Extraction of left ventricular ejection fraction information from various types of clinical reports.

Efforts to improve the treatment of congestive heart failure, a common and serious medical condition, include the use of quality measures to assess guideline-concordant care. The goal of this study is to identify left ventricular ejection fraction (LVEF) information from various types of clinical notes, and to then use this information for heart failure quality measurement. We analyzed the annotation differences between a new corpus of clinical notes from the Echocardiography, Radiology, and Text Integrated Utility package and other corpora annotated for natural language processing (NLP) research in the Department of Veterans Affairs. These reports contain varying degrees of structure. To examine whether existing LVEF extraction modules we developed in prior research improve the accuracy of LVEF information extraction from the new corpus, we created two sequence-tagging NLP modules trained with a new data set, with or without predictions from the existing LVEF extraction modules. We also conducted a set of experiments to examine the impact of training data size on information extraction accuracy. We found that less training data is needed when reports are highly structured, and that combining predictions from existing LVEF extraction modules improves information extraction when reports have less structured formats and a rich set of vocabulary.

Resolution of stuttering during ketamine treatment: a case report.

BACKGROUND: Stuttering may include repetition of words in whole or part, difficulty saying words, and elongated pauses in speech. Approximately 5% of children stutter for a period lasting 6 months or more. Most of those children stop stuttering as they approach adulthood, but the condition persists in approximately 1% of adults. The cause of stuttering is unknown. Adults who stutter face substantial burdens in many aspects of their lives. Stutterers may choose not to pursue meaningful employment opportunities, may not be hired for positions they seek, or may be denied promotions or positive performance evaluations. Stuttering can cause physical tension from fear of speaking. Social challenges arise when a person who stutters is seen as less capable or of lower intelligence than fluent speakers. Stuttering causes emotional difficulties through the frustration and embarrassment that disfluent speakers feel. Stutterers may experience a general loss of self-esteem and personal satisfaction in life. Speech therapy is the primary intervention for stuttering. Medications have been investigated as treatments for stuttering, but no medication has been identified that has widespread effectiveness. CASE PRESENTATION: A 60-year-old white non-Hispanic woman who had been a near lifelong stutterer was prescribed ketamine for an unrelated condition and experienced an almost immediate resolution of her stuttering. CONCLUSIONS: Many possible pharmacological treatments for stuttering have been studied. Some medications appear to be effective in some patients; some appear to be more generally effective but have negative side effects. No reporting in relevant literature has addressed a possible role for ketamine in stuttering treatment. On the basis of this case report, research on the effect of ketamine on stuttering would be useful. Any effective treatment for stuttering would have a significant positive effect on quality of life for persons who stutter.

Feasibility of Capturing Cancer Treatment Data in the Utah All-Payer Claims Database.

PURPOSE: Incompleteness of treatment data is a recognized limitation of cancer registry data. An all-payer claims database (APCD) is a tool that states use to capture health care information across systems and payer. We linked the Utah Cancer Registry (UCR) records to Utah's statewide APCD and evaluated how this linkage led to improvements in the capture of cancer treatment information. METHODS: We linked cancers diagnosed and reported to the UCR with Utah APCD claims for the calendar years 2013 and 2014 using LinkPlus Software. For patients with breast or colorectal cancers, manual abstraction was completed to provide a gold-standard comparison for the treatment data obtained from the claims. RESULTS: Among 10,759 reportable cancer occurrences linked to the APCD, the claims identified additional patients with cancer who received therapies that had been unknown to the registry, increasing the proportion treated with chemotherapy from 23.7% to 27.6%, hormone therapy from 14.1% to 18.8%, immunotherapy from 4.3% to 13.2%, and radiation therapy from 24.9% to 27.5%. The APCD increased the sensitivity of treatment variables compared with the abstraction gold standard. Notably, sensitivity of hormonal therapy for breast cancer increased from 78.6% to 95.2% when augmented with APCD claims data. However, the APCD alone did not achieve as high specificity for treatment data as did the data collected through traditional registry methods. CONCLUSIONS: This is the first study, to our knowledge, showing that linking cancer registry data with a statewide claims database that covers multiple insurance companies improves cancer treatment data collection. Linking of cancer registry and APCD data can improve comprehensiveness of cancer registry treatment data.

Linkage between Utah All Payers Claims Database and Central Cancer Registry.

OBJECTIVE: To evaluate the linkage of claims from the Utah All Payers Claims Database (APCD) and Utah Cancer Registry (UCR). DATA SOURCES: Secondary data from 2013 and 2014 Utah APCD and 2013 UCR cases. STUDY DESIGN: This is a descriptive analysis of the quality of linkage between APCD claims data and cancer registry cases. DATA COLLECTION/EXTRACTION METHODS: We used the LinkPlus software to link Utah APCD and UCR data. PRINCIPAL FINDINGS: We were able to link 82.4 percent (9441/11 453) of the UCR reportable cancer cases with APCD claims. Of those linked, 66 percent were perfect matches. CONCLUSIONS: The quality of identifiers is high, evidence that claims data can potentially supplement cancer registry data for use in research.

Automating Quality Measures for Heart Failure Using Natural Language Processing: A Descriptive Study in the Department of Veterans Affairs.

BACKGROUND: We developed an accurate, stakeholder-informed, automated, natural language processing (NLP) system to measure the quality of heart failure (HF) inpatient care, and explored the potential for adoption of this system within an integrated health care system. OBJECTIVE: To accurately automate a United States Department of Veterans Affairs (VA) quality measure for inpatients with HF. METHODS: We automated the HF quality measure Congestive Heart Failure Inpatient Measure 19 (CHI19) that identifies whether a given patient has left ventricular ejection fraction (LVEF) <40%, and if so, whether an angiotensin-converting enzyme inhibitor or angiotensin-receptor blocker was prescribed at discharge if there were no contraindications. We used documents from 1083 unique inpatients from eight VA medical centers to develop a reference standard (RS) to train (n=314) and test (n=769) the Congestive Heart Failure Information Extraction Framework (CHIEF). We also conducted semi-structured interviews (n=15) for stakeholder feedback on implementation of the CHIEF. RESULTS: The CHIEF classified each hospitalization in the test set with a sensitivity (SN) of 98.9% and positive predictive value of 98.7%, compared with an RS and SN of 98.5% for available External Peer Review Program assessments. Of the 1083 patients available for the NLP system, the CHIEF evaluated and classified 100% of cases. Stakeholders identified potential implementation facilitators and clinical uses of the CHIEF. CONCLUSIONS: The CHIEF provided complete data for all patients in the cohort and could potentially improve the efficiency, timeliness, and utility of HF quality measurements.

Evaluation of PHI Hunter in Natural Language Processing Research.

OBJECTIVES: We introduce and evaluate a new, easily accessible tool using a common statistical analysis and business analytics software suite, SAS, which can be programmed to remove specific protected health information (PHI) from a text document. Removal of PHI is important because the quantity of text documents used for research with natural language processing (NLP) is increasing. When using existing data for research, an investigator must remove all PHI not needed for the research to comply with human subjects' right to privacy. This process is similar, but not identical, to de-identification of a given set of documents. MATERIALS AND METHODS: PHI Hunter removes PHI from free-form text. It is a set of rules to identify and remove patterns in text. PHI Hunter was applied to 473 Department of Veterans Affairs (VA) text documents randomly drawn from a research corpus stored as unstructured text in VA files. RESULTS: PHI Hunter performed well with PHI in the form of identification numbers such as Social Security numbers, phone numbers, and medical record numbers. The most commonly missed PHI items were names and locations. Incorrect removal of information occurred with text that looked like identification numbers. DISCUSSION: PHI Hunter fills a niche role that is related to but not equal to the role of de-identification tools. It gives research staff a tool to reasonably increase patient privacy. It performs well for highly sensitive PHI categories that are rarely used in research, but still shows possible areas for improvement. More development for patterns of text and linked demographic tables from electronic health records (EHRs) would improve the program so that more precise identifiable information can be removed. CONCLUSIONS: PHI Hunter is an accessible tool that can flexibly remove PHI not needed for research. If it can be tailored to the specific data set via linked demographic tables, its performance will improve in each new document set.

Exploration of ICD-9-CM coding of chronic disease within the Elixhauser Comorbidity Measure in patients with chronic heart failure.

INTRODUCTION: International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) codes capture comorbidities that can be used to risk adjust nonrandom patient groups. We explored the accuracy of capturing comorbidities associated with one risk adjustment method, the Elixhauser Comorbidity Measure (ECM), in patients with chronic heart failure (CHF) at one Veterans Affairs (VA) medical center. We explored potential reasons for the differences found between the original codes assigned and conditions found through retrospective review. METHODS: This descriptive, retrospective study used a cohort of patients discharged with a principal diagnosis coded as CHF from one VA medical center in 2003. One admission per patient was used in the study; with multiple admissions, only the first admission was analyzed. We compared the assignment of original codes assigned to conditions found in a retrospective, manual review of the medical record conducted by an investigator with coding expertise as well as by physicians. Members of the team experienced with assigning ICD-9-CM codes and VA coding processes developed themes related to systemic reasons why chronic conditions were not coded in VA records using applied thematic techniques. RESULTS: In the 181-patient cohort, 388 comorbid conditions were identified; 305 of these were chronic conditions, originally coded at the time of discharge with an average of 1.7 comorbidities related to the ECM per patient. The review by an investigator with coding expertise revealed a total of 937 comorbidities resulting in 618 chronic comorbid conditions with an average of 3.4 per patient; physician review found 872 total comorbidities with 562 chronic conditions (average 3.1 per patient). The agreement between the original and the retrospective coding review was 88 percent. The kappa statistic for the original and the retrospective coding review was 0.375 with a 95 percent confidence interval (CI) of 0.352 to 0.398. The kappa statistic for the retrospective coding review and physician review was 0.849 (CI, 0.823-0.875). The kappa statistic for the original coding and the physician review was 0.340 (CI, 0.316-0.364). Several systemic factors were identified, including familiarity with inpatient VA and non-VA guidelines, the quality of documentation, and operational requirements to complete the coding process within short time frames and to identify the reasons for movement within a given facility. CONCLUSION: Comorbidities within the ECM representing chronic conditions were significantly underrepresented in the original code assignment. Contributing factors potentially include prioritization of codes related to acute conditions over chronic conditions; coders' professional training, educational level, and experience; and the limited number of codes allowed in initial coding software. This study highlights the need to evaluate systemic causes of underrepresentation of chronic conditions to improve the accuracy of risk adjustment used for health services research, resource allocation, and performance measurement.

Epinome - a novel workbench for epidemic investigation and analysis of search strategies in public health practice.

We present a novel user-centric visual analytics system that supports investigation of simulated disease outbreak and the study of decision-making. We developed Epinome as part of our research on decision making in public health and in particular, on the evaluation of information search strategies in public health practice. Epinome is a highly dynamic web-based system that provides a platform to track and study subjects' decision making and information search strategies, under controlled and repeatable conditions using simulated disease outbreaks. In this paper we focus on the design and implementation of Epinome and present relevant results from field tests we conducted in Utah and Colorado.