PT  - JOURNAL ARTICLE
AU  - Mark Rolfe
AU  - Jo Powell-Bright
AU  - Daniel Dodd
TI  - Semantic pattern analysis of patient perceptions using automated co-occurrence information mapping
DP  - 2012 Sep 01
TA  - European Respiratory Journal
PG  - P1288
VI  - 40
IP  - Suppl 56
4099  - http://erj.ersjournals.com/content/40/Suppl_56/P1288.short
4100  - http://erj.ersjournals.com/content/40/Suppl_56/P1288.full
SO  - Eur Respir J2012 Sep 01; 40
AB  - Background: The majority of patients with chronic diseases seek additional information from the internet following medical consultations. Text analytics is a widely recognized, validated system for modelling the structure and information content of text. Here we describe a unique method for identifying and analysing linguistic information from the internet to provide quantitative, unprompted insights into patients' sentiments about their conditions.Methods: Boolean- and thesaurus-based, machine-learning software is used to conduct an iterative nonlinear search of the web for all relevant texts containing broad keywords related to a given chronic disease. Texts are then analysed by Leximancer (v4.0), a text-mining tool that identifies themes and concepts from large bodies of text using a statistics-based algorithm.Results: This innovative approach ensures that a comprehensive disease-specific dataset is captured from the web. Leximancer automatically identifies commonly occurring concepts (weighted combinations of words that co-occur within the text). These are presented visually as maps, showing the strength of the relationship between different concepts (relative frequency and inter-connectedness) to facilitate semantic classification. Positive and negative sentiments about specific aspects of the disease and its management can be identified and selected for statistical analysis, demonstrating the validity of this technique.Conclusion: The internet is a data-rich source of patient-to-patient and patient-to-healthcare professional communications. The sentiment analysis method described can facilitate broader understanding of patient perceptions of their disease and its management.