Validation of Semantic Analyses of Unstructured Medical Data for Research Purposes

Abstract

Background In secondary data there are often unstructured free texts. The aim of this study was to validate a text mining system to extract unstructured medical data for research purposes. Methods From a radiological department, 1,000 out of 7,102 CT findings were randomly selected. These were manually divided into defined groups by 2 physicians. For automated tagging and reporting, the text analysis software Averbis Extraction Platform (AEP) was used. Special features of the system are a morphological analysis for the decomposition of compound words as well as the recognition of noun phrases, abbreviations and negated statements. Based on the extracted standardized keywords, findings reports were assigned to the given findings groups using machine learning methods. To assess the reliability and validity of the automated process, the automated and two independent manual mappings were compared for matches in multiple runs. Results Manual classification was too time-consuming. In the case of automated keywording, the classification according to ICD-10 turned out to be unsuitable for our data. It also showed that the keyword search does not deliver reliable results. Computer-aided text mining and machine learning resulted in reliable results. The inter-rater reliability of the two manual classifications, as well as the machine and manual classification was very high. Both manual classifications were consistent in 93% of all findings. The kappa coefficient is 0.89 [95% confidence interval (CI) 0.87–0.92]. The automatic classification agreed with the independent, second manual classification in 86% of all findings (Kappa coefficient 0.79 [95% CI 0.75–0.81]). Discussion The classification of the software AEP was very good. In our study, however, it followed a systematic pattern. Most misclassifications were found in findings that indicate an increased risk of cancer. The free-text structure of the findings raises concerns about the feasibility of a purely automated analysis. The combination of human intellect and intelligent, adaptive software appears most suitable for mining unstructured but important textual information for research.

Association between chiropractic spinal manipulation and cauda equina syndrome in adults with low back pain: Retrospective cohort study of US academic health centers

“…This feature includes mechanisms to understand negation (e.g., absence of a condition), intent, and context, thus aiding us in (1) excluding patients with prior CES as noted in free text in medical records, and (2) identifying new diagnoses of CES during follow-up. Prior studies have demonstrated that this software has acceptable accuracy, reliability, and agreement when compared to manual chart review for extracting clinical concepts related to diagnoses, laboratory values, medications, and symptoms [ 41 , 42 ].…”

Association between spinal manipulative therapy and lumbar spine reoperation after discectomy: a retrospective cohort study

“…The natural language processing feature of TriNetX uses Averbis software (Averbis, Freiburg im Breisgau, DE), which employs machine learning and rules-based algorithms to extract meaning from unstructured clinical text while incorporating mechanisms to interpret context, intent, and negation [ 21 , 22 ]. In previous studies, this software has demonstrated acceptable accuracy, reliability, and concordance with manual chart review in extracting clinical concepts related to diagnoses, symptoms, medications, and laboratory values [ 22 , 23 ]. Specifically, studies estimated an overall Kappa value of 0.79 (good agreement) [ 22 ] and F1 values up to 0.89 representing the harmonic mean of recall and precision [ 22 , 23 ].…”

“…In previous studies, this software has demonstrated acceptable accuracy, reliability, and concordance with manual chart review in extracting clinical concepts related to diagnoses, symptoms, medications, and laboratory values [ 22 , 23 ]. Specifically, studies estimated an overall Kappa value of 0.79 (good agreement) [ 22 ] and F1 values up to 0.89 representing the harmonic mean of recall and precision [ 22 , 23 ]. However, performance of this software may vary across different clinical contexts.…”

BeoNet-Halle—development of a multifunctional database for the automated extraction of healthcare data from general practitioner and specialist practices

“…Dieser Datenschatz ist in Deutschland weitgehend unerforscht, da hierzulande kaum geeignete Instrumente existieren, mit denen sich solche Angaben ausreichend anonymisieren lassen. Mittels Merkmalsextraktion werden zurzeit erste Ansätze zur Aufbereitung von unstrukturierten medizinischen Daten erforscht [ 23 , 24 ].…”