Clinical decision support system, using expert consensus-derived logic and natural language processing, decreased...

Prior research has shown that endoscopy sedation errors can occur due to manual workflow limitations. It was already known that manual review of sedation orders is inefficient in high-volume settings. No prior work had resolved how to intercept rare but significant ordering errors effectively. This gap motivated the development of automated tools to enhance safety. Existing studies focused on general error rates but lacked specific interventions. The challenge lies in balancing high case volume with low error prevalence. Manual review becomes impractical when errors are rare. That uncertainty drove the need for a scalable solution.

The aim of this study was to reduce sedation-type order errors in outpatient endoscopy. The specific problem addressed was workflow limitations in manual error detection. The motivation was to intercept rare but critical ordering errors efficiently. Traditional methods failed due to high volume and low error prevalence. The researchers proposed using a clinical decision support system (CDSS) to automate detection. The CDSS was designed to integrate into existing workflows. It was intended to reduce the number of chart reviews needed per error. The goal was to improve safety without increasing manual workload.

The CDSS was developed by an expert panel using an agile approach. Patient-specific historical endoscopy records were queried by the system. Expert consensus-derived logic was applied to detect potential errors. Natural language processing was used to analyze unstructured data. The system flagged possible sedation order errors for human review. A retrospective analysis compared pre-pilot and pilot periods. Four months of pre-pilot data were analyzed alongside twelve months of pilot data. A total of 22,755 endoscopy cases were included in the evaluation.

The CDSS reduced the sedation-type order error rate on the day of endoscopy. The pre-pilot error rate was 0.39%, and the pilot rate was 0.037%. The odds ratio was 0.094, with a p-value less than 1e-8. Background prevalence of erroneous orders remained unchanged at 0.39% and 0.34%. The number of chart reviews needed per error decreased from 296.7 to 3.5. This reduction allowed for efficient integration into existing workflows. The CDSS intercepted rare but important ordering errors effectively. The system did not increase the overall error prevalence during the pilot period.

The authors stated that the CDSS significantly reduced endoscopy sedation-type order errors. The system achieved this without increasing manual workload or error prevalence. The CDSS integrated into existing workflows to intercept rare errors. The use of expert consensus-derived logic and natural language processing was key. The system's performance was validated through a retrospective analysis. The reduction in chart reviews per error improved workflow efficiency. The CDSS was able to function in a high-volume, low-error-prevalence setting. The findings suggest that workflow-integrated CDSS can enhance procedural safety.