Dean F. Sittig

With the recent spate of articles proclaiming significant improvements in the delivery of high-quality health care achieved by using real-time clinical decision support systems (CDSS), many health care organizations are either trying to buy or build such capabilities. Whichever they decide to do, the following five elements are prerequisites: an integrated, real-time patient database, a data-drive mechanism, a time-drive mechanism, a knowledge engineer, and a long-term data repository. The following paragraphs describe each of these key elements and give examples of their utility.

INTEGRATED, REAL-TIME PATIENT DATA BASE

Such a data base integrates data under a common patient identifier from a variety of clinical and administrative sources including the pharmacy, clinical laboratory, admissions, discharges, transfers, nursing notes, and radiology reports. It stores and updates all data as soon as laboratory results are available, forming the basis of any real-time CDSS effort and enabling the knowledge engineer to implement logic that involves patient-specific data from multiple data sources. For example, to develop a CDSS that helps a clinician interpret blood gas data, the patient’s level of respiratory support at the time of the measurement is essential.

DATA-DRIVE MECHANISM

A data-drive mechanism in a computer system enables a knowledge engineer to set a flag so that a program can be activated when a particular type of data or data item (e.g., clinical laboratory results or a chest radiograph report) is stored in the database. Such a "triggering" event enables a KE to create automatic, real-time, asynchronous, clinical decision support systems. They are automatic in the sense that the clinician does not have to ask the computer to execute a particular program, rather the simple act of storing data in the patient's medical record starts the CDSS. Such systems are real-time in that they run as soon as the data is stored, instead of at a specified time of day.

KNOWLEDGE ENGINEER

The knowledge engineer (KE) is an informatics expert who is responsible for extracting and then translating the clinical knowledge into machine executable logic. The KE must have in-depth knowledge of the structure and meaning of the data recorded in the patient’s electronic medical record, a thorough understanding of the various knowledge representation schemes, and the analytical and social ability to discuss and help others choose between complex options in clinical and patient care scenarios. It is not necessary that the KE be a clinician or a computer programmer, although either skill set would be useful.

TIME-DRIVE MECHANISM

The time-drive mechanism on a computer allows the knowledge engineer to develop programs that will be executed automatically at a specific time in the future (e.g., 2am) or after a specific time interval has passed (24 hours after ICU admission). By running a program based on time, the knowledge engineer can create logic to remind clinicians to perform specific activities or to check that the appropriate action has been performed.

LONG-TERM, CLINICAL DATA REPOSITORY

The long-term clinical data repository contains patient-specific data from a variety of clinical sources collected over a period of several years. It allows the knowledge engineer, in conjunction with the clinical advisory group, to develop reliable statistical predictors of specific events. For example, one could develop a logistic regression equation that identifies the pathogen most likely to be found in a particular specimen and recommends the least expensive antibiotic. The database could also be used to identify potential problem areas, such as the percentage of patients with diabetes who have not had an HBA1C test performed within the last six months. Finally, one could use the database to test the newly created logic, a crucial step in the development of any CDSS.

dfs 2/1/99