Abstract 549

Clinical pediatric databases have become commonplace in healthcare environments. However, clinicians and researchers have been unable to easily explore and utilize these rich sources of information, as currently available data retrieval tools are beyond the expertise of most physicians.

To address this, we have developed DXtractor, an easy to use web-based computer application implemented in the Java language that allows clinicians to search large clinical databases for specific groups of patients. These patients are grouped on the basis of one or more clinical and/or demographic concepts that users select. Boolean combinations of patient groups, as well as complex time-based relationships, can be used to further specify the target subpopulation of patients. DXtractor's user interface consists of a series of clickable buttons, each of which is labeled with a clinical concept, like "Diagnosis", "Labs", or "Age". When invoked, these buttons cause the display of a screen which allows the user to specify details for that type of query.

A clinical example is illustrative of the ease of use of the application. To find the group of patients seen at Children's Hospital who are female, have diabetes mellitus, and have delayed puberty, we construct the following series of serially linked patient subpopulations. We first generate the group of all patients with diabetes mellitus by clicking "Diagnosis", and choosing "Diabetes Mellitus" from the displayed list. We then narrow down this patient subpopulation to just female patients with diabetes by using the "Gender" query and choosing the "Female" option. We now identify girls with delayed puberty, first isolating all female patients ("Gender" query), and then limiting this set to those girls who have been noted to have Tanner Stage I breasts ("Clinical" query). Finally, we limit this set to those girls whose age ("Age" query) was greater than 14 years at the last time they were noted to have Tanner I breasts - i.e. delayed puberty. We have now generated the group of diabetic girls, as well as the group of girls with delayed puberty; all that remains is to find the logical intersection of the two groups. This is performed in DXtractor using the Boolean "AND" operator. Further refinement of this type of query is possible; for instance, we could have identified only the poorly controlled diabetic girls, by limiting the set of diabetic girls to only those who have had a history of elevated hemoglobin A1c values ("Lab" query).

DXtractor does not require its users to understand any of the underlying database structures, nor any particulars of the SQL querying language required to retrieve the data from the database. It allows non-programming clinicians and researchers to rapidly identify specific patient subpopulations, for either routine clinical needs, or for further investigation in study protocols. It allows for significantly improved access to large stores of clinical data which have previously been underutilized.