Computer assisted pregnancy management.

The overall goal of this project has been to design and implement computer-based decision making logic for use in the practice of obstetrics. This goal has been accomplished by using the knowledge of medical experts to develop a number of HELP sectors (computer decision algorithms) that diagnose and suggest patient-specific management actions for maternity patients in the clinician's office. Information obtained from physical examination, health history, laboratory data, and test results are all recorded in the patient's computer file to be evaluated by these sectors. Management decisions in caring for the maternity patient during pregnancy were analyzed to determine what criteria prompt an action by the obstetrician. When such criteria are satisfied (judging from the information stored in the patient file), the HELP system sends an appropriate message to the physician. These messages, which include action-oriented suggestions, alert the physician to potential problems and their solutions. HELP's ability to access a broad spectrum of data about a patient makes it well suited for patient-specific management decisions. In order to understand how well the system performed and to what extent the computer actually aided the clinician, the HELP sectors were tested in the offices of three obstetricians. These same obstetricians provided the medical knowledge used to construct the HELP sectors. During the nine month segment of 1982-1983, the HELP sectors were run on 185 maternity patients. It was found that for 102 of these patients (55%), the computer suggested one or more appropriate management actions that would assist in the care of the patient during her pregnancy. This implies that 83 of the patients (45%) needed no special attention from HELP. (All 185 of the patient reports supplied information regarding the duration of the pregnancy, the estimated data of confinement, EDC, and useful Rh compatibility information). The most common suggestions offered by the computer were reminders to screen for potential patient problems such as diabetes and rubella, although other clinically important suggestions were also made. In this project we have shown that expert medical knowledge can be modeled in a computer and suggest appropriate management actions and flag pertinent findings for the care of maternity patients in an office setting.