Keywording Patient Data by Computer

Abstract

The Vascular Surgery Laboratory at Harper Hospital is both a clinical and a research lab working as part of Wayne State University's Medical School. The lab sees patients with vascular disorders who are part of the general surgery service as well as seeing patients under the care of the physician in charge of the lab. To form a good data base, information gathered on patients is carefully recorded and filed. Much of this information is already being put to use in one or more of the research projects in the works. Most of the ongoing projects have to do with experimentation into non-invasive methods for diagnosis of arterial and vascular diseases. Some of these techniques are already pretty far advanced and are used frequently such as directional doppler measurements. For each patient seen in the lab, that patient's history and current diagnosis plus all of the laboratory's measurements become part of the patient's permanent record. The hope is that with a large data base it will be possible to make conclusions about disease-treatment trends based on some pattern of lab measurements. This is where the computer comes in. The process of informing the computer as to which are and which are not keywords in a given report can be done in several ways. It can be done manually where the author of a report indicates what keywords are applicable to that report. Another way of accomplishing this is to decide upon a list of all possible keywords and store these in the computer so that when a report is scanned by the computer words matching entries in the keyword list are saved as the keywords of the current report. The keywording method finally decided upon was a compromise between these two methods. There are advantages and disadvantages to both methods. The manual method is too inconsistant. It is difficult enough for one .author to remember what words have been used as keywords (eg. was it "occlusive", "occlusion", or "occluded") much less several different authors. On the other hand the automatic method requires that once a keyword has been entered in the permanent list it will always be picked out as a keyword. This may not be desireable. For example the word "stenosis" is sometimes used as a keyword but often it is just used in a very general, descriptive sense. The resulting compromise is based on the observation that most keywords can be "modified" with words like "upper", "lower", "arm", "leg", "carotid", "brachial", "left", "right", etc. The compromise result was that a list of all possible modifiers was stored in the computer but the author selects the keywords in each report. In the scanning process any time a keyword is indicated the computer's response is to search for any modifiers of that keyword. These keyword modifier combinations are then stored with the corresponding patient name. This solution turned out to allow the most felxible retrieval system. The retrieval system was designed to be an extremely flexible, conversational program. Its basic function is to select only those patients satisfying certain keywords and modifiers chosen by the operator but it is also equiped with many other helpful axillary features.