. Medical billing was on the basis of "fee for service," and doctors autonomously determined what services the patient needed and therefore what services the patient paid for. Mostly through the spread of healthcare coverage as a benefit of employment, much of what was paid did not come directly out of patients' pockets but was paid by their insurers, who simply passed the costs along through increased premiums on future policies. At a time of sustained economic growth, this appears not to have occasioned much comment or generated much resistance.

The creation of Medicare under the Johnson administration in the 1960's made government-paid medical insurance available to all senior citizens, to be followed shortly by Medicaid and its State-specific variants, which extended similar coverage to the poor. These government programs raised the visibility of health care expenditures tremendously, because total spending was now visible as a large and growing item in each year's Federal budget, whereas previously the payment of direct fees and insurance premiums was very broadly distributed. HCFA, the government's Health Care Financing Agency, which ultimately pays the bills (through a system of private insurers who actually administer Medicare), has been the focus of several innovations that are having profound effects on health care financing. First, HCFA mounted the initial opposition to fee for service payments and introduced the notion that care providers must take some of the financial risks of care. Second, and more recently, the government has been urging the adoption of the Health Maintenance Organization (HMO) model of health care delivery.

If fee for service encourages physicians to do and charge for as much as possible, then what countervailing force can rationally oppose this arrangement? HCFA's notion was that they would pay a standard amount for each disease, and then leave it up to each provider to figure out how best to spend that amount to provide good care for the patient. Under such a system, for example, all patients admitted to the emergency department with chest pain and a suspicion of a heart attack (myocardial infarction) would be assigned the same diagnostic code and the institution would be paid the same amount for caring for all of them. This system, which has become known as the DRG (diagnostically related groups) system, is currently how HCFA pays for medical services. Thus, the unit of service is no longer the individual lab test, x-ray, surgical procedure, etc., but the totality of care for roughly one "episode." Payments for standard episodes are determined by looking at what it had cost to treat such cases under the fee for service system, by making rational arguments for what should actually be done, on average, for such patients, etc. The care provider can obviously make a profit on some patients in the DRG who happen to recover with little care (or die of their condition quickly) and can lose large sums on extended care for patients whose situation does not resolve rapidly. The definition of DRG's has evolved to differentiate among conditions with vastly different typical expenses to offset this problem, but at the heart of the system remains an insurance-like notion that reimbursement is based on services needed by a population, not the individual patient.

The other major innovation, the HMO, continues to move the care reimbursement model further in this direction. In an HMO, a care provider takes on responsibility for the total care of any needed medical condition by a population of so-called "covered lives," for a negotiated fixed fee. This model is also called "capitation," because payment is "per head." Reimbursement under this system is neither for the individual service, as in the fee for service model, not for the individual episode, as in the DRG model, but for taking care of an entire patient population for a fixed period of time. The HMO was pioneered in the post-WWII period by Kaiser Permanente, originally as a captive health care provider for employees of the Kaiser Aluminum and Steel factories in California. The integration of primary care, specialy clinics, hospitalization, convalescent care and insurance was sufficiently attractive that Kaiser's HMO was eventually spun off and began to provide coverage for many others, still usually as a benefit through their employers. Today, capitation is used not only in HMO's but even in large government-run health delivery systems such as the Veterans' Administration hospitals, whose budget is determined by the number of veterans who live in their catchment basin.

In the HMO model, as with DRG's, care for some patients costs less to provide than the agreed-on fee, but care for others may be much more expensive. The care provider must, therefore, think in actuarial terms, and of necessity takes on many of the characteristics of an insurance company as well as of a care provider. Indeed, some companies have specialized in creating "virtual HMO's." They have no doctors, no clinics, no hospitals; all such services are contracted, and the core business of such an organization is solely risk management. Indeed, it has been the success of such lightweight (some might say evanescent) companies to compete successfully with more traditional integrated providers such as Kaiser Permanente that has raised alarms about the instability of future health care provider organizations.

Both the DRG and the HMO model put a premium on better information systems, because profitability (and survival) of the providers is directly related to being able to provider adequate care for low cost. Thus, in addition to documentation requirements needed to satisfy payers that care is in fact adequate, providers have strong incentives to understand where their costs are and to discover what the cheapest adequately effective treatments are. Because effectiveness can only be judged by examining clinical aspects of the patient's care, this change has motivated including clinical aspects of the patient record into the medical computing system.

All interventions have side effects as well as direct ones, and the move to DRG-based payments and capitation has also spawned interesting unintended developments. For example, thriving businesses have been created to help provider organizations most advantageously characterize their patients' DRGs for reimbursement purposes. If a complication of one disease can be cast as an independent one, then the DRG scheme will pay for both. Thus, expertise in reducing the real-world complexity of an actual patient chart to a simplistic coding scheme can be very well rewarded. The HMO's natural response to underwriting risk is to try to assure that it covers the least sick of its potential patient population. If it can identify those most likely to wind up costing the most for their care and avoid them as members, then on average it will do much better financially. This desire to "cherry pick" establishes the conditions that motivate serious systematic invasions of patients' privacy, a topic to which we will turn later.

Institutional Agglomeration

Quality of Care

Outcomes

Consequences of History

One consequence of the genesis of medical computing systems in billing is that they have typically been designed, developed and controlled by individuals with financial, not clinical or technical training. Even in the mid-1990's, it is not unusual to see large hospital information systems being developed either at hosptials or vendors by teams that include no one with clinical experience. The traditional view is that if the system winds up awkward to use for clinicians, that's too bad, but if it winds up inappropriate for running the finances of the institution, that is a complete disaster!