ABSTRACT
BACKGROUND: There are approximately 8,000 different rare diseases that affect roughly 400 million people worldwide. Many of them suffer from delayed diagnosis. Ciliopathies are rare monogenic disorders characterized by a significant phenotypic and genetic heterogeneity that raises an important challenge for clinical diagnosis. Diagnosis support systems (DSS) applied to electronic health record (EHR) data may help identify undiagnosed patients, which is of paramount importance to improve patients' care. Our objective was to evaluate three online-accessible rare disease DSSs using phenotypes derived from EHRs for the diagnosis of ciliopathies. METHODS: Two datasets of ciliopathy cases, either proven or suspected, and two datasets of controls were used to evaluate the DSSs. Patient phenotypes were automatically extracted from their EHRs and converted to Human Phenotype Ontology terms. We tested the ability of the DSSs to diagnose cases in contrast to controls based on Orphanet ontology. RESULTS: A total of 79 cases and 38 controls were selected. Performances of the DSSs on ciliopathy real world data (best DSS with area under the ROC curve = 0.72) were not as good as published performances on the test set used in the DSS development phase. None of these systems obtained results which could be described as "expert-level". Patients with multisystemic symptoms were generally easier to diagnose than patients with isolated symptoms. Diseases easily confused with ciliopathy generally affected multiple organs and had overlapping phenotypes. Four challenges need to be considered to improve the performances: to make the DSSs interoperable with EHR systems, to validate the performances in real-life settings, to deal with data quality, and to leverage methods and resources for rare and complex diseases. CONCLUSION: Our study provides insights into the complexities of diagnosing highly heterogenous rare diseases and offers lessons derived from evaluation existing DSSs in real-world settings. These insights are not only beneficial for ciliopathy diagnosis but also hold relevance for the enhancement of DSS for various complex rare disorders, by guiding the development of more clinically relevant rare disease DSSs, that could support early diagnosis and finally make more patients eligible for treatment.
Subject(s)
Ciliopathies , Electronic Health Records , Rare Diseases , Humans , Ciliopathies/diagnosis , Rare Diseases/diagnosis , Decision Support Systems, Clinical , PhenotypeABSTRACT
BACKGROUND: Gastrointestinal (GI) discomfort is prevalent and known to be associated with impaired quality of life. Real-world information on factors of GI discomfort and solutions used by people is, however, limited. Social media, including online forums, have been considered a new source of information to examine the health of populations in real-life settings. OBJECTIVE: The aims of this retrospective infodemiology study are to identify discussion topics, characterize users, and identify perceived determinants of GI discomfort in web-based messages posted by users of French social media. METHODS: Messages related to GI discomfort posted between January 2003 and August 2018 were extracted from 14 French-speaking general and specialized publicly available online forums. Extracted messages were cleaned and deidentified. Relevant medical concepts were determined on the basis of the Medical Dictionary for Regulatory Activities and vernacular terms. The identification of discussion topics was carried out by using a correlated topic model on the basis of the latent Dirichlet allocation. A nonsupervised clustering algorithm was applied to cluster forum users according to the reported symptoms of GI discomfort, discussion topics, and activity on online forums. Users' age and gender were determined by linear regression and application of a support vector machine, respectively, to characterize the identified clusters according to demographic parameters. Perceived factors of GI discomfort were classified by a combined method on the basis of syntactic analysis to identify messages with causality terms and a second topic modeling in a relevant segment of phrases. RESULTS: A total of 198,866 messages associated with GI discomfort were included in the analysis corpus after extraction and cleaning. These messages were posted by 36,989 separate web users, most of them being women younger than 40 years. Everyday life, diet, digestion, abdominal pain, impact on the quality of life, and tips to manage stress were among the most discussed topics. Segmentation of users identified 5 clusters corresponding to chronic and acute GI concerns. Diet topic was associated with each cluster, and stress was strongly associated with abdominal pain. Psychological factors, food, and allergens were perceived as the main causes of GI discomfort by web users. CONCLUSIONS: GI discomfort is actively discussed by web users. This study reveals a complex relationship between food, stress, and GI discomfort. Our approach has shown that identifying web-based discussion topics associated with GI discomfort and its perceived factors is feasible and can serve as a complementary source of real-world evidence for caregivers.
Subject(s)
Gastrointestinal Diseases/therapy , Quality of Life/psychology , Telemedicine/methods , Adult , Female , Humans , Internet , Language , Male , Middle Aged , Retrospective Studies , Social Media , Time Factors , Young AdultABSTRACT
BACKGROUND: Immune checkpoint inhibitors (ICIs) are increasingly used to treat several types of tumors. Impact of this emerging therapy on patients' health-related quality of life (HRQoL) is usually collected in clinical trials through standard questionnaires. However, this might not fully reflect HRQoL of patients under real-world conditions. In parallel, users' narratives from social media represent a potential new source of research concerning HRQoL. OBJECTIVE: The aim of this study is to assess and compare coverage of ICI-treated patients' HRQoL domains and subdomains in standard questionnaires from clinical trials and in real-world setting from social media posts. METHODS: A retrospective study was carried out by collecting social media posts in French language written by internet users mentioning their experiences with ICIs between January 2011 and August 2018. Automatic and manual extractions were implemented to create a corpus where domains and subdomains of HRQoL were classified. These annotations were compared with domains covered by 2 standard HRQoL questionnaires, the EORTC QLQ-C30 and the FACT-G. RESULTS: We identified 150 users who described their own experience with ICI (89/150, 59.3%) or that of their relative (61/150, 40.7%), with 137 users (91.3%) reporting at least one HRQoL domain in their social media posts. A total of 8 domains and 42 subdomains of HRQoL were identified: Global health (1 subdomain; 115 patients), Symptoms (13; 76), Emotional state (10; 49), Role (7; 22), Physical activity (4; 13), Professional situation (3; 9), Cognitive state (2; 2), and Social state (2; 2). The QLQ-C30 showed a wider global coverage of social media HRQoL subdomains than the FACT-G, 45% (19/42) and 29% (12/42), respectively. For both QLQ-C30 and FACT-G questionnaires, coverage rates were particularly suboptimal for Symptoms (68/123, 55.3% and 72/123, 58.5%, respectively), Emotional state (7/49, 14% and 24/49, 49%, respectively), and Role (17/22, 77% and 15/22, 68%, respectively). CONCLUSIONS: Many patients with cancer are using social media to share their experiences with immunotherapy. Collecting and analyzing their spontaneous narratives are helpful to capture and understand their HRQoL in real-world setting. New measures of HRQoL are needed to provide more in-depth evaluation of Symptoms, Emotional state, and Role among patients with cancer treated with immunotherapy.
Subject(s)
Immune Checkpoint Inhibitors/therapeutic use , Quality of Life/psychology , Social Media/standards , Data Analysis , Female , Humans , Immune Checkpoint Inhibitors/pharmacology , Male , Retrospective Studies , Surveys and QuestionnairesABSTRACT
BACKGROUND: Medication nonadherence is a major impediment to the management of many health conditions. A better understanding of the factors underlying noncompliance to treatment may help health professionals to address it. Patients use peer-to-peer virtual communities and social media to share their experiences regarding their treatments and diseases. Using topic models makes it possible to model themes present in a collection of posts, thus to identify cases of noncompliance. OBJECTIVE: The aim of this study was to detect messages describing patients' noncompliant behaviors associated with a drug of interest. Thus, the objective was the clustering of posts featuring a homogeneous vocabulary related to nonadherent attitudes. METHODS: We focused on escitalopram and aripiprazole used to treat depression and psychotic conditions, respectively. We implemented a probabilistic topic model to identify the topics that occurred in a corpus of messages mentioning these drugs, posted from 2004 to 2013 on three of the most popular French forums. Data were collected using a Web crawler designed by Kappa Santé as part of the Detec't project to analyze social media for drug safety. Several topics were related to noncompliance to treatment. RESULTS: Starting from a corpus of 3650 posts related to an antidepressant drug (escitalopram) and 2164 posts related to an antipsychotic drug (aripiprazole), the use of latent Dirichlet allocation allowed us to model several themes, including interruptions of treatment and changes in dosage. The topic model approach detected cases of noncompliance behaviors with a recall of 98.5% (272/276) and a precision of 32.6% (272/844). CONCLUSIONS: Topic models enabled us to explore patients' discussions on community websites and to identify posts related with noncompliant behaviors. After a manual review of the messages in the noncompliance topics, we found that noncompliance to treatment was present in 6.17% (276/4469) of the posts.
Subject(s)
Internet/instrumentation , Medication Adherence/statistics & numerical data , Social Media/instrumentation , HumansABSTRACT
BACKGROUND: While traditional signal detection methods in pharmacovigilance are based on spontaneous reports, the use of social media is emerging. The potential strength of Web-based data relies on their volume and real-time availability, allowing early detection of signals of disproportionate reporting (SDRs). OBJECTIVE: This study aimed (1) to assess the consistency of SDRs detected from patients' medical forums in France compared with those detected from the traditional reporting systems and (2) to assess the ability of SDRs in identifying earlier than the traditional reporting systems. METHODS: Messages posted on patients' forums between 2005 and 2015 were used. We retained 8 disproportionality definitions. Comparison of SDRs from the forums with SDRs detected in VigiBase was done by describing the sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), accuracy, receiver operating characteristics curve, and the area under the curve (AUC). The time difference in months between the detection dates of SDRs from the forums and VigiBase was provided. RESULTS: The comparison analysis showed that the sensitivity ranged from 29% to 50.6%, the specificity from 86.1% to 95.5%, the PPV from 51.2% to 75.4%, the NPV from 68.5% to 91.6%, and the accuracy from 68% to 87.7%. The AUC reached 0.85 when using the metric empirical Bayes geometric mean. Up to 38% (12/32) of the SDRs were detected earlier in the forums than that in VigiBase. CONCLUSIONS: The specificity, PPV, and NPV were high. The overall performance was good, showing that data from medical forums may be a valuable source for signal detection. In total, up to 38% (12/32) of the SDRs could have been detected earlier, thus, ensuring the increased safety of patients. Further enhancements are needed to investigate the reliability and validation of patients' medical forums worldwide, the extension of this analysis to all possible drugs or at least to a wider selection of drugs, as well as to further assess performance against established signals.
Subject(s)
Databases, Factual , France , Humans , Internet , PharmacovigilanceABSTRACT
Rare diseases pose significant challenges due to their heterogeneity and lack of knowledge. This study develops a comprehensive pipeline interoperable with a document-oriented clinical data warehouse, integrating cohort characterization, patient clustering and interpretation. Leveraging NLP, semantic similarity, machine learning and visualization, the pipeline enables the identification of prevalent phenotype patterns and patient stratification. To enhance interpretability, discriminant phenotypes characterizing each cluster are provided. Users can visually test hypotheses by marking patients exhibiting specific keywords in the EHR like genes, drugs and procedures. Implemented through a web interface, the pipeline enables clinicians to navigate through different modules, discover intricate patterns and generate interpretable insights that may advance rare diseases understanding, guide decision-making, and ultimately improve patient outcomes.
Subject(s)
Electronic Health Records , Phenotype , Rare Diseases , Humans , Machine Learning , Data Warehousing , Natural Language Processing , Cluster Analysis , User-Computer InterfaceABSTRACT
BACKGROUND: Rare diseases affect approximately 400 million people worldwide. Many of them suffer from delayed diagnosis. Among them, NPHP1-related renal ciliopathies need to be diagnosed as early as possible as potential treatments have been recently investigated with promising results. Our objective was to develop a supervised machine learning pipeline for the detection of NPHP1 ciliopathy patients from a large number of nephrology patients using electronic health records (EHRs). METHODS AND RESULTS: We designed a pipeline combining a phenotyping module re-using unstructured EHR data, a semantic similarity module to address the phenotype dependence, a feature selection step to deal with high dimensionality, an undersampling step to address the class imbalance, and a classification step with multiple train-test split for the small number of rare cases. The pipeline was applied to thirty NPHP1 patients and 7231 controls and achieved good performances (sensitivity 86% with specificity 90%). A qualitative review of the EHRs of 40 misclassified controls showed that 25% had phenotypes belonging to the ciliopathy spectrum, which demonstrates the ability of our system to detect patients with similar conditions. CONCLUSIONS: Our pipeline reached very encouraging performance scores for pre-diagnosing ciliopathy patients. The identified patients could then undergo genetic testing. The same data-driven approach can be adapted to other rare diseases facing underdiagnosis challenges.
Subject(s)
Ciliopathies , Rare Diseases , Humans , Electronic Health Records , Semantics , Supervised Machine Learning , Ciliopathies/diagnosis , Ciliopathies/genetics , AlgorithmsABSTRACT
In the context of medical concept extraction, it is critical to determine if clinical signs or symptoms mentioned in the text were present or absent, experienced by the patient or their relatives. Previous studies have focused on the NLP aspect but not on how to leverage this supplemental information for clinical applications. In this paper, we aim to use the patient similarity networks framework to aggregate different phenotyping modalities. NLP techniques were applied to extract phenotypes and predict their modalities from 5470 narrative reports of 148 patients with ciliopathies (a group of rare diseases). Patient similarities were computed using each modality separately for aggregation and clustering. We found that aggregating negated phenotypes improved patient similarity, but further aggregating relatives' phenotypes worsened the result. We suggest that different modalities of phenotypes can contribute to patient similarity, but they should be aggregated carefully and with appropriate similarity metrics and aggregation models.
Subject(s)
Electronic Health Records , Narration , Humans , Phenotype , Rare Diseases , Natural Language ProcessingABSTRACT
BACKGROUND: During the unprecedented COVID-19 pandemic, social media has been extensively used to amplify the spread of information and to express personal health-related experiences regarding symptoms, including anosmia and ageusia, 2 symptoms that have been reported later than other symptoms. OBJECTIVE: Our objective is to investigate to what extent Twitter users reported anosmia and ageusia symptoms in their tweets and if they connected them to COVID-19, to evaluate whether these symptoms could have been identified as COVID-19 symptoms earlier using Twitter rather than the official notice. METHODS: We collected French tweets posted between January 1, 2020, and March 31, 2020, containing anosmia- or ageusia-related keywords. Symptoms were detected using fuzzy matching. The analysis consisted of 3 parts. First, we compared the coverage of anosmia and ageusia symptoms in Twitter and in traditional media to determine if the association between COVID-19 and anosmia or ageusia could have been identified earlier through Twitter. Second, we conducted a manual analysis of anosmia- and ageusia-related tweets to obtain quantitative and qualitative insights regarding their nature and to assess when the first associations between COVID-19 and these symptoms were established. We randomly annotated tweets from 2 periods: the early stage and the rapid spread stage of the epidemic. For each tweet, each symptom was annotated regarding 3 modalities: symptom (yes or no), associated with COVID-19 (yes, no, or unknown), and whether it was experienced by someone (yes, no, or unknown). Third, to evaluate if there was a global increase of tweets mentioning anosmia or ageusia in early 2020, corresponding to the beginning of the COVID-19 epidemic, we compared the tweets reporting experienced anosmia or ageusia between the first periods of 2019 and 2020. RESULTS: In total, 832 (respectively 12,544) tweets containing anosmia (respectively ageusia) related keywords were extracted over the analysis period in 2020. The comparison to traditional media showed a strong correlation without any lag, which suggests an important reactivity of Twitter but no earlier detection on Twitter. The annotation of tweets from 2020 showed that tweets correlating anosmia or ageusia with COVID-19 could be found a few days before the official announcement. However, no association could be found during the first stage of the pandemic. Information about the temporality of symptoms and the psychological impact of these symptoms could be found in the tweets. The comparison between early 2020 and early 2019 showed no difference regarding the volumes of tweets. CONCLUSIONS: Based on our analysis of French tweets, associations between COVID-19 and anosmia or ageusia by web users could have been found on Twitter just a few days before the official announcement but not during the early stage of the pandemic. Patients share qualitative information on Twitter regarding anosmia or ageusia symptoms that could be of interest for future analyses.
Subject(s)
Ageusia , COVID-19 , Social Media , Humans , Retrospective Studies , Ageusia/diagnosis , Anosmia/epidemiology , Pandemics , COVID-19/diagnosisABSTRACT
The wide adoption of Electronic Health Records (EHR) in hospitals provides unique opportunities for high throughput phenotyping of patients. The phenotype extraction from narrative reports can be performed by using either dictionary-based or data-driven methods. We developed a hybrid pipeline using deep learning to enrich the UMLS Metathesaurus for automatic detection of phenotypes from EHRs. The pipeline was evaluated on a French database of patients with a rare disease characterized by skeletal abnormalities, Jeune syndrome. The results showed a 2.5-fold improvement regarding the number of detected skeletal abnormalities compared to the baseline extraction using the standard release of UMLS. Our method can help enrich the coverage of the UMLS and improve phenotyping, especially for languages other than English.
Subject(s)
Deep Learning , Unified Medical Language System , Algorithms , Electronic Health Records , Ellis-Van Creveld Syndrome , Humans , Rare Diseases/diagnosisABSTRACT
A timely diagnosis is a key challenge for many rare diseases. As an expanding group of rare and severe monogenic disorders with a broad spectrum of clinical manifestations, ciliopathies, notably renal ciliopathies, suffer from important underdiagnosis issues. Our objective is to develop an approach for screening large-scale clinical data warehouses and detecting patients with similar clinical manifestations to those from diagnosed ciliopathy patients. We expect that the top-ranked similar patients will benefit from genetic testing for an early diagnosis. The dependence and relatedness between phenotypes were taken into account in our similarity model through medical concept embedding. The relevance of each phenotype to each patient was also considered by adjusted aggregation of phenotype similarity into patient similarity. A ranking model based on the best-subtype-average similarity was proposed to address the phenotypic overlapping and heterogeneity of ciliopathies. Our results showed that using less than one-tenth of learning sources, our language and center specific embedding provided comparable or better performances than other existing medical concept embeddings. Combined with the best-subtype-average ranking model, our patient-patient similarity-based screening approach was demonstrated effective in two large scale unbalanced datasets containing approximately 10,000 and 60,000 controls with kidney manifestations in the clinical data warehouse (about 2 and 0.4% of prevalence, respectively). Our approach will offer the opportunity to identify candidate patients who could go through genetic testing for ciliopathy. Earlier diagnosis, before irreversible end-stage kidney disease, will enable these patients to benefit from appropriate follow-up and novel treatments that could alleviate kidney dysfunction.
ABSTRACT
To identify patients with similar clinical profiles and derive insights from the records and outcomes of similar patients can help fast and precise diagnosis and other clinical decisions for rare diseases. Similarity methods are required to take into account the semantic relations between medical concepts and also the different relevance of all medical concepts presented in patients' medical records. In this paper, we introduce the methods developed in the context of rare disease screening/diagnosis from clinical data warehouse using medical concept embedding and adjusted aggregations. Our methods provided better preliminary results than baseline methods, with a significant improvement of precision among the top ranked similar patients, which is encouraging for further fine-tuning and application on a large-scale dataset for new/candidate patient identification.
Subject(s)
Electronic Health Records , Rare Diseases , Data Warehousing , Feasibility Studies , Humans , Rare Diseases/diagnosis , SemanticsABSTRACT
BACKGROUND: During the COVID-19 pandemic, numerous countries, including China and France, have implemented lockdown measures that have been effective in controlling the epidemic. However, little is known about the impact of these measures on the population as expressed on social media from different cultural contexts. OBJECTIVE: This study aims to assess and compare the evolution of the topics discussed on Chinese and French social media during the COVID-19 lockdown. METHODS: We extracted posts containing COVID-19-related or lockdown-related keywords in the most commonly used microblogging social media platforms (ie, Weibo in China and Twitter in France) from 1 week before lockdown to the lifting of the lockdown. A topic model was applied independently for three periods (prelockdown, early lockdown, and mid to late lockdown) to assess the evolution of the topics discussed on Chinese and French social media. RESULTS: A total of 6395; 23,422; and 141,643 Chinese Weibo messages, and 34,327; 119,919; and 282,965 French tweets were extracted in the prelockdown, early lockdown, and mid to late lockdown periods, respectively, in China and France. Four categories of topics were discussed in a continuously evolving way in all three periods: epidemic news and everyday life, scientific information, public measures, and solidarity and encouragement. The most represented category over all periods in both countries was epidemic news and everyday life. Scientific information was far more discussed on Weibo than in French tweets. Misinformation circulated through social media in both countries; however, it was more concerned with the virus and epidemic in China, whereas it was more concerned with the lockdown measures in France. Regarding public measures, more criticisms were identified in French tweets than on Weibo. Advantages and data privacy concerns regarding tracing apps were also addressed in French tweets. All these differences were explained by the different uses of social media, the different timelines of the epidemic, and the different cultural contexts in these two countries. CONCLUSIONS: This study is the first to compare the social media content in eastern and western countries during the unprecedented COVID-19 lockdown. Using general COVID-19-related social media data, our results describe common and different public reactions, behaviors, and concerns in China and France, even covering the topics identified in prior studies focusing on specific interests. We believe our study can help characterize country-specific public needs and appropriately address them during an outbreak.
ABSTRACT
The exhaustive automatic detection of symptoms in social media posts is made difficult by the presence of colloquial expressions, misspellings and inflected forms of words. The detection of self-reported symptoms is of major importance for emergent diseases like the Covid-19. In this study, we aimed to (1) develop an algorithm based on fuzzy matching to detect symptoms in tweets, (2) establish a comprehensive list of Covid-19-related symptoms and (3) evaluate the fuzzy matching for Covid-19-related symptom detection in French tweets. The Covid-19-related symptom list was built based on the aggregation of different data sources. French Covid-19-related tweets were automatically extracted using a dedicated data broker during the first wave of the pandemic in France. The fuzzy matching parameters were finetuned using all symptoms from MedDRA and then evaluated on a subset of 5000 Covid-19-related tweets in French for the detection of symptoms from our Covid-19-related list. The fuzzy matching improved the detection by the addition of 42% more correct matches with an 81% precision.
Subject(s)
COVID-19 , Social Media , France/epidemiology , Humans , Pandemics , SARS-CoV-2ABSTRACT
INTRODUCTION: Rare diseases affect approximately 350 million people worldwide. Delayed diagnosis is frequent due to lack of knowledge of most clinicians and a small number of expert centers. Consequently, computerized diagnosis support systems have been developed to address these issues, with many relying on rare disease expertise and taking advantage of the increasing volume of generated and accessible health-related data. Our objective is to perform a review of all initiatives aiming to support the diagnosis of rare diseases. METHODS: A scoping review was conducted based on methods proposed by Arksey and O'Malley. A charting form for relevant study analysis was developed and used to categorize data. RESULTS: Sixty-eight studies were retained at the end of the charting process. Diagnosis targets varied from 1 rare disease to all rare diseases. Material used for diagnosis support consisted mostly of phenotype concepts, images or fluids. Fifty-seven percent of the studies used expert knowledge. Two-thirds of the studies relied on machine learning algorithms, and one-third used simple similarities. Manual algorithms were encountered as well. Most of the studies presented satisfying performance of evaluation by comparison with references or with external validation. Fourteen studies provided online tools, most of which aimed to support the diagnosis of all rare diseases by considering queries based on phenotype concepts. CONCLUSION: Numerous solutions relying on different materials and use of various methodologies are emerging with satisfying preliminary results. However, the variability of approaches and evaluation processes complicates the comparison of results. Efforts should be made to adequately validate these tools and guarantee reproducibility and explicability.
Subject(s)
Rare Diseases , Humans , Rare Diseases/diagnosis , Reproducibility of ResultsABSTRACT
BACKGROUND: Social media is a potential source of information on postmarketing drug safety surveillance that still remains unexploited nowadays. Information technology solutions aiming at extracting adverse reactions (ADRs) from posts on health forums require a rigorous evaluation methodology if their results are to be used to make decisions. First, a gold standard, consisting of manual annotations of the ADR by human experts from the corpus extracted from social media, must be implemented and its quality must be assessed. Second, as for clinical research protocols, the sample size must rely on statistical arguments. Finally, the extraction methods must target the relation between the drug and the disease (which might be either treated or caused by the drug) rather than simple co-occurrences in the posts. OBJECTIVE: We propose a standardized protocol for the evaluation of a software extracting ADRs from the messages on health forums. The study is conducted as part of the Adverse Drug Reactions from Patient Reports in Social Media project. METHODS: Messages from French health forums were extracted. Entity recognition was based on Racine Pharma lexicon for drugs and Medical Dictionary for Regulatory Activities terminology for potential adverse events (AEs). Natural language processing-based techniques automated the ADR information extraction (relation between the drug and AE entities). The corpus of evaluation was a random sample of the messages containing drugs and/or AE concepts corresponding to recent pharmacovigilance alerts. A total of 2 persons experienced in medical terminology manually annotated the corpus, thus creating the gold standard, according to an annotator guideline. We will evaluate our tool against the gold standard with recall, precision, and f-measure. Interannotator agreement, reflecting gold standard quality, will be evaluated with hierarchical kappa. Granularities in the terminologies will be further explored. RESULTS: Necessary and sufficient sample size was calculated to ensure statistical confidence in the assessed results. As we expected a global recall of 0.5, we needed at least 384 identified ADR concepts to obtain a 95% CI with a total width of 0.10 around 0.5. The automated ADR information extraction in the corpus for evaluation is already finished. The 2 annotators already completed the annotation process. The analysis of the performance of the ADR information extraction module as compared with gold standard is ongoing. CONCLUSIONS: This protocol is based on the standardized statistical methods from clinical research to create the corpus, thus ensuring the necessary statistical power of the assessed results. Such evaluation methodology is required to make the ADR information extraction software useful for postmarketing drug safety surveillance. INTERNATIONAL REGISTERED REPORT IDENTIFIER (IRRID): RR1-10.2196/11448.
ABSTRACT
Background: The Food and Drug Administration (FDA) in the United States and the European Medicines Agency (EMA) have recognized social media as a new data source to strengthen their activities regarding drug safety. Objective: Our objective in the ADR-PRISM project was to provide text mining and visualization tools to explore a corpus of posts extracted from social media. We evaluated this approach on a corpus of 21 million posts from five patient forums, and conducted a qualitative analysis of the data available on methylphenidate in this corpus. Methods: We applied text mining methods based on named entity recognition and relation extraction in the corpus, followed by signal detection using proportional reporting ratio (PRR). We also used topic modeling based on the Correlated Topic Model to obtain the list of the matics in the corpus and classify the messages based on their topics. Results: We automatically identified 3443 posts about methylphenidate published between 2007 and 2016, among which 61 adverse drug reactions (ADR) were automatically detected. Two pharmacovigilance experts evaluated manually the quality of automatic identification, and a f-measure of 0.57 was reached. Patient's reports were mainly neuro-psychiatric effects. Applying PRR, 67% of the ADRs were signals, including most of the neuro-psychiatric symptoms but also palpitations. Topic modeling showed that the most represented topics were related to Childhood and Treatment initiation, but also Side effects. Cases of misuse were also identified in this corpus, including recreational use and abuse. Conclusion: Named entity recognition combined with signal detection and topic modeling have demonstrated their complementarity in mining social media data. An in-depth analysis focused on methylphenidate showed that this approach was able to detect potential signals and to provide better understanding of patients' behaviors regarding drugs, including misuse.
ABSTRACT
BACKGROUND: Adverse drug reactions (ADRs) are an important cause of morbidity and mortality. Classical Pharmacovigilance process is limited by underreporting which justifies the current interest in new knowledge sources such as social media. The Adverse Drug Reactions from Patient Reports in Social Media (ADR-PRISM) project aims to extract ADRs reported by patients in these media. We identified 5 major challenges to overcome to operationalize the analysis of patient posts: (1) variable quality of information on social media, (2) guarantee of data privacy, (3) response to pharmacovigilance expert expectations, (4) identification of relevant information within Web pages, and (5) robust and evolutive architecture. OBJECTIVE: This article aims to describe the current state of advancement of the ADR-PRISM project by focusing on the solutions we have chosen to address these 5 major challenges. METHODS: In this article, we propose methods and describe the advancement of this project on several aspects: (1) a quality driven approach for selecting relevant social media for the extraction of knowledge on potential ADRs, (2) an assessment of ethical issues and French regulation for the analysis of data on social media, (3) an analysis of pharmacovigilance expert requirements when reviewing patient posts on the Internet, (4) an extraction method based on natural language processing, pattern based matching, and selection of relevant medical concepts in reference terminologies, and (5) specifications of a component-based architecture for the monitoring system. RESULTS: Considering the 5 major challenges, we (1) selected a set of 21 validated criteria for selecting social media to support the extraction of potential ADRs, (2) proposed solutions to guarantee data privacy of patients posting on Internet, (3) took into account pharmacovigilance expert requirements with use case diagrams and scenarios, (4) built domain-specific knowledge resources embeding a lexicon, morphological rules, context rules, semantic rules, syntactic rules, and post-analysis processing, and (5) proposed a component-based architecture that allows storage of big data and accessibility to third-party applications through Web services. CONCLUSIONS: We demonstrated the feasibility of implementing a component-based architecture that allows collection of patient posts on the Internet, near real-time processing of those posts including annotation, and storage in big data structures. In the next steps, we will evaluate the posts identified by the system in social media to clarify the interest and relevance of such approach to improve conventional pharmacovigilance processes based on spontaneous reporting.
ABSTRACT
Suspected adverse drug reactions (ADR) reported by patients through social media can be a complementary source to current pharmacovigilance systems. However, the performance of text mining tools applied to social media text data to discover ADRs needs to be evaluated. In this paper, we introduce the approach developed to mine ADR from French social media. A protocol of evaluation is highlighted, which includes a detailed sample size determination and evaluation corpus constitution. Our text mining approach provided very encouraging preliminary results with F-measures of 0.94 and 0.81 for recognition of drugs and symptoms respectively, and with F-measure of 0.70 for ADR detection. Therefore, this approach is promising for downstream pharmacovigilance analysis.
Subject(s)
Data Mining , Drug-Related Side Effects and Adverse Reactions , Semantics , Social Media , Adverse Drug Reaction Reporting Systems , Humans , PharmacovigilanceABSTRACT
BACKGROUND AND OBJECTIVES: Suspected adverse drug reactions (ADR) reported by patients through social media can be a complementary tool to already existing ADRs signal detection processes. However, several studies have shown that the quality of medical information published online varies drastically whatever the health topic addressed. The aim of this study is to use an existing rating tool on a set of social network web sites in order to assess the capabilities of these tools to guide experts for selecting the most adapted social network web site to mine ADRs. METHODS: First, we reviewed and rated 132 Internet forums and social networks according to three major criteria: the number of visits, the notoriety of the forum and the number of messages posted in relation with health and drug therapy. Second, the pharmacist reviewed the topic-oriented message boards with a small number of drug names to ensure that they were not off topic. Six experts have been chosen to assess the selected internet forums using a French scoring tool: Net scoring. Three different scores and the agreement between experts according to each set of scores using weighted kappa pooled using mean have been computed. RESULTS: Three internet forums were chosen at the end of the selection step. Some criteria get high score (scores 3-4) no matter the website evaluated like accessibility (45-46) or design (34-36), at the opposite some criteria always have bad scores like quantitative (40-42) and ethical aspect (43-44), hyperlinks actualization (30-33). Kappa were positives but very small which corresponds to a weak agreement between experts. CONCLUSION: The personal opinion of the expert seems to have a major impact, undermining the relevance of the criterion. Our future work is to collect results given by this evaluation grid and proposes a new scoring tool for Internet social networks assessment.