Improving identification of fall-related injuries in ambulatory care using statistical text mining.

OBJECTIVES: We determined whether statistical text mining (STM) can identify fall-related injuries in electronic health record (EHR) documents and the impact on STM models of training on documents from a single or multiple facilities. METHODS: We obtained fiscal year 2007 records for Veterans Health Administration (VHA) ambulatory care clinics in the southeastern United States and Puerto Rico, resulting in a total of 26 010 documents for 1652 veterans treated for fall-related injury and 1341 matched controls. We used the results of an STM model to predict fall-related injuries at the visit and patient levels and compared them with a reference standard based on chart review. RESULTS: STM models based on training data from a single facility resulted in accuracy of 87.5% and 87.1%, F-measure of 87.0% and 90.9%, sensitivity of 92.1% and 94.1%, and specificity of 83.6% and 77.8% at the visit and patient levels, respectively. Results from training data from multiple facilities were almost identical. CONCLUSIONS: STM has the potential to improve identification of fall-related injuries in the VHA, providing a model for wider application in the evolving national EHR system.

Use of artificial neural networks to determine cognitive impairment and therapeutic effectiveness in Alzheimer's transgenic mice.

Behavioral testing of transgenic mouse models of Alzheimer's disease (AD) is the functional endpoint for determining the effectiveness of therapeutic interventions and elucidating AD pathogenesis. Utilizing these mouse models, there have been remarkably few attempts to analyze multiple behavioral measures/tasks with higher-level computation techniques, either to distinguish performance between transgenic groups or to reveal any "overall" cognitive benefit of a given therapeutic. The present study compared the classificatory accuracy of artificial neural networks (ANNs) versus more traditional discriminant function analysis (DFA) using multiple behavioral measures/tasks from two AD transgenic mouse investigations. These investigations were to determine if AD transgenic mice could be cognitively-protected by either long-term caffeine administration (CA) or by a cognitively-stimulating environment (SE). Both the entire set of behavioral measures and a subset of 8 cognitive-based measures were analyzed. Both classifiers revealed a beneficial "overall" effect of CA and SE to protect AD transgenic mice across multiple cognitive measures/tasks. However, for both CA and SE datasets, the ANN was superior to DFA for discerning transgenicity (non-transgenic vs. transgenic-controls) across multiple behavioral measures. These results indicate that ANNs have an excellent capacity to discriminate cognitive impairment in AD transgenic mice and thus designate ANNs as a novel, sensitive method for cognitive assessment in Alzheimer's research.

Special healthcare taxing districts: association with population health status.

BACKGROUND: Locally generated special healthcare taxes are an important component of community infrastructure, but their impact on the health status of populations has not been systematically addressed. METHODS: Florida counties were segmented on the basis of the use/nonuse of locally generated tax dollars for health care during the 1992-1996 period and analyzed in 2004. Linear mixed-effects regression analysis was used to test a model in which taxing behavior served as the primary predictor variable for total age-adjusted and selected cause-specific mortality. Race/ethnicity, rurality, poverty, access to a public hospital, and physician availability were controlled. RESULTS: Local taxation was associated with lower total age-adjusted mortality, and lower mortality for the major causes of death, except stroke, when compared to the state mean. Local taxation is protective relative to total age-adjusted mortality (odds ratio [OR]=0.63, 95% confidence interval [CI]=0.40-0.98) and age-adjusted mortality from chronic obstructive lung disease (OR=0.50, CI=0.32-0.79), cancers (OR=0.53, CI=0.34-0.084), and intentional injury (OR=0.50, CI=0.38-0.92). CONCLUSIONS: Locally generated tax revenues used for the provision of healthcare services are consistently associated with improved health outcomes of major public health importance. The means by which this advantage is achieved will require additional research.

The application of volume-outcome contouring in data warehousing.

Despite a compelling body of published research on the nature of provider volume and clinical outcomes, healthcare executives and policymakers have not managed to develop and implement systems that are useful in directing patients to higher volume providers via selective referral or avoidance. A specialized data warehouse application, utilizing hospital discharge data linked to physician biographical information, allows detailed analysis of physician and hospital volume and the resulting pattern (contour) of related outcomes such as mortality, complications, and medical errors. The approach utilizes a historical repository of hospital discharge data in which the outcomes of interest, important patient characteristics and risk factors used in severity-adjusting of the outcomes are derived from the coding structure of the data.

Finding falls in ambulatory care clinical documents using statistical text mining.

OBJECTIVE: To determine how well statistical text mining (STM) models can identify falls within clinical text associated with an ambulatory encounter. MATERIALS AND METHODS: 2241 patients were selected with a fall-related ICD-9-CM E-code or matched injury diagnosis code while being treated as an outpatient at one of four sites within the Veterans Health Administration. All clinical documents within a 48-h window of the recorded E-code or injury diagnosis code for each patient were obtained (n=26 010; 611 distinct document titles) and annotated for falls. Logistic regression, support vector machine, and cost-sensitive support vector machine (SVM-cost) models were trained on a stratified sample of 70% of documents from one location (dataset Atrain) and then applied to the remaining unseen documents (datasets Atest-D). RESULTS: All three STM models obtained area under the receiver operating characteristic curve (AUC) scores above 0.950 on the four test datasets (Atest-D). The SVM-cost model obtained the highest AUC scores, ranging from 0.953 to 0.978. The SVM-cost model also achieved F-measure values ranging from 0.745 to 0.853, sensitivity from 0.890 to 0.931, and specificity from 0.877 to 0.944. DISCUSSION: The STM models performed well across a large heterogeneous collection of document titles. In addition, the models also generalized across other sites, including a traditionally bilingual site that had distinctly different grammatical patterns. CONCLUSIONS: The results of this study suggest STM-based models have the potential to improve surveillance of falls. Furthermore, the encouraging evidence shown here that STM is a robust technique for mining clinical documents bodes well for other surveillance-related topics.

Using ensemble models to classify the sentiment expressed in suicide notes.

In 2007, suicide was the tenth leading cause of death in the U.S. Given the significance of this problem, suicide was the focus of the 2011 Informatics for Integrating Biology and the Bedside (i2b2) Natural Language Processing (NLP) shared task competition (track two). Specifically, the challenge concentrated on sentiment analysis, predicting the presence or absence of 15 emotions (labels) simultaneously in a collection of suicide notes spanning over 70 years. Our team explored multiple approaches combining regular expression-based rules, statistical text mining (STM), and an approach that applies weights to text while accounting for multiple labels. Our best submission used an ensemble of both rules and STM models to achieve a micro-averaged F(1) score of 0.5023, slightly above the mean from the 26 teams that competed (0.4875).

Throw the bath water out, keep the baby: keeping medically-relevant terms for text mining.

The purpose of this research is to answer the question, can medically-relevant terms be extracted from text notes and text mined for the purpose of classification and obtain equal or better results than text mining the original note? A novel method is used to extract medically-relevant terms for the purpose of text mining. A dataset of 5,009 EMR text notes (1,151 related to falls) was obtained from a Veterans Administration Medical Center. The dataset was processed with a natural language processing (NLP) application which extracted concepts based on SNOMED-CT terms from the Unified Medical Language System (UMLS) Metathesaurus. SAS Enterprise Miner was used to text mine both the set of complete text notes and the set represented by the extracted concepts. Logistic regression models were built from the results, with the extracted concept model performing slightly better than the complete note model.

Using ontology network structure in text mining.

Statistical text mining treats documents as bags of words, with a focus on term frequencies within documents and across document collections. Unlike natural language processing (NLP) techniques that rely on an engineered vocabulary or a full-featured ontology, statistical approaches do not make use of domain-specific knowledge. The freedom from biases can be an advantage, but at the cost of ignoring potentially valuable knowledge. The approach proposed here investigates a hybrid strategy based on computing graph measures of term importance over an entire ontology and injecting the measures into the statistical text mining process. As a starting point, we adapt existing search engine algorithms such as PageRank and HITS to determine term importance within an ontology graph. The graph-theoretic approach is evaluated using a smoking data set from the i2b2 National Center for Biomedical Computing, cast as a simple binary classification task for categorizing smoking-related documents, demonstrating consistent improvements in accuracy.

Hispanic health status in Orange County, Florida.

PURPOSE: To assess the health status of the Hispanic population of Orange County, Florida. METHODS: The methodology utilized secondary data for 66 ethnically identified indicators in a comparative framework applied for a 5-year period (1997-2001). FINDINGS: Orange County Hispanics are younger with lower per capita income than their Florida peers, less likely to be White, and much more likely to be of Puerto Rican origin. Relative to the Hispanic populations in the selected peer counties and statewide, Orange County Hispanics have higher age-adjusted death rates for a majority of disease categories and conditions, such as breast, lung, and prostate cancers; chronic liver disease and cirrhosis; diabetes mellitus; pneumonia and influenza; stroke; acquired immunodeficiency syndrome; motor vehicle accidents; and infant, neonatal, and child mortality. Orange County Hispanics did better in comparison to Orange non-Hispanics, with lower age-adjusted death rates for major causes of death such as heart disease, cancer, and stroke. However, for many indicators, the 5-year trends for Orange County Hispanics are moving in an unfavorable direction in contrast to the trends for non-Hispanics, which are either stable or improving. CONCLUSION: Comparative assessments of Hispanic populations using secondary data enable the development of a comprehensive health status profile. However, this approach is currently constrained by the limited number of ethnically identified indicators and, especially for Hispanics, problems in the accuracy and consistency of the assignment to racial categories and subsequent reporting.