Assessing Data Quality in Disease-Specific Patient Information Systems

BCB Medical provides disease-specific patient information systems for collection of patient data, and information services for clinical research as well as for public and private hospitals. There are over 100 disease-specific systems, each collecting unique and defined information for entire treatment path of a patient. As the data is entered to the systems by healthcare professionals during clinical work, or by patients through validated questionnaires for outcome measurement, the quality of the data needs to be carefully assessed before analyses. In this thesis, a quality report assessing the completeness and accuracy of the data collected by two of the largest disease-specific systems is created. For creating a report, that can be used to assess data quality in various different systems with diverse structures, fields common in most systems were selected for completeness assessment, and general rules for plausibility of the values were created for measuring data accuracy. The results of quality report were comparable between two systems studied in the thesis and enabled the analysis of features affecting the results. Also, the possibility of using machine learning methods in completeness assessment for complex form data was studied by comparing Random Forest and Logistic Regression, and by creating validation rules for the models. The resulting quality report will be deployed in other ongoing internal projects in BCB Medical