N-gram support vector machines for scalable procedure and diagnosis classification, with applications to clinical free text data from the intensive care unit.

BACKGROUND: Existing risk adjustment models for intensive care unit (ICU) outcomes rely on manual abstraction of patient-level predictors from medical charts. Developing an automated method for abstracting these data from free text might reduce cost and data collection times. OBJECTIVE: To develop a support vector machine (SVM) classifier capable of identifying a range of procedures and diagnoses in ICU clinical notes for use in risk adjustment. MATERIALS AND METHODS: We selected notes from 2001-2008 for 4191 neonatal ICU (NICU) and 2198 adult ICU patients from the MIMIC-II database from the Beth Israel Deaconess Medical Center. Using these notes, we developed an implementation of the SVM classifier to identify procedures (mechanical ventilation and phototherapy in NICU notes) and diagnoses (jaundice in NICU and intracranial hemorrhage (ICH) in adult ICU). On the jaundice classification task, we also compared classifier performance using n-gram features to unigrams with application of a negation algorithm (NegEx). RESULTS: Our classifier accurately identified mechanical ventilation (accuracy=0.982, F1=0.954) and phototherapy use (accuracy=0.940, F1=0.912), as well as jaundice (accuracy=0.898, F1=0.884) and ICH diagnoses (accuracy=0.938, F1=0.943). Including bigram features improved performance on the jaundice (accuracy=0.898 vs 0.865) and ICH (0.938 vs 0.927) tasks, and outperformed NegEx-derived unigram features (accuracy=0.898 vs 0.863) on the jaundice task. DISCUSSION: Overall, a classifier using n-gram support vectors displayed excellent performance characteristics. The classifier generalizes to diverse patient populations, diagnoses, and procedures. CONCLUSIONS: SVM-based classifiers can accurately identify procedure status and diagnoses among ICU patients, and including n-gram features improves performance, compared to existing methods.

Adequacy of lymph node examination in colorectal surgery: contribution of the hospital versus the surgeon.

BACKGROUND: Examination of at least 12 lymph nodes (LNs) in the staging of colon cancer (CC) was recommended by the National Comprehensive Cancer Network in 2000; however, rates of an adequate examination remain low. This study compares the impact of the hospital contextual variance against that of the operating surgeon on delivery of an adequate LN examination. STUDY DESIGN: Retrospective analysis of California Cancer Registry data for all CC operations (2001-2006). Hierarchical models predicted the adequacy of LN examination as a function of patient, surgeon, and hospital characteristics. Models were created using penalized quasi-likelihood approximation with second order Taylor linearization as implemented in MLwiN 2.15. RESULTS: A total of 25,606 resections involving 3376 surgeons operating in 346 hospitals were analyzed. Half of cases had an adequate examination. Hierarchical models showed the median odds of an adequate examination associated with the hospital context [(MORhosp 2.05; 95% confidence interval, 1.9-2.2) was much higher than that associated with the surgeon (MORsurg 1.34; 95% confidence interval, 1.2-1.4)]. Hospital characteristics teaching and high volume predicted higher odds of an adequate examination. There was no association with hospital revenue. CONCLUSIONS: Approximately half of patients undergoing surgery for CC received an adequate LN examination. Hospital contextual factors had a stronger association with receipt of an adequate examination than surgeon factors. Our results suggest that quality improvement initiatives and incentives should be targeted at the hospital level to achieve the highest impact. Furthermore, we have identified nonteaching and low volume settings as rational targets for these efforts.

Complex visual hallucinations and attentional performance in eye disease and dementia: a test of the Perception and Attention Deficit model.

OBJECTIVE: This study aimed to test the prediction from the Perception and Attention Deficit model of complex visual hallucinations (CVH) that impairments in visual attention and perception are key risk factors for complex hallucinations in eye disease and dementia. METHODS: Two studies ran concurrently to investigate the relationship between CVH and impairments in perception (picture naming using the Graded Naming Test) and attention (Stroop task plus a novel Imagery task). The studies were in two populations-older patients with dementia (n = 28) and older people with eye disease (n = 50) with a shared control group (n = 37). The same methodology was used in both studies, and the North East Visual Hallucinations Inventory was used to identify CVH. RESULTS: A reliable relationship was found for older patients with dementia between impaired perceptual and attentional performance and CVH. A reliable relationship was not found in the population of people with eye disease. CONCLUSIONS: The results add to previous research that object perception and attentional deficits are associated with CVH in dementia, but that risk factors for CVH in eye disease are inconsistent, suggesting that dynamic rather than static impairments in attentional processes may be key in this population.

Specific attentional impairments and complex visual hallucinations in eye disease.

OBJECTIVE: To test the prediction by the Perception and Attention Deficit (PAD) model of complex visual hallucinations that cognitive impairment, specifically in visual attention, is a key risk factor for complex hallucinations in eye disease. METHODS: Two studies of elderly patients with acquired eye disease investigated the relationship between complex visual hallucinations (CVH) and impairments in general cognition and verbal attention (Study 1) and between CVH, selective visual attention and visual object perception (Study 2). The North East Visual Hallucinations Inventory was used to classify CVH. RESULTS: In Study 1, there was no relationship between CVH (n=10/39) and performance on cognitive screening or verbal attention tasks. In Study 2, participants with CVH (n=11/31) showed poorer performance on a modified Stroop task (p<0.05), a novel imagery-based attentional task (p<0.05) and picture (p<0.05) but not silhouette naming (p=0.13) tasks. Performance on these tasks correctly classified 83% of the participants as hallucinators or non-hallucinators. CONCLUSIONS: The results suggest that, consistent with the PAD model, complex visual hallucinations in people with acquired eye disease are associated with visual attention impairment.

Extreme makeover: Transformation of the veterans health care system.

The veterans health care system administered by the U.S. Department of Veterans Affairs (VA) was established after World War I to provide health care for veterans who suffered from conditions related to their military service. It has grown to be the nation's largest integrated health care system. As the system grew, a number of factors contributed to its becoming increasingly dysfunctional. By the mid-1990s, VA health care was widely criticized for providing fragmented and disjointed care of unpredictable and irregular quality, which was expensive, difficult to access, and insensitive to individual needs. Between 1995 and 1999, the VA health care system was reengineered, focusing especially on management accountability, care coordination, quality improvement, resource allocation, and information management. Numerous systemic changes were implemented, producing dramatically improved quality, service, and operational efficiency. VA health care is now considered among the best in America, and the VA transformation is viewed as a model for health care reform.

Why physicians favor use of percutaneous coronary intervention to medical therapy: a focus group study.

BACKGROUND: Percutaneous coronary intervention (PCI) is performed in many patients with stable coronary artery disease, despite evidence of little clinical benefit over optimal medical therapy. OBJECTIVE: To examine physicians' beliefs, practices, and decision-making regarding elective PCI. DESIGN: Six focus groups, three with primary care physicians and three with cardiologists. Participants discussed PCI using hypothetical case scenarios. Transcripts were analyzed using grounded theory, and commonly expressed themes regarding the decision-making pathway to PCI were identified. PARTICIPANTS: Twenty-eight primary care physicians and 20 interventional and non-interventional cardiologists in Butte County, Orange County, and San Francisco Bay Area, California, in 2006. RESULTS: A number of factors led primary care physicians to evaluate non-symptomatic or minimally symptomatic patients for coronary artery disease and refer them to a cardiologist. The use of screening tests often led to additional testing and referral, as well as fear of missing a coronary stenosis, perceived patient expectations, and medicolegal concerns. The end result was a cascade such that any positive test would generally lead to the catheterization lab, where an "oculostenotic reflex" made PCI a virtual certainty. CONCLUSIONS: The widespread use of PCI in patients with stable coronary artery disease--despite evidence of little benefit in outcomes over medical therapy--may in part be due to psychological and emotional factors leading to a cascade effect wherein testing leads inevitably to PCI. Determining how to help physicians better incorporate evidence-based medicine into decision-making has important implications for patient outcomes and the optimal use of new technologies.