Analysis of data quality for integration into therapy decision-supporting systems in the patient model "larynx carcinoma"

Introduction:

Clinical Decision Support Systems (CDSS) based on Bayesian networks (BN) have the potential to map complex diseases and to simulate treatment options and outcomes. Data quality is a known problem in the development of CDSS. Validating a TNM staging network as part of a digital patient model "laryngeal carcinoma" identified four data quality issues. To collect data of optimal quality, model-based data entry forms have been developed.

Methods:

The prototype of an input system was implemented, which extracts all parameters of the BN and prepares them in structured questionnaires. All relevant findings can be set in discrete parameters. Since findings are of varying reliability, a percentage reliability can be set for each parameter. The quality of the data and the user-friendliness of the questionnaires were tested in a study with four clinicians.

Results:

Initial prototype enhancements were made after initial evaluation, simplifying the process. The study showed that the data is collected completely and in the required quality. The usability was found to be adequate. Most of the time was needed for processing patient records.

Conclusions:

The developed prototype demonstrates the possibilities of optimized and high-quality data collection. In order to reduce a loss of time during file preparation, data entry should take place parallel to the reporting. For this reason and in order to enable the clinical use of CDSS, an integration in the hospital information system is aspired in the future.

Publication History

Publication Date:
18 April 2018 (online)