Web-based textual analysis of free-text patient experience comments from a survey in primary care.

摘要: Background: Open-ended questions eliciting free-text comments have been widely adopted in surveys of patient experience. Analysis free text can provide deeper or new insight, identify areas for action, and initiate further investigation. Also, they may be a promising way to progress from documentation experience achieving quality improvement. The usual methods analyzing are known time resource intensive. To efficiently deal with large amount free-text, rapidly summarizing characterizing the being explored. Objective: aim this study was investigate feasibility using freely available Web-based processing tools (text clouds, distinctive word extraction, key words context) extracting useful information amounts commentary about experience, as an alternative more intensive analytic methods. Methods: We collected responses broad, open-ended question on patients’ primary care cross-sectional postal survey patients recently consulting doctors 25 English general practices. encoded files which were then uploaded three textual tools. we used two cloud creators: TagCrowd unigrams, Many Eyes bigrams; Voyant Tools, reading tool that extract perform Keyword Context (KWIC) analysis. association scores occurrence certain tested logistic regression KWIC analysis also performed gain insight into use significant word. Results: In total, 3426 received 7721 (comment rate: 44.4%). five most frequent “doctor”, “appointment”, “surgery”, “practice”, “time”. two-word combinations “reception staff”, “excellent service”, “two weeks”. showed “excellent” significantly associated better (OR=1.96, 95%CI=1.63-2.34), while “rude” worse (OR=0.53, 95%CI=0.46-0.60). results revealed 49 78 (63%) occurrences related receptionists 17(22%) doctors. Conclusions: output serve springboard Text extraction show promise quick evaluation unstructured feedback. easily understandable, but require probing such establish context. Future research should explore whether sophisticated (eg, sentiment analysis, natural language processing) could add additional levels understanding. [JMIR Med Inform 2015;3(2):e20]