Clinical Decision Support Systems to Improve Clinical Practice and Quality of Care

Glowing predictions about the all-encompassing and beneficial role of computers in medicine have appeared with increasing frequency in the scientific literature since the 1970s. However, this optimistic vision has not yet been realized almost 25 years later. A prescient commentary 15 years ago predicted the numerous obstacles that have prevented these rosy scenarios from coming true in clinical practice.¹ Several factors continue to echo the challenges faced in this area, including lack of investment; lack of leadership from practicing physicians, medical schools, and professional societies; and continuing control of information services in most health care organizations by chief information officers and other administrators.

Several articles in THE JOURNAL^{2 - 5} offer hope that the vision is indeed becoming reality. These articles address clinical decision support systems (CDSSs), which are clinical consultation systems that use population statistics and expert knowledge to offer real-time information for clinicians. These systems aid patient management and consultation through analysis of patient-specific information in comparison with an expert knowledge base. Clinical decision support systems should be clearly distinguished from operational decision support systems, which are corporate repositories (ie, data warehouses) of clinical and financial information that are scrutinized for utilization review, component cost evaluation, and clinician performance evaluation. Often referred to as health care archeology, these data warehouses are used to evaluate clinical and financial performance in groups of patients after care has been given. In contrast, CDSSs offer the potential to improve the quality and reduce the cost of care by influencing medical decisions at the time and place decisions are made. Clinical decision support systems run the gamut from simple alerting of potential drug interactions to interpreting results of blood gas analysis, critiquing orders for administration of blood products, automated medical diagnostic programs, and electronic disease management programs such as for antibiotic prescribing.⁶ Because these programs include patient-specific information in the analysis, they can transform often-ignored guidelines into dynamic programs that offer real-time patient-specific management advice. Indeed, our experience suggests that embedding guidelines in CDSSs rather than presenting them as static documents improves physician acceptance, markedly changes behavior and practice, and significantly improves quality of patient care.⁷

In this issue of THE JOURNAL, Hunt et al² review clinical trials of CDSSs published during the last 25 years. The authors selected studies for their relative methodologic purity and evaluation of the impact on physician performance and patient outcomes. The good news is that the number and quality of studies of CDSSs are increasing and in certain clinical areas, such as drug use and preventive medicine, these systems have been shown to improve physician performance and, less frequently, improve patient outcomes. Unfortunately, these studies cover a small fraction of all the CDSSs currently marketed. As the number of new companies developing CDSS products increases, clinicians may be bewildered by the number of different vendors with slick program demonstrations, some of which may have exaggerated and unsubstantiated claims. Aside from prompting skepticism on the part of practicing physicians, the extreme variability in quality and amount of evaluation of these systems underscores the need for all health centers to conduct in-house evaluation of any system before it is purchased and installed. Whereas a mistake in banking software may misplace several million dollars, a mistake in a CDSS may lead to the death of a patient. Enhanced oversight of these systems both locally and at the national level seems warranted, but by whom? This issue has grown large enough to attract the attention of the Food and Drug Administration, which has suggested regulation of CDSSs as medical devices.⁸ Miller and Gardner⁸ have proposed recommendations that, if implemented, may offer practicing clinicians some measure of confidence in marketed CDSSs.

Drug use represents the most common intervention in medicine and has the potential for costly and deadly consequences.⁹ Clinical decision support systems may help improve many aspects of the medication use process.^{3 - 5} In this issue of JAMA, Bates et al³ evaluate the effect of a computerized order entry system for physicians and multidisciplinary team intervention on preventing medication errors and adverse drug events (ADEs). Medication errors introduced at each step of the process were significantly reduced in the computerized order entry group. Although a multidisciplinary team intervention group, which included several medication-use process changes and an enhanced role for the clinical pharmacist, did not further reduce the ADE rate, previous studies have shown team interventions to be effective.³ Given the sheer complexity of the drug-use process and the multitude of potential failure points, computerized order-entry and drug management systems offer one of the most promising tools for decreasing medication errors, preventing ADEs, and improving drug use.^{9 - 10}

Also in this issue of JAMA, Raschke et al⁴ outline a program to identify and prevent ADEs in a community hospital (with far less information-system infrastructure than in the study by Bates et al³ ) using a commercially available clinical information system locally tailored for the project. The system not only identified potential problems before they recurred, but also established that in about a quarter of cases physicians were unaware of the potential problem. This study illustrates several important issues. First, local customization of CDSSs is necessary to fit local practice patterns, local standards of care, and local workflow issues. This approach is in stark contrast to common vendor claims of "turnkey" or "plug and play" clinical information systems that need little local modification. Second, rather than sending alerts as e-mail messages to physicians (with a predictably slow response), this study used pharmacists as a human link to interpret the volume of alerts, contact physicians by telephone with only appropriate alerts, and discuss management options. Accordingly, this study offers a generalizable method for detection and prevention of ADEs.

Monane et al⁵ present a different view of CDSSs through the use of data routinely collected by a pharmacy benefit manager organization to improve pharmaceutical prescribing in elderly patients. Although changes in prescriptions occurred in only 24% of patients in response to alerts for potentially inappropriate drug use, given the patient population and the expected baseline rate of change of 2%, a clear impact on prescribing behavior occurred. This is a refreshing application of a system that has often been used by managed care organizations to triage patients or otherwise minimize costs.

That CDSSs will increasingly affect medical decisions is a foregone conclusion in the health care setting.¹¹ Who develops, implements, evaluates, monitors, and maintains these systems is an open question. With the rapid growth of the Web, the ability of CDSSs to affect clinical practice and health care quality is enormous, but it is unclear whether these systems will be used to control medical care or improve it. The articles in this issue of THE JOURNAL should be a clarion call to medical schools to develop active research and training in medical informatics and to professional societies to get deeply involved in the process of development, implementation, and evaluation of this potent new medical software. Resources beyond MEDLINE are available to the practicing physician confronted with CDSSs (www.amia.org, www.amdis.org).¹² Although the promises of the past are not yet met, their realization will depend to a large extent on effective leadership by practicing physicians, medical schools, and professional societies. Only through careful application and objective assessment based on outcome data will the potential for computer-based CDSSs to advance clinical practice and improve the quality of care be realized.