Search

Motivations: It has recently been argued [1] that the effectiveness of a cure depends on the doctor-patient shared understanding of an illness and its treatment. Although a better communication between doctor and patient can be pursued through dedicated training programs, or by collecting patients' experiences and symptoms by means of questionnaires, the impact of these actions is limited by time and resources. In this paper we suggest that a patient-centered view of a disease - as well as potential misalignment between patient and doctor focuses - can be inferred at a larger scale through automated textual analysis of health-related forums. People are generating an enormous amount of social data to describe their health care experiences, and continuously search information about diseases, symptoms, diagnoses, doctors, treatment options and medicines. By automatically collecting, analyzing and exploiting this information, it is possible to obtain a more detailed and nuanced vision of patients' experience, that we call the "social phenotype" of diseases.Materials and Methods: As a use-case for our analysis, we consider diabetes, a widespread disease in most industrialized countries. We create a high quality data sample of diabetic patients' messages in Italy, extracted from popular medical forums during more than 10 years. Next, we use a state-of-the-art topic extraction technique based on generative statistical models improved with word embeddings, to identify the main complications, the frequently reported symptoms and the common concerns of these patients. Finally, in order to detect differences in focus, we compare the results of our analysis with available quality of life (QoL) assessments obtained with standard methodologies, such as questionnaires and survey studies.Results: We show that patients with diabetes, when accessing on-line forums, express a perception of their disease in a way that might be noticeably different from what is inferred from published QoL assessments on diabetes. In our study, we found that issues reported to have a daily impact on these patients are diet, glycemic control, drugs and clinical tests. These problems are not commonly considered in QoL assessments, since they are not perceived by doctors as representing severe limitations. Although limited to the case of Italian diabetic patients, we suggest that the methodology described in this paper, which is language and disease agnostic, could be applied to other diseases and countries, since misalignment between doctor and patients, and the importance of collecting unbiased patient perceptions, has been emphasized in many studies ([2,3]inter alia). Extracting the social phenotype of a disease might help acquiring patient-centered information on health care experiences on a much wider scale. [ABSTRACT FROM AUTHOR]

Background: Radiology reports are typically written in a free-text format, making clinical information difficult to extract and use. Recently, the adoption of structured reporting (SR) has been recommended by various medical societies thanks to the advantages it offers, e.g. standardization, completeness, and information retrieval. We propose a pipeline to extract information from Italian free-text radiology reports that fits with the items of the reference SR registry proposed by a national society of interventional and medical radiology, focusing on CT staging of patients with lymphoma., Methods: Our work aims to leverage the potential of Natural Language Processing and Transformer-based models to deal with automatic SR registry filling. With the availability of 174 Italian radiology reports, we investigate a rule-free generative Question Answering approach based on the Italian-specific version of T5: IT5. To address information content discrepancies, we focus on the six most frequently filled items in the annotations made on the reports: three categorical (multichoice), one free-text (free-text), and two continuous numerical (factual). In the preprocessing phase, we encode also information that is not supposed to be entered. Two strategies (batch-truncation and ex-post combination) are implemented to comply with the IT5 context length limitations. Performance is evaluated in terms of strict accuracy, f1, and format accuracy, and compared with the widely used GPT-3.5 Large Language Model. Unlike multichoice and factual, free-text answers do not have 1-to-1 correspondence with their reference annotations. For this reason, we collect human-expert feedback on the similarity between medical annotations and generated free-text answers, using a 5-point Likert scale questionnaire (evaluating the criteria of correctness and completeness)., Results: The combination of fine-tuning and batch splitting allows IT5 ex-post combination to achieve notable results in terms of information extraction of different types of structured data, performing on par with GPT-3.5. Human-based assessment scores of free-text answers show a high correlation with the AI performance metrics f1 (Spearman's correlation coefficients>0.5, p-values<0.001) for both IT5 ex-post combination and GPT-3.5. The latter is better at generating plausible human-like statements, even if it systematically provides answers even when they are not supposed to be given., Conclusions: In our experimental setting, a fine-tuned Transformer-based model with a modest number of parameters (i.e., IT5, 220 M) performs well as a clinical information extraction system for automatic SR registry filling task. It can extract information from more than one place in the report, elaborating it in a manner that complies with the response specifications provided by the SR registry (for multichoice and factual items), or that closely approximates the work of a human-expert (free-text items); with the ability to discern when an answer is supposed to be given or not to a user query., Competing Interests: Declaration of competing interest The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper., (Copyright © 2024 Elsevier B.V. All rights reserved.)