Use of free text clinical records in identifying syndromes and analysing health data.

This study discusses the development of a systematic technique for analyzing large volumes of clinical records to assist in epidemiological research. Implementing this method, the researchers categorised reasons for retirement of thoroughbred racehorses in Hong Kong, showing high success rates and efficiency.

Introduction

• The article focuses on the potential of utilizing clinical records as a resource for planning and deriving new research initiatives in epidemiology. It discusses how large volumes of clinical data can be used to identify patterns and trends in health data.

Method Development

• The researchers devised a method to systematically reduce the large volume of textual data into fewer content categories. This was achieved through developing explicit rules for user-defined coding.
• The goal of this process was to facilitate data analysis by generating a more manageable volume of information to handle. This technique sought to preserve as much original information from the records as possible, while making the data easier to work with for epidemiological studies.

Implementation of the Method

• The methodology was applied to categorize the reasons for retirement among a large dataset of 3727 thoroughbred racehorses in Hong Kong across several racing seasons.
• This process was conducted via the use of an automated user-defined dictionary, created to sort and classify the reasonings behind these retirements.

Results of the Implementation

• The article reports that the automated process was able to successfully categorize 95% of the records.
• The remaining 5% of records that could not be automatically categorized were manually assigned to one of the dictionary categories, demonstrating the necessity of human intervention despite the high efficiency of the system.
• The entire process from initial screening to final categorization of all records took approximately 100 man-hours to complete. This indicates that the method developed was not only effective but also quite efficient considering the large volume of data it was designed to handle.

Conclusion

• The study concludes by highlighting the potential of the developed method in efficiently handling and analyzing large volumes of free-text clinical records. The successful implementation and high success rate shown in this study indicate the method’s potential utility for epidemiologists and other researchers analyzing similar types of data.