Using a Text Mining Approach to Explore the Recording Quality of a Nursing Record System

Most nursing records in Taiwan have been computerized, resulting in a large amount of unstructured text data. The quality of these records has rarely been discussed.

This study used a text mining method to analyze the quality of a nursing record system to establish an auditing model and associated tools for nursing records, with the ultimate objective of improving the quality of electronic nursing records.

This study utilized a retrospective method to collect the electronic nursing records of 6,277 patients who had been discharged from the internal medicine departments of a medical center in northern Taiwan from January to June 2014. SAS Enterprise Guide Version 6.1 and SAS Text Miner Version 13.2 software were used to perform text mining. Nursing experts were invited to examine the electronic nursing records. The text mining results were compared against a benchmark that was developed by the experts, and the efficiency of SAS Text Miner was examined using the criteria of specificity, sensitivity, and accuracy.

In this study, 27,356 nurse-formulated events were used in the analysis. The results of the nurse-formulated events showed an 8.08% similar error with system-formulated events, 29.72% were identified as necessary and appropriate names, 17.53% were retained, 10.15% involved error event names, and 34.52% were not classified. In this study, the sensitivity of SAS text mining in the training (testing) data set was 96% (95%), and the specificity and accuracy were both 99% (99%).

The results of this study show that text mining is an effective approach to auditing the quality of electronic nursing records. SAS Text Miner software was shown to identify inappropriate nursing record content quickly and efficiently. Furthermore, the results of this study may be included in in-service education teaching materials to promote the writing of better nursing records to improve the quality of electronic nursing records.