Free-text Narratives Research Articles

Natural Language Processing in Radiology: A Systematic Review.

Radiological reporting has generated large quantities of digital content within the electronic health record, which is potentially a valuable source of information for improving clinical care and supporting research. Although radiology reports are stored for communication and documentation of diagnostic imaging, harnessing their potential requires efficient and automated information extraction: they exist mainly as free-text clinical narrative, from which it is a major challenge to obtain structured data. Natural language processing (NLP) provides techniques that aid the conversion of text into a structured representation, and thus enables computers to derive meaning from human (ie, natural language) input. Used on radiology reports, NLP techniques enable automatic identification and extraction of information. By exploring the various purposes for their use, this review examines how radiology benefits from NLP. A systematic literature search identified 67 relevant publications describing NLP methods that support practical applications in radiology. This review takes a close look at the individual studies in terms of tasks (ie, the extracted information), the NLP methodology and tools used, and their application purpose and performance results. Additionally, limitations, future challenges, and requirements for advancing NLP in radiology will be discussed.

Structured pathology reporting improves the macroscopic assessment of rectal tumour resection specimens

We examined whether introduction of a structured macroscopic reporting template for rectal tumour resection specimens improved the completeness and efficiency in collecting key macroscopic data elements. Fifty free text (narrative) macroscopic reports retrieved from 2012 to 2014 were compared with 50 structured macroscopic reports from 2013 to 2015, all of which were generated at John Hunter Hospital, Newcastle, NSW. The six standard macroscopic data elements examined in this study were reported in all 50 anatomical pathology reports using a structured macroscopic reporting dictation template. Free text reports demonstrated significantly impaired data collection when recording intactness of mesorectum (p<0.001), relationship to anterior peritoneal reflection (p=0.028) and distance of tumour to the non-peritonealised circumferential margin (p<0.001). The number of words used was also significantly (p<0.001) reduced using pre-formatted structured reports compared to free text reports. The introduction of a structured reporting dictation template improves data collection and may reduce the subsequent administrative burden when macroscopically evaluating rectal resections.

Identification of Nonresponse to Treatment Using Narrative Data in an Electronic Health Record Inflammatory Bowel Disease Cohort

Electronic health records, increasingly a part of healthcare, provide a wealth of untapped narrative free text data that have the potential to accurately inform clinical outcomes. From a validated cohort of patients with Crohn's disease or ulcerative colitis, we identified patients with ≥1 coded or narrative mention of monoclonal antibodies to tumor necrosis factor α. Chart review by ascertained true use of therapy, time of initiation, and cessation of treatment, and also clinical response stratified as nonresponse, partial, or complete response at 1 year. Internal consistency was assessed in an independent validation cohort. A total of 3087 patients had a mention of an antibodies to tumor necrosis factor α. Actual therapy initiation was within 60 days of the first coded mention in 74% of patients. In the derivation cohort, 18% of antibodies to tumor necrosis factor α starts were classified as nonresponse at 1 year, 21% as partial, and 56% as complete response. On multivariate analysis, the number of narrative mentions of diarrhea (odds ratio 1.08; 95% confidence interval, 1.02-1.14) and fatigue (odds ratio 1.16; 95% confidence interval, 1.02-1.32) was independently associated with nonresponse at 1 year (area under the curve 0.82). A likelihood of nonresponse score comprising a weighted sum of both demonstrated a good dose-response relationship across nonresponders (2.18), partial (1.20), and complete (0.50) responders (P < 0.0001) and correlated well with need for surgery or hospitalizations. Narrative data in an electronic health record offer considerable potential to define temporally evolving disease outcomes such as nonresponse to treatment.

Comparison of methods for auto-coding causation of injury narratives

Manually reading free-text narratives in large databases to identify the cause of an injury can be very time consuming and recently, there has been much work in automating this process. In particular, the variations of the naïve Bayes model have been used to successfully auto-code free text narratives describing the event/exposure leading to the injury of a workers’ compensation claim. This paper compares the naïve Bayes model with an alternative logistic model and found that this new model outperformed the naïve Bayesian model. Further modest improvements were found through the addition of sequences of keywords in the models as opposed to consideration of only single keywords. The programs and weights used in this paper are available upon request to researchers without a training set wishing to automatically assign event codes to large data-sets of text narratives. The utility of sharing this program was tested on an outside set of injury narratives provided by the Bureau of Labor Statistics with promising results.

Data-Driven Information Extraction from Chinese Electronic Medical Records.

ObjectiveThis study aims to propose a data-driven framework that takes unstructured free text narratives in Chinese Electronic Medical Records (EMRs) as input and converts them into structured time-event-description triples, where the description is either an elaboration or an outcome of the medical event.Materials and MethodsOur framework uses a hybrid approach. It consists of constructing cross-domain core medical lexica, an unsupervised, iterative algorithm to accrue more accurate terms into the lexica, rules to address Chinese writing conventions and temporal descriptors, and a Support Vector Machine (SVM) algorithm that innovatively utilizes Normalized Google Distance (NGD) to estimate the correlation between medical events and their descriptions.ResultsThe effectiveness of the framework was demonstrated with a dataset of 24,817 de-identified Chinese EMRs. The cross-domain medical lexica were capable of recognizing terms with an F1-score of 0.896. 98.5% of recorded medical events were linked to temporal descriptors. The NGD SVM description-event matching achieved an F1-score of 0.874. The end-to-end time-event-description extraction of our framework achieved an F1-score of 0.846.DiscussionIn terms of named entity recognition, the proposed framework outperforms state-of-the-art supervised learning algorithms (F1-score: 0.896 vs. 0.886). In event-description association, the NGD SVM is superior to SVM using only local context and semantic features (F1-score: 0.874 vs. 0.838).ConclusionsThe framework is data-driven, weakly supervised, and robust against the variations and noises that tend to occur in a large corpus. It addresses Chinese medical writing conventions and variations in writing styles through patterns used for discovering new terms and rules for updating the lexica.

Estimation of Agricultural and Logging Injury Incidence in Maine Using Electronic Administrative Data Sets

ABSTRACTAgriculture and forestry rank among industries with the highest rates of occupational fatality and injury. Establishing a nonfatal injury surveillance system is a top priority in the National Occupational Research Agenda. Recently, new sources of data such as Pre-Hospital Care Reports (PCRs) and hospitalization data have transitioned to electronic databases. Using narrative free text and location codes from Maine PCRs, along with International Classification of Diseases (ICD)-9 External Cause of Injury Codes (E-codes) in Maine hospital data, researchers are designing a surveillance system to track farm and forestry injury that utilizes electronic match-merging of the two data sources. For 2008, PCR records produced a total of 104 true agricultural cases. Of these, 66 (63%) were identified from the keyword/visual inspection process alone, 25 (24%) were identified by the farm checkbox only, and the remaining 13 (13%) by both methods. For the 150 unique injury events found in hospitalization data, 146 had the initial episode of care documented in only one of the three hospital files. The emergency department (ED) file had the largest number of these (123/146 = 84.2%), followed by the outpatient file (12/146 = 8.2%) and the inpatient file (11/146 = 7.5%). Of the 250 unique agricultural injuries identified (100 PCR only + 146 hospital only + 4 from both), 66 (26%) would not have been identified without free text review of PCR narrative. The false-positive rate (97.14%) keyword searches underscores that without visual inspection, it is not an effective strategy. Both sources of data (PCR and hospital data) need to be used in a continued surveillance system.

Current Development and Technology in the Information Extraction for Clinical Narrative Text

Most of the clinical narratives are free-text forms. The information extractions from clinical narrative text are more complicated than those from other biomedical texts, such as books, articles, literature abstracts, and so on. In this paper, we review recent published researches on the implication and technology of information extraction from free-text clinical narratives. We mainly introduce the Hidden Markov Model, Support Vector Machine, Ontology and other combined method used in information extraction for clinical narratives. The objectives, typical systems, applications, technologies and future challenge of information extraction for clinical narrative are introduced in this paper.

Support patient search on pathology reports with interactive online learning based data extraction

Background: Structural reporting enables semantic understanding and prompt retrieval of clinical findings about patients. While synoptic pathology reporting provides templates for data entries, information in pathology reports remains primarily in narrative free text form. Extracting data of interest from narrative pathology reports could significantly improve the representation of the information and enable complex structured queries. However, manual extraction is tedious and error-prone, and automated tools are often constructed with a fixed training dataset and not easily adaptable. Our goal is to extract data from pathology reports to support advanced patient search with a highly adaptable semi-automated data extraction system, which can adjust and self-improve by learning from a user’s interaction with minimal human effort. Methods: We have developed an online machine learning based information extraction system called IDEAL-X. With its graphical user interface, the system’s data extraction engine automatically annotates values for users to review upon loading each report text. The system analyzes users’ corrections regarding these annotations with online machine learning, and incrementally enhances and refines the learning model as reports are processed. The system also takes advantage of customized controlled vocabularies, which can be adaptively refined during the online learning process to further assist the data extraction. As the accuracy of automatic annotation improves overtime, the effort of human annotation is gradually reduced. After all reports are processed, a built-in query engine can be applied to conveniently define queries based on extracted structured data. Results: We have evaluated the system with a dataset of anatomic pathology reports from 50 patients. Extracted data elements include demographical data, diagnosis, genetic marker, and procedure. The system achieves F-1 scores of around 95% for the majority of tests. Conclusions: Extracting data from pathology reports could enable more accurate knowledge to support biomedical research and clinical diagnosis. IDEAL-X provides a bridge that takes advantage of online machine learning based data extraction and the knowledge from human’s feedback. By combining iterative online learning and adaptive controlled vocabularies, IDEAL-X can deliver highly adaptive and accurate data extraction to support patient search.

Giving voice to cancer patients: assessing non-specific effects of an integrative oncology therapeutic program via short patient narratives.

The aim of this study was to assess patient perspectives regarding non-specific effects of a complementary medicine (CM) consultation and intervention within an integrative oncology setting. Patients undergoing chemotherapy in a community-based oncology service were referred by oncology healthcare providers to an integrative oncology physician trained in CM-oriented supportive care. Assessment of concerns and well-being was made using the Measure Yourself Concerns and Wellbeing questionnaire, at baseline and after 3 months of CM treatments, which were designed to improve quality of life (QoL) outcomes. Patients were asked to describe the most important aspects of the integrative treatment process. Free-text narratives were examined using content analysis with ATLAS.Ti software for systematic coding. Of 152 patients' narratives analyzed, 44% reported an experience of patient-centered care, including CM practitioners' approach of togetherness, uniqueness, and the invoking of an internal process. CM practitioner approach was experienced within a context of an enhanced sense of confidence; gaining a different perspective; and acquiring emotional resilience and empowerment. Short patient narratives should be considered for patient-reported outcomes, expressing perspectives of both effects and experience of care. CM may promote patient QoL-related outcomes through non-specific effects, enhancing patient-centered care. The benefits of CM dependent on general therapeutic incidental aspects (i.e., common factors) warrant attention regarding non-specific components of treatment.

Creating accountability in image quality analysis part 3: creation of a standardized image-centric mark-up and annotation tool.

While quality assurance (QA) has existed in some form from the inception of modern medicine, its importance and degree of scrutiny is greater than ever today, as evidenced by numerous quality-centric mandates issued from the Institute of Medicine (IOM) [1–3]. These IOM mandates have in turn fostered numerous healthcare and legislative quality initiatives including evidence-based medicine (EBM), meaningful use, and pay for performance [4–7]. The common denominator to these quality initiatives is data, which serves as the means with which measurement takes place and performance is judged. The goal of this data-driven analysis and intervention is to improve quality in healthcare delivery, which in turn is expected to improve clinical outcomes. Unfortunately, in its present form, a great deal of medical data exists in nonstandardized formats, which precludes creation of referenceable databases and meta-analysis [8, 9]. At the same time, despite the “digitization” of vast amounts of medical data, data integration and accessibility remains a problem due to the relative lack of integration between disparate information systems [10, 11]. The combined inabilities to record, access, correlate, and analyze standardized data in medical practice adversely affects these quality initiatives. Despite almost universal support for the principles behind EBM, its widespread applicability is limited by these existing data deficiencies. Attempts to improve quality and standardize data in medical imaging practice have been limited to date by a number of factors including the preferred method of radiology reporting (i.e., nonstandardized and narrative free text), inconsistency of QA standards (with the exception of mammography), lack of supporting quality-centric technology, and heightened emphasis on productivity and workflow enhancement in the face of declining reimbursements and increasing data volume and complexity [12–15]. If the IOM mandates are to be addressed in medical imaging practice, radical innovation is required which simultaneously addresses issues of data standardization and quality improvement without sacrificing workflow and productivity.

Automatic generation of case‐detection algorithms to identify children with asthma from large electronic health record databases

Most electronic health record databases contain unstructured free-text narratives, which cannot be easily analyzed. Case-detection algorithms are usually created manually and often rely only on using coded information such as International Classification of Diseases version 9 codes. We applied a machine-learning approach to generate and evaluate an automated case-detection algorithm that uses both free-text and coded information to identify asthma cases. The Integrated Primary Care Information (IPCI) database was searched for potential asthma patients aged 5-18 years using a broad query on asthma-related codes, drugs, and free text. A training set of 5032 patients was created by manually annotating the potential patients as definite, probable, or doubtful asthma cases or non-asthma cases. The rule-learning program RIPPER was then used to generate algorithms to distinguish cases from non-cases. An over-sampling method was used to balance the performance of the automated algorithm to meet our study requirements. Performance of the automated algorithm was evaluated against the manually annotated set. The selected algorithm yielded a positive predictive value (PPV) of 0.66, sensitivity of 0.98, and specificity of 0.95 when identifying only definite asthma cases; a PPV of 0.82, sensitivity of 0.96, and specificity of 0.90 when identifying both definite and probable asthma cases; and a PPV of 0.57, sensitivity of 0.95, and specificity of 0.67 for the scenario identifying definite, probable, and doubtful asthma cases. The automated algorithm shows good performance in detecting cases of asthma utilizing both free-text and coded data. This algorithm will facilitate large-scale studies of asthma in the IPCI database.

Health care professionals' pain narratives in hospitalized children's medical records. Part 1: pain descriptors.

Although documentation of children's pain by health care professionals is frequently undertaken, few studies have explored the nature of the language used to describe pain in the medical records of hospitalized children. To describe health care professionals' use of written language related to the quality and quantity of pain experienced by hospitalized children. Free-text pain narratives documented during a 24 h period were collected from the medical records of 3822 children (0 to 18 years of age) hospitalized on 32 inpatient units in eight Canadian pediatric hospitals. A qualitative descriptive exploration using a content analysis approach was used. Pain narratives were documented a total of 5390 times in 1518&nbsp;of the 3822 children's medical records (40%). Overall, word choices represented objective and subjective descriptors. Two major categories were identified, with their respective subcategories of word indicators and associated cues: indicators of pain, including behavioural (e.g., vocal, motor, facial and activities cues), affective and physiological cues, and children's descriptors; and word qualifiers, including intensity, comparator and temporal qualifiers. The richness and complexity of vocabulary used by clinicians to document children's pain lend support to the concept that the word 'pain' is a label that represents a myriad of different experiences. There is potential to refine pediatric pain assessment measures to be inclusive of other cues used to identify children's pain. The results enhance the discussion concerning the development of standardized nomenclature. Further research is warranted to determine whether there is congruence in interpretation across time, place and individuals.

Development and evaluation of a Naïve Bayesian model for coding causation of workers’ compensation claims

IntroductionTracking and trending rates of injuries and illnesses classified as musculoskeletal disorders caused by ergonomic risk factors such as overexertion and repetitive motion (MSDs) and slips, trips, or falls (STFs) in different industry sectors is of high interest to many researchers. Unfortunately, identifying the cause of injuries and illnesses in large datasets such as workers’ compensation systems often requires reading and coding the free form accident text narrative for potentially millions of records. MethodTo alleviate the need for manual coding, this paper describes and evaluates a computer auto-coding algorithm that demonstrated the ability to code millions of claims quickly and accurately by learning from a set of previously manually coded claims. ConclusionsThe auto-coding program was able to code claims as a musculoskeletal disorders, STF or other with approximately 90% accuracy. Impact on industryThe program developed and discussed in this paper provides an accurate and efficient method for identifying the causation of workers' compensation claims as a STF or MSD in a large database based on the unstructured text narrative and resulting injury diagnoses. The program coded thousands of claims in minutes. The method described in this paper can be used by researchers and practitioners to relieve the manual burden of reading and identifying the causation of claims as a STF or MSD. Furthermore, the method can be easily generalized to code/classify other unstructured text narratives.

Exploring the role narrative free-text plays in discrepancies between physician coding and the InterVA regarding determination of malaria as cause of death, in a malaria holo-endemic region.

BackgroundIn countries where tracking mortality and clinical cause of death are not routinely undertaken, gathering verbal autopsies (VA) is the principal method of estimating cause of death. The most common method for determining probable cause of death from the VA interview is Physician-Certified Verbal Autopsy (PCVA). A recent alternative method to interpret Verbal Autopsy (InterVA) is a computer model using a Bayesian approach to derive posterior probabilities for causes of death, given an a priori distribution at population level and a set of interview-based indicators. The model uses the same input information as PCVA, with the exception of narrative text information, which physicians can consult but which were not inputted into the model. Comparing the results of physician coding with the model, large differences could be due to difficulties in diagnosing malaria, especially in holo-endemic regions. Thus, the aim of the study was to explore whether physicians' access to electronically unavailable narrative text helps to explain the large discrepancy in malaria cause-specific mortality fractions (CSMFs) in physician coding versus the model.MethodsFree-texts of electronically available records (N = 5,649) were summarised and incorporated into the InterVA version 3 (InterVA-3) for three sub-groups: (i) a 10%-representative subsample (N = 493) (ii) records diagnosed as malaria by physicians and not by the model (N = 1035), and (iii) records diagnosed by the model as malaria, but not by physicians (N = 332). CSMF results before and after free-text incorporation were compared.ResultsThere were changes of between 5.5-10.2% between models before and after free-text incorporation. No impact on malaria CSMFs was seen in the representative sub-sample, but the proportion of malaria as cause of death increased in the physician sub-sample (2.7%) and saw a large decrease in the InterVA subsample (9.9%). Information on 13/106 indicators appeared at least once in the free-texts that had not been matched to any item in the structured, electronically available portion of the Nouna questionnaire.DiscussionFree-texts are helpful in gathering information not adequately captured in VA questionnaires, though access to free-text does not explain differences in physician and model determination of malaria as cause of death.

Characteristics of Finnish and Swedish intensive care nursing narratives: a comparative analysis to support the development of clinical language technologies

BackgroundFree text is helpful for entering information into electronic health records, but reusing it is a challenge. The need for language technology for processing Finnish and Swedish healthcare text is therefore evident; however, Finnish and Swedish are linguistically very dissimilar. In this paper we present a comparison of characteristics in Finnish and Swedish free-text nursing narratives from intensive care. This creates a framework for characterising and comparing clinical text and lays the groundwork for developing clinical language technologies.MethodsOur material included daily nursing narratives from one intensive care unit in Finland and one in Sweden. Inclusion criteria for patients were an inpatient period of least five days and an age of at least 16 years. We performed a comparative analysis as part of a collaborative effort between Finnish- and Swedish-speaking healthcare and language technology professionals that included both qualitative and quantitative aspects. The qualitative analysis addressed the content and structure of three average-sized health records from each country. In the quantitative analysis 514 Finnish and 379 Swedish health records were studied using various language technology tools.ResultsAlthough the two languages are not closely related, nursing narratives in Finland and Sweden had many properties in common. Both made use of specialised jargon and their content was very similar. However, many of these characteristics were challenging regarding development of language technology to support producing and using clinical documentation.ConclusionsThe way Finnish and Swedish intensive care nursing was documented, was not country or language dependent, but shared a common context, principles and structural features and even similar vocabulary elements. Technology solutions are therefore likely to be applicable to a wider range of natural languages, but they need linguistic tailoring.AvailabilityThe Finnish and Swedish data can be found at: http://www.dsv.su.se/hexanord/data/.

Using Safety Data to Describe Common Injury-Producing Events: Examples from the U.S. Air Force

The U.S. military leadership has recently increased its efforts to reduce the number of lost-workday injuries for both the active duty and civilian employee components of the total force. The detailed causes and circumstances of those nonfatal injuries-information needed for injury prevention-has largely been unexplored. The purpose of this project was to determine the utility of Air Force safety data for nonfatal injury prevention. In 2004, events associated with injury-producing mishaps reported through the U.S. Air Force (USAF) Ground Safety Automated System from 1993-2002 (n = 32,812 injuries) were reconstructed. Essential data elements necessary to reconstruct event causes and circumstances were identified in both coded data and in free-text mishap narratives. Activities and mechanisms were coded in a format similar to that of the ICD-10. A taxonomy was then developed to identify hazard scenarios associated with injury-producing activities or mechanisms. Coded data provided only four data elements (activity, injury event/exposure, nature of injury/body part, and outcome) that were sufficiently descriptive for prevention purposes. Therefore, narrative information was coded and analyzed to obtain additional information. The assembled data enabled identification and description of hazard scenarios associated with the most common injury-producing activities and mechanisms. Safety reports from the USAF provide detailed mishap descriptions for lost-workday injuries that could support in-depth analysis and more effective preventive efforts. However, some of the most valuable information is found in the pre-text narratives that require coding and classification, such as was conducted for this report in order to be optimally useful for injury epidemiology and prevention.

Innovations in the Utilization of Health Information Technology in Psychiatric Services

There is ever-increasing activity in applying information technology to various areas of healthcare, including mental health. Examples of innovations include applications in screening, treatment, clinical decision support, communication and coordination, telemedicine, Internet-based education and services, public health research, training and education, and bioinformatics. Issues and challenges include protection of privacy, managing narrative free text, assessing the reliability of information found online, and mitigating impact on clinical workflow. While many of the innovations described will not be fully realized until national information systems reach a larger scale, many are having a positive impact on mental healthcare today.

Negation recognition in medical narrative reports

Substantial medical data, such as discharge summaries and operative reports are stored in electronic textual form. Databases containing free-text clinical narratives reports often need to be retrieved to find relevant information for clinical and research purposes. The context of negation, a negative finding, is of special importance, since many of the most frequently described findings are such. When searching free-text narratives for patients with a certain medical condition, if negation is not taken into account, many of the documents retrieved will be irrelevant. Hence, negation is a major source of poor precision in medical information retrieval systems. Previous research has shown that negated findings may be difficult to identify if the words implying negations (negation signals) are more than a few words away from them. We present a new pattern learning method for automatic identification of negative context in clinical narratives reports. We compare the new algorithm to previous methods proposed for the same task, and show its advantages: accuracy improvement compared to other machine learning methods, and much faster than manual knowledge engineering techniques with matching accuracy. The new algorithm can be applied also to further context identification and information extraction tasks.

The Radiology Report of the Future: A Summary of the 2007 Intersociety Conference

A radiology report is the official record documenting the contribution of a radiologist to a patient's care. The use of structured reports and a common lexicon will help referring physicians better understand the contents of reports. These same features in electronic health records will enable radiologists to mine reports for utilization management information as well as form the basis for clinical investigations.

Injury surveillance in Northern Ireland

A recent development in Northern Ireland shows much promise for injury control in the province. A new electronic injury surveillance system has been introduced at the Royal Belfast Hospital for...