Search

The conclusions produced by an expert system were presented to the attending clinicians in a pediatric hospital to verify if it could improve the detection of infections. Clinician interventions were compared before and after the implementation of the system. The results failed to demonstrate a statistical difference between the clinicians' practice patterns before and after implementation. The clinicians' opinions about the system were assessed through a paper questionnaire following the experiment. The questionnaire results demonstrated that the respondents appreciated the conclusions presented by the system.

Computerized protocols that standardize clinical decision making for mechanical ventilation of ICU patients provided the tools for evaluation of the usefulness of SpO2, in clinical decision making. Systematic examination of the variability between clinical decision makers and of the variability of the link between SpO2, and PaO2, make it unlikely that SpO2, is actually used in a systematic or rigorous manner in clinical decision making.

The HELP computerized hospital information system routinely provides clinical, administrative, financial, and decision support functions at LDS Hospital, a 475-bed tertiary care facility in Salt Lake City, Utah. Computerized patient data is stored in a centralized database. Paper records are the permanent patient record with the computer record serving as a supplement. Our study sought to determine what fraction of the traditional medical record was contained in electronic form in the HELP database. Medical records from 100 hospital discharges were examined. Ten (10) hospital services were represented. Each page of each record was examined and classified as: 1) computerized data, 2) handwritten data, 3) typed data, 4) a form, or 5) other. On the basis of the number of pages, the computerized fraction of the chart ranged from 10.2% to 39.5%, depending on the service. The average was 25.6%. The amount of disk storage required to store the computerized data ranged from 2,400 to 83,000 bytes per patient, depending on the service. Plans for further computerization of the medical record are discussed.

Computerized protocols were created to direct the management of arterial hypoxemia in critically ill patients with adult respiratory distress syndrome (ARDS) and have now been applied routinely, 24 hours a day, in the care of 36 such patients. We evaluated the performance of the protocols by measuring how often the clinical staff followed protocol instructions. Since compliance with the protocols could also vary depending on the type of therapy, we assessed compliance with the protocols as a function of 1) the method of artificially ventilating the patients, 2) whether the protocol instruction was to increase or decrease the intensity of therapy or to wait for an interval of time and 3) whether the computer instruction was “correct” or “incorrect.” A total of 7,663 instructions were evaluated. The clinical staff followed protocol instructions 63.8% of the time in the first 8 patients and 90.8% of the time in the subsequent patients. Instructions to wait were more likely to be followed than instructions to change therapy. There was no difference in compliance between instructions to increase and those to decrease the intensity of therapy. “Incorrect” instructions were followed 27% of the time. These instructions may not have been clinically important. The mode of ventilation therapy did not affect compliance with protocol instructions. We conclude that protocols can direct the clinical care of critically ill patients in a manner that is acceptable to experienced clinicians.

Hospital-acquired infections (HAIs) and the selection of empiric antibiotics continue to create problems for physicians. We developed computerized methods to identify patients at high risk of acquiring HAIs. Data from 3,151 patients with HAIs were compared with 3,152 control patients. Logistic regression was used to create a model to predict HAIs. Computer programs now monitor hospitalized patients every day and infection control personnel are notified of high risk patients before the onset of infections. During a six month study, between 60 and 70 percent of patients with HAIs were identified before infection onset. We used similar methods to identify patient factors that can be used to predict the pathogens and to help select empiric antibiotics for individual patients. A computerized antibiotic assistant program was developed to be used by physicians both inside and outside the hospital. The computer program utilizes five years of microbiology data (over 13,500 cultures) and a knowledge base containing empiric logic from infectious disease experts to help physicians select antibiotics. The computer predicted a susceptible antibiotic in 238 of 250 (95%) patients with infections.