"The computer is illiterate when it comes to research evidence," says Ida Sim. Sim, a medical informatics specialist and physician, wants to nourish underachieving computers with a more palatable form of research data, and then whip them into shape so that they can help users find the best and most up-to-date research results to answer important questions about clinical care.

Computer scientists have made advances in a field called natural language processing, Sim notes, which now make it possible to program a computer so that it can understand the gist of a newspaper article. But a computer still cannot read a published research study. Nor can it then organize, present, or interpret information in ways that are useful to professional groups that review medical practice guidelines, or to health care policymakers, physicians and others interested in improving medical care. Sim aims to change this state of affairs, in part by changing the way research is published.

The hottest new medical findings are often published results from clinical trials, especially "randomized" trials, the gold standard of clinical research. Similar groups of patients in these studies are randomly assigned to receive an experimental treatment, or else a standard treatment or inactive placebo, and clinical researchers monitor how patients respond over time. Optimally, the researchers, like the patients themselves, are "blinded," meaning they do not know which participants have taken the experimental drug until the trial ends and the collected data have been prepared for analysis. But evidence painstakingly gathered from these clinical trials often gets lost in the shuffle. More than 10,000 randomized trials are published and indexed each year in Medline, a universally used online medical reference and abstract service maintained by the National Library of Medicine. What are the chances that a physician will be familiar with, or be able to quickly reference, read and interpret the studies that are relevant to the individual patients he or she will see during any given day? Next to none, according to Sim.

Electronic records

"It's just impossible for any doctor to keep up with the evidence," she says. "There is no reason why computers couldn't or shouldn't be helping us with that task."

Sim envisions a day when the patient data entered into an electronic medical record as part of the daily work will be used to instantly present physicians not just with up-to-date treatment guidelines, but also with automatically generated, pertinent summaries of the clinical literature with access to specific data from specific clinical trials. More ambitious physicians and researchers would thus have the option to probe deeper, not only within clinical-trial data -- to compile and analyze information on subsets of patients of interest, for example -- but also within data sets containing genetic and biological information. Such a capability will come in handy during an era of what some are referring to as "post-genomics medicine," when a patient's genetic profile will be matched with treatments that work best for individuals with that profile.

But Sim is taking one step at a time. With support from the National Library of Medicine and from a prestigious United States Presidential Early Career Award in Science and Engineering, she has launched an informatics initiative she calls the Trial Bank Project, a collaboration with the Journal of the American Medical Association and the Annals of Internal Medicine. "The aim is to publish clinical research not only in text form, so that humans can read it, but also to put it into a computerized knowledge base, so that computers can also read the evidence, understand it, apply it, and help us to apply it in turn," Sim says.

The reasoning incorporated into computer-based systems to support decision-making is only as good as the evidence from clinical trials or whatever other knowledge base it is applied to, Sim explains. As it stands, many groups are building physician-support systems so that doctors can use computers to help them in their practices. But each group has to build its own knowledge base, and each knowledge base almost immediately becomes out of date as new research is published. Moreover, there is no mechanism to automatically update a knowledge base; it has to be done by hand. Sim wants to show that it is possible to update automatically and make the information available as a common resource, by building a prototype system that does precisely that. Such a resource could serve as a foundation for groups that want to build their own systems.

Journal authors who choose to participate in the Trial Bank Project are publishing their research studies not only as text but also in a computer format that captures information about the study - such as facts about participants, protocols, references and data - in standardized fields using numbers and a selected medical vocabulary.

Sim has developed software for the retrieval, organization and presentation of the information, as well as a web interface (http://rctbank.ucsf.edu/Presenter/). The initial target users of the website are primarily "peer" reviewers of journal articles and biostatisticians, although article summaries geared to physicians are also posted. More consistent and complete presentation of research in this format will make it easier for reviewers to compare studies, Sim asserts.

Sim does not exhibit the kind of narrow focus or frustration with institutional constraints on achieving technical goals sometimes attributed to productive computer "geeks." In fact, she chose medicine, even though she already knew she wanted to also do research, in part because she has a strong interest in liberal arts and broader issues. About half of her research now is in the area of informatics, but she also collaborates on studies to evaluate the quality and efficiency of health care services, and she works in the clinic once a week.

"Informatics is a technical solution, but it is important that technical solutions be used appropriately in a much larger context," Sim says. "We need to know who pays for the technology, who gets reimbursed, who gets rewarded for using the technology. Unless you understand that, it doesn't matter how neat your computer systems are."

Sim lived in Hong Kong until she was 10, then in Canada, before moving to the US to attend college. "I was happy to come to UCSF, because there is a Chinese population here, and I can relate to that part of my culture in the clinic, which is important to me," she says. Chinese patients seek her out and are a majority of the people she sees in the clinic. "They are usually poorer people who don't speak English at all, who are very culturally Chinese, and who have great trouble navigating our health care system," she adds.

Manipulating information

medical informatics. "I'm a weird bird - I'm really the only formally trained medical informatics person on campus." Her unique training has enabled her to carry out duties usually assigned to more senior faculty. For instance, she is the associate director of the growing Program in Biological and Medical Informatics, one of only a handful of programs in the nation that are trying to train people in both bio- and medical informatics.

"We want to show students that the way that we generate, address and represent biological information - from the Human Genome Project - for instance, needs to be put in the content of how we use that information for clinical care, and vice versa," Sim says. "Medicine is a human activity, but it is also an activity in which we spend a great deal of time manipulating information. We need the computer to help us. There is a lot of need, and a lot of opportunity."