Natural language processing of symptoms documented in free-text narratives of electronic health records: a systematic review

Abstract

Objective Natural language processing (NLP) of symptoms from electronic health records (EHRs) could contribute to the advancement of symptom science. We aim to synthesize the literature on the use of NLP to process or analyze symptom information documented in EHR free-text narratives. Materials and Methods Our search of 1964 records from PubMed and EMBASE was narrowed to 27 eligible articles. Data related to the purpose, free-text corpus, patients, symptoms, NLP methodology, evaluation metrics, and quality indicators were extracted for each study. Results Symptom-related information was presented as a primary outcome in 14 studies. EHR narratives represented various inpatient and outpatient clinical specialties, with general, cardiology, and mental health occurring most frequently. Studies encompassed a wide variety of symptoms, including shortness of breath, pain, nausea, dizziness, disturbed sleep, constipation, and depressed mood. NLP approaches included previously developed NLP tools, classification methods, and manually curated rule-based processing. Only one-third (n = 9) of studies reported patient demographic characteristics. Discussion NLP is used to extract information from EHR free-text narratives written by a variety of healthcare providers on an expansive range of symptoms across diverse clinical specialties. The current focus of this field is on the development of methods to extract symptom information and the use of symptom information for disease classification tasks rather than the examination of symptoms themselves. Conclusion Future NLP studies should concentrate on the investigation of symptoms and symptom documentation in EHR free-text narratives. Efforts should be undertaken to examine patient characteristics and make symptom-related NLP algorithms or pipelines and vocabularies openly available.

Medical Information Extraction in the Age of Deep Learning

“…For instance, Sheikhalishahi et al [40] discussed NLP methods targeted at chronic diseases and found that shallow ML and rule-based approaches (as opposed to more sophisticated DL-based ones) prevail. Koleck et al [41] summarized the use of NLP to analyze symptom information documented in EHR free-text narratives as an indication of diseases and similar to the previous survey found little coverage of DL methods in this application area as well. Savova et al [42] reviewed the current state of clinical NLP with respect to oncology and cancer phenotyping from EHR.…”

Advancing biomedical and health informatics knowledge through reviews of existing research

“…The methods varied in our 5 examples, ranging from author-developed spreadsheets 6 and specialized software such as Covidence from the Cochrane Collaboration (https://community.cochrane.org/help/tools-and-software/covidence). 8…”

“…13 In contrast, Blackley et al 10 characterized the quality of studies by their reporting of metrics related to speech recognition for clinical documents, including number of speakers, number of documents, and a variety of accuracy measures. Similarly, Koleck et al 8 reported on the number of documents and a set of metrics suitable for evaluating an natural language processing algorithm or pipeline including sensitivity, specificity, precision, recall, F measure, kappa statistic, area under the receiver-operating characteristic curve, and C-statistic. In 1 instance, 6 the actual purposes of the review were to discover what evaluation measures were used in the literature and to make recommendations related to quality assessments.…”

“…Synthesis approaches were primarily tabular along with narrative text. However, 3 reviews contained graphical components, including a histogram of studies over time, 10 a chord diagram showing the associations between symptoms and studies, 8 and a heatmap displaying the relationships among PICOS (participants, interventions, comparisons, outcomes, and study design) 7 components and setting and measurement classifications. 6 Now that JAMIA is an online journal, there are no charges for color, so I encourage authors to include graphs when appropriate.…”

Proposal of the First International Workshop on Semantic Indexing and Information Retrieval for Health from Heterogeneous Content Types and Languages (SIIRH)

“…Efficient retrieval of biomedical documents is key for evidence-based medicine, preparing systematic reviews or retrieval of particular clinical case studies. Due to particular search conditions of caregivers and healthcare professionals (limited amount of time spent per patient), they are also in need of more sophisticated retrieval approaches applied to electronic health records [11], a type of content highly challenging due to its telegraphic and domain specific language and the presence of negations and abbreviations. There is also interest in processing patient-generated content like social media and patient fora, a key resource for rare disease research, clinical trials patient selection/stratification or for discovering new patient-reported symptoms and treatment-related adverse effects.…”