Rationale: Within an information society, where everyone should be able to access all available information, improving access to written language is becoming more and more a central issue. This is the case for health-related information which should be accessible to all members of the society, including people who have reading difficulties as a result of a low education level or of language-based learning disabilities or because the language of the text is not their native language. Moreover, the breakdown of doctor-patient communication is one of the most frequent cause of adverse events. Research questions: We conducted a preliminary investigation to assess the readability of a corpus of informed consent forms used before a clinical procedure in the hospitals of a Regional Healthcare Service. Secondary goals include the comparison of readability across specialties and healthcare trusts. Methods: Providing complex scientific information in a way that is comprehensible to a lay person is a challenge that nowadays can be addressed by resorting to advanced Natural Language Processing (NLP) techniques, which make it possible to monitor the linguistic complexity of texts at the syntactic and lexical levels and to support their simplification, whenever needed. The study has been carried out by combining NLP-enabled feature extraction and state-of-the-art machine learning algorithms. To this end we used READ-IT, the first NLP-based readability assessment tool for Italian. Results: We analysed 584 documents, covering 29 specialties, for a total of 607.790 word tokens, currently used at the 36 public hospitals in Tuscany. Although the readability level of all documents in the corpus is low, both at the lexical and syntactic level, significant differences can be observed between specialties and healthcare trust releasing the forms. With the readability level ranging between 0 (easy-to-read) and 100 (difficult-to-read), it resulted that the pediatric informed consent documents are the most easy-to-read forms (with an average score of 75) while the most difficult-to read documents are documents of the surgical area (whose average score is 80) (standard deviation 2). Discussion: The state of the art resulting from this preliminary study shows that NLP-based readability assessment tools can help to measure the linguistic complexity of informed consent forms and guide the editor to identify linguistically complex passages that need to be simplified, either syntactically or lexically. The use of an assessment tool designed for the general language is the main limitation of the study and should be addressed through the customization of the tool to assess the readability of the healthcare jargon. A further step of the research consider also the design of a guidance to prepare readable informed consent forms.
Language technologies for automatic readability assessment of health-related Information: a preliminary investigation into the informed consent forms used in a regional health service
Giulia Venturi;Simonetta Montemagni;Manuela Sassi;
2015
Abstract
Rationale: Within an information society, where everyone should be able to access all available information, improving access to written language is becoming more and more a central issue. This is the case for health-related information which should be accessible to all members of the society, including people who have reading difficulties as a result of a low education level or of language-based learning disabilities or because the language of the text is not their native language. Moreover, the breakdown of doctor-patient communication is one of the most frequent cause of adverse events. Research questions: We conducted a preliminary investigation to assess the readability of a corpus of informed consent forms used before a clinical procedure in the hospitals of a Regional Healthcare Service. Secondary goals include the comparison of readability across specialties and healthcare trusts. Methods: Providing complex scientific information in a way that is comprehensible to a lay person is a challenge that nowadays can be addressed by resorting to advanced Natural Language Processing (NLP) techniques, which make it possible to monitor the linguistic complexity of texts at the syntactic and lexical levels and to support their simplification, whenever needed. The study has been carried out by combining NLP-enabled feature extraction and state-of-the-art machine learning algorithms. To this end we used READ-IT, the first NLP-based readability assessment tool for Italian. Results: We analysed 584 documents, covering 29 specialties, for a total of 607.790 word tokens, currently used at the 36 public hospitals in Tuscany. Although the readability level of all documents in the corpus is low, both at the lexical and syntactic level, significant differences can be observed between specialties and healthcare trust releasing the forms. With the readability level ranging between 0 (easy-to-read) and 100 (difficult-to-read), it resulted that the pediatric informed consent documents are the most easy-to-read forms (with an average score of 75) while the most difficult-to read documents are documents of the surgical area (whose average score is 80) (standard deviation 2). Discussion: The state of the art resulting from this preliminary study shows that NLP-based readability assessment tools can help to measure the linguistic complexity of informed consent forms and guide the editor to identify linguistically complex passages that need to be simplified, either syntactically or lexically. The use of an assessment tool designed for the general language is the main limitation of the study and should be addressed through the customization of the tool to assess the readability of the healthcare jargon. A further step of the research consider also the design of a guidance to prepare readable informed consent forms.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
