A clinical data analytics language (CliniDAL)

Abstract

This work aims to propose CliniDAL as a generic solution for a special purpose query language for Clinical Data Analytics, which can compute answers to any question that is answerable from a Clinical Information System (CIS), with a specific focus on addressing temporal issues in CISs. The research begins with the investigation of common questions in the clinical domain, which led to the definition of 6 classes of question and answer categories. A set of syntactic rules has been generated to support these query categories which enables direct question answering and supports answering research-oriented queries requiring data analytics functionalities. Also, a framework is proposed for scientific experimentation which resolves time-event dependencies in the queries. A free text search facility for string searching and semantic concept searching (using SNOMED-CT) is provided to enhance the results. In addition, a temporal model is proposed and integrated into CliniDAL which enables the use of natural language temporal expressions in query composition. A generic mapping algorithm is proposed using a similarity based top-k algorithm (accuracy of more than 84%), which automatically maps the query terms to the underlying data or schema of CISs with design models of ER, EAV or hybrid models of ER and EAV. Also, a generic translation algorithm is proposed to translate an initial text query to a set of SQL queries, mainly to resolve complexity of data extraction from CISs with an EAV model. The experimental results based on testing more than 270 clinical queries and 4 case studies reflect the capability of the language at creating the desired queries via the restricted natural language of the provided web-interface. So, it is not only much easier for naïve database users like clinicians to apply CliniDAL’s approach in comparison to an SQL approach, but also they do not need to have any knowledge of data or schema of the underlying CIS for the composition of queries.