Databases in the Next Millennium

Because I am a clinician-educator-editor rather than someone who works with large databases, my contribution to this issue must necessarily be the product of an outsider looking in and consists primarily of putting database work into a larger context. The papers in this Supplement convince me that large clinical databases have begun to occupy a unique niche in the spectrum of clinical research, somewhere between the scientifically rigorous, highly controlled work (off-line research) exemplified by randomized, controlled trials at one extreme and the applied, pragmatic work (on-line or real-time research) found in quality improvement circles at the other. Prognosticating about the future role of databases (Do such databases even have a working future? If so, what will it look like?) is, of course, more difficult, primarily because prognosticating is generically such a terrible job. After all, even weather prediction more than 24 hours in advance is not very accurate, particularly when measured against chance, that is, the prediction that the weather tomorrow will be like the weather today; chaos theory has helped us understand why [1]. At the same time, I am lucky because, like most prognosticators, I am not likely to be held accountable for my predictions. Data and Quality: What Are the Questions? The principal question that concerns us here is whether databases will have a role in determining the quality of health care-and their costs. There are, of course, several points of view on quality, particularly on the question of whose perspective is most important in deciding what quality is (my own view is that the patient's perspective matters most) [2]. But the definition of quality and its importance in the future of the health care system, like the nature of the system itself, go well beyond questions of perspective; the very concept of quality is inextricably intertwined with our most basic cultural traditions, our social values, our political philosophies, our economy, and even our spiritual life. This makes the health care system incredibly complex and hard to understand and, correspondingly, hard to change. Lying just behind this principal question are two deeper questions. The first is Can data serve as the agent for meaningful improvement? The authors who have contributed to this issue clearly feel that the answer is yes. That is, they assume that if you can only get your hands on the right data (accurate, risk-adjusted data) and enough of it, people and systems will change, evolve, improve. The second question is the converse, namely, Is meaningful improvement possible without data? The authors give us less to work with here, but many who take medical quality improvement seriously have little doubt that the answer is no. For example, Kathleen Goonan, in The Juran Prescription: Clinical Quality Management, says it this way: You can't manage what you can't measure [3]. Unfortunately, each of these questions forces us to think narrowly and dichotomously, oversimplifying a complex situation. From what we know about the way in which innovations are adopted [4], either question is probably the wrong one, much as the question of whether women are more important than men (or vice versa) in creating children is the wrong one. For our purposes, the better question appears to be, What elements are necessary and sufficient for improving the quality of medical care? Many examples of change in the explosively developing quality improvement era have involved the use of data [5]. Even these examples, however, do not prove that the data themselves were the principal lever for change; other factors may have been at least as important, perhaps more so. In this essay, I suggest, first, that although data may often be necessary, by themselves they are usually not sufficient to bring about change. Second, I suggest that medicine, even at its cognitive, scientific, data-driven best, is always a social and emotional act. To the extent that this is true, new and better ways to practice are adopted only when data are tightly linked to the appropriate emotional and social forces. Third, I suggest that a respectable understanding of the role that social and emotional forces play in the diffusion of innovations has developed; indeed, it has become an entire discipline within the social sciences [4]. As in other areas, however, medicine has unfortunately had difficulty recognizing and accepting the lessons that this discipline has to teach, particularly those involving the crucial importance of attitude and values [6]. Finally, therefore, I suggest that a major task, if not the major task, for medicine in the next millennium with regard to the use of databases is to develop a deeper understanding of these social and emotional forces. Only then, I believe, will it be possible to put these forces to work in applying database information to problems of quality. The Use of Data as an Innovation: A Historical Glimpse Before turning to the future, it may be helpful to look to the past. If we can look forward to databases serving in the future as a major engine of change and progress, we should be able to find at least some convincing evidence that they have served in this capacity over a respectable period of time, that they have a historical track record. When did people start to use large databases (as distinguished from knowledge, such as that found in libraries) as the basis for improved management of complex systems? I am not a historian, particularly not a historian of management systems, so I cannot pretend to provide anything approaching a definitive answer. But I recently did come across a striking example of an early application of database information as an instrument of large system reform in Simon Schama's extraordinary text on the history of human concepts of nature, Landscape and Memory [7]. In the 1660s in France, Schama tells us, one Jean-Baptiste Colbert was appointed by Louis XIV to be the administrator of all French forests. Soon thereafter, Colbert warned his king that: La France perira, faute de bois (France will perish for lack of wood). Colbert was talking about the almost insatiable need at the time for wood for building ships of the line, as well as for heating and for fueling foundries and manufacturing of all sorts. Wood was, in fact, the principal source of energy (and some of the construction material) for the French economy and, hence, for French imperial power. Lack of wood was, in a sense, the fuel crisis of its time. Colbert also knew that a serious quality problem underlay this crisis. That is, he knew that the management of the French forests, royal and otherwise, was in chaos. He intended to set things right and, in some fashion still not completely understood by historians, he decided that to bring about the sweeping management reforms needed to avert a major fuel crisis, he needed data-big data-and he acted accordingly. In the French forests of the 1660s, the data collection scene was very different from that of today's health care system. There were no data abstractors in record rooms, no minicomputers humming quietly in corners, and certainly no practitioners entering data on the fly into handheld computers. Rather, Carriage loads of men in long wigs and long coats, carrying surveying rods and spools of horsehair twine descended on the forests of Normandy, Lower Burgundy, and the Ile de France. But data collection and analysis there were. Colbert's minions worked at the task for years, and (here comes the clincher): By the end of the 1660s, Colbert had the data he needed to act (emphasis added). The object, says Schama, as always in Cartesian France, was to bring order out of chaos; an object lesson, perhaps, for our less-than-Cartesian times. The result of Colbert's data-driven process was the great Ordinance of 1669: some 500 regulations, in more than 100 printed pages, that served as the Bible of French forestry for more than 100 years, well after the French revolution. Colbert's strategy was, by all measures, a bureaucrat's dream, because for many years it effectively did improve the use of French forest resources. Data and the People Who Use Them In Colbert's case, a large database seems to have been necessary to bring about change. This point should be of particular interest to us because without those data, it seems, even the absolute power of the French monarchy was not enough to bring French forest resources under control. But it is also arguable that the data alone would have had little impact if they had not been linked to the monarchy's awesome force. In sum, neither the social, administrative, and military power of the monarchy nor the data alone were enough to bring about the needed reforms. Although each was necessary, neither, by itself, was sufficient; improvement required both. Indeed, 30 years of recent intensive study has made it clear that both knowledge and emotional engagement are universally required for the diffusion of innovations, no matter what kind of new idea is involved, from the sterilization of contaminated drinking water by boiling, to the adoption of new fertilizers and seed in farming, to the use of fax machines and the Internet [5]. As Rogers conceives of it, the diffusion process for innovations always involves five stages: knowledge, persuasion, decision, implementation, and confirmation [8]. Turning to medicine, the Rogers diffusion model finds echoes in continuing medical education, which, until recently, was the major way in which the profession approached quality improvement. The model most widely accepted in medical education sees clinical work as consisting of four distinct elements (Table 1): knowledge, skills, performance, and outcomes. According to this model, the practice of medicine, particularly high-quality medicine, requires four related activities: understanding a clinical practice, knowing how to carry it out, doing it, and doing it well. Less