A PubMed search resulted in multiple instances of coding radiology reports into terminologies, but only one example of coding radiology reports using SNOMED-CT [
22–
27]. As far as we know, there have been no previously published attempts to map these types of patient imaging information to a standard terminology. While our approach to storing the study-related data in PACS involves multiple steps and interactions with different systems, it accomplishes an important task that has been previously hindered because of just this complexity. The data transformations and update steps have been automated to provide a convenient way of collecting patient history and safety information and storing it along with study images.
Good coverage of the needed elements was obtained using the SNOMED-CT terminology, with 93% of questionnaire elements able to be manually mapped to appropriate concepts in the SNOMED-CT terminology. The small percentage of elements that did not have corresponding SNOMED-CT concepts were almost exclusively anatomically based and had to be extrapolated to a more generic concept (i.e., the SNOMED-CT concept of “thigh” was used to represent the concept of “anterior thigh” from the questionnaire). This was very similar to other studies matching clinical information [
13–
19,
28,
29]. In cases where concepts also required location delineation, secondary concepts were created and submitted into the forms_data table as separate rows. For example, left leg numbness along with begin mapped to “Numbness of lower limb” was also mapped to “Entire left lower extremity”, “Entire right lower extremity”, or “Both lower extremities” depending on the patient answer.
The advantages of coding this information are multifaceted. If the patient comes to the imaging department in the future to have additional or repeat examinations, the information for the new questionnaire can be pre-populated and the patient will only have to review the information and add any interval changes. Answers are propagated over multiple questionnaires, which is especially important for text entry as the patient can look at entered information and add to them if additional details were remembered. This should increase the efficiency of the imaging department, although at this current time it has not been tested. In addition, this coded information can be manipulated by computer applications for decision support, research, and patient care both within our application, and potentially in an electronic medical record or other application that understand SNOMED-CT codes.
With this patient information now available to the radiologist digitally, there are two distinct advantages. Firstly, the radiologist responsible to protocol the study can review this information and further tailor each study to get the best possible images to evaluate the patient’s pathology. Secondly, the interpreting radiologist can use this clinical information to provide a more appropriate interpretation of the findings and correlate them to a more definitive diagnosis. These are areas of interest that we can explore after full implementation.
Data Mining
One of the more important yet less heralded outcomes of this project is the creation of a large store of patient history and safety information that can be used for data mining. Data mining is the searching of coded patient information to try and find patterns of cause and effect [
30]. There are approximately 56 elements per questionnaire available for data mining. With approximately 60,000 CTs and MRIs performed at our facilities per year that amounts to 3.4 million searchable data points per year that can be used with for data mining and outcomes research [
31–
34].
This project provides an example of the utility for collecting patient information using a coded terminology. The wide-spread us of computers in patient care, both by healthcare providers and patients, allows for multiple opportunities for acquiring coded patient information. Gathering this data supports outcome related evidence based healthcare. The recent push and monetary influx by the federal government to computerize the patient health record makes creating and using technologies such as these feasible.
1 Scott L. DuVall,2 and Richard H. Wiggins, III2
