Antidepressant Use in the Elderly Is Associated With an Increased Risk of Dementia.

A retrospective cohort study was conducted including 3688 patients age 60 years or older without dementia enrolled in a depression screening study in primary care clinics. Information on antidepressant use and incident dementia during follow-up was retrieved from electronic medical records. The Cox proportional hazard models were used to compare the risk for incident dementia among 5 participant groups: selective serotonin re-uptake inhibitors (SSRI) only, non-SSRI only (non-SSRI), mixed group of SSRI and non-SSRI, not on antidepressants but depressed, and not on antidepressants and not depressed. SSRI and non-SSRI users had significantly higher dementia risk than the nondepressed nonusers (hazard ratio [HR]=1.83, P=0.0025 for SSRI users and HR=1.50, P=0.004 for non-SSRI users). In addition, SSRIs users had significantly higher dementia risk than non-users with severe depression (HR=2.26, P=0.0005). Future research is needed to confirm our results in other populations and to explore potential mechanism underlying the observed association.

Identification of Patients with Family History of Pancreatic Cancer--Investigation of an NLP System Portability.

In this study we have developed a rule-based natural language processing (NLP) system to identify patients with family history of pancreatic cancer. The algorithm was developed in a Unstructured Information Management Architecture (UIMA) framework and consisted of section segmentation, relation discovery, and negation detection. The system was evaluated on data from two institutions. The family history identification precision was consistent across the institutions shifting from 88.9% on Indiana University (IU) dataset to 87.8% on Mayo Clinic dataset. Customizing the algorithm on the the Mayo Clinic data, increased its precision to 88.1%. The family member relation discovery achieved precision, recall, and F-measure of 75.3%, 91.6% and 82.6% respectively. Negation detection resulted in precision of 99.1%. The results show that rule-based NLP approaches for specific information extraction tasks are portable across institutions; however customization of the algorithm on the new dataset improves its performance.

DEEPEN: A negation detection system for clinical text incorporating dependency relation into NegEx.

In Electronic Health Records (EHRs), much of valuable information regarding patients' conditions is embedded in free text format. Natural language processing (NLP) techniques have been developed to extract clinical information from free text. One challenge faced in clinical NLP is that the meaning of clinical entities is heavily affected by modifiers such as negation. A negation detection algorithm, NegEx, applies a simplistic approach that has been shown to be powerful in clinical NLP. However, due to the failure to consider the contextual relationship between words within a sentence, NegEx fails to correctly capture the negation status of concepts in complex sentences. Incorrect negation assignment could cause inaccurate diagnosis of patients' condition or contaminated study cohorts. We developed a negation algorithm called DEEPEN to decrease NegEx's false positives by taking into account the dependency relationship between negation words and concepts within a sentence using Stanford dependency parser. The system was developed and tested using EHR data from Indiana University (IU) and it was further evaluated on Mayo Clinic dataset to assess its generalizability. The evaluation results demonstrate DEEPEN, which incorporates dependency parsing into NegEx, can reduce the number of incorrect negation assignment for patients with positive findings, and therefore improve the identification of patients with the target clinical findings in EHRs.

Redefined blood pressure variability measure and its association with mortality in elderly primary care patients.

Visit-to-visit blood pressure (BP) variability has received considerable attention recently. The objective of our study is to define a variability measure that is independent of change over time and determine the association between longitudinal summary measures of BP measurements and mortality risk. Data for the study came from a prospective cohort of 2906 adults, aged ≥60 years, in an urban primary care system with ≤15 years of follow-up. Dates of death for deceased participants were retrieved from the National Death Index. Systolic and diastolic BP measurements from outpatient clinic visits were extracted from the Regenstrief Medical Record System. For each patient, the intercept, regression slope, and root mean square error for visit-to-visit variability were derived using linear regression models and used as independent variables in Cox proportional hazards models for both all-cause mortality and mortality attributable to coronary heart disease or stroke. Rate of change was associated with mortality risk in a U-shaped relationship and that participants with little or no change in BP had the lowest mortality risk. BP variability was not an independent predictor of mortality risk. By separating change over time from visit-to-visit variability in studies with relatively long follow-up, we demonstrated in this elderly primary care patient population that BP changes over time, not variability, were associated with greater mortality risk. Future research is needed to confirm our findings in other populations.

An efficient pancreatic cyst identification methodology using natural language processing.

Pancreatic cancer is one of the deadliest cancers, mostly diagnosed at late stages. Patients with pancreatic cysts are at higher risk of developing cancer and their surveillance can help to diagnose the disease in earlier stages. In this retrospective study we collected a corpus of 1064 records from 44 patients at Indiana University Hospital from 1990 to 2012. A Natural Language Processing (NLP) system was developed and used to identify patients with pancreatic cysts. NegEx algorithm was used initially to identify the negation status of concepts that resulted in precision and recall of 98.9% and 89% respectively. Stanford Dependency parser (SDP) was then used to improve the NegEx performance resulting in precision of 98.9% and recall of 95.7%. Features related to pancreatic cysts were also extracted from patient medical records using regex and NegEx algorithm with 98.5% precision and 97.43% recall. SDP improved the NegEx algorithm by increasing the recall to 98.12%.

Natural language processing for the development of a clinical registry: a validation study in intraductal papillary mucinous neoplasms.

BACKGROUND: Medical natural language processing (NLP) systems have been developed to identify, extract and encode information within clinical narrative text. However, the role of NLP in clinical research and patient care remains limited. Pancreatic cysts are common. Some pancreatic cysts, such as intraductal papillary mucinous neoplasms (IPMNs), have malignant potential and require extended periods of surveillance. We seek to develop a novel NLP system that could be applied in our clinical network to develop a functional registry of IPMN patients. OBJECTIVES: This study aims to validate the accuracy of our novel NLP system in the identification of surgical patients with pathologically confirmed IPMN in comparison with our pre-existing manually created surgical database (standard reference). METHODS: The Regenstrief EXtraction Tool (REX) was used to extract pancreatic cyst patient data from medical text files from Indiana University Health. The system was assessed periodically by direct sampling and review of medical records. Results were compared with the standard reference. RESULTS: Natural language processing detected 5694 unique patients with pancreas cysts, in 215 of whom surgical pathology had confirmed IPMN. The NLP software identified all but seven patients present in the surgical database and identified an additional 37 IPMN patients not previously included in the surgical database. Using the standard reference, the sensitivity of the NLP program was 97.5% (95% confidence interval [CI] 94.8-98.9%) and its positive predictive value was 95.5% (95% CI 92.3-97.5%). CONCLUSIONS: Natural language processing is a reliable and accurate method for identifying selected patient cohorts and may facilitate the identification and follow-up of patients with IPMN.

Natural language processing to extract follow-up provider information from hospital discharge summaries.

OBJECTIVE: We evaluate the performance of a Natural Language Processing (NLP) application designed to extract follow-up provider information from free-text discharge summaries at two hospitals. EVALUATION: We compare performance by the NLP application, called the Regenstrief EXtracion tool (REX), to performance by three physician reviewers at extracting follow-up provider names, phone/fax numbers and location information. Precision, recall, and F-measures are reported, with 95% CI for pairwise comparisons. RESULTS: Of 556 summaries with follow-up information, REX performed as follows in precision, recall, F-measure respectively: Provider Name 0.96, 0.92, 0.94; Phone/Fax 0.99, 0.92, 0.96; Location 0.83, 0.82, 0.82. REX was as good as all physician-reviewers in identifying follow-up provider names and phone/fax numbers, and slightly inferior to two physicians at identifying location information. REX took about four seconds (vs. 3-5 minutes for physician-reviewers) to extract follow-up information. CONCLUSION: A NLP program had physician-like performance at extracting provider follow-up information from discharge summaries.

Comparing methods for identifying pancreatic cancer patients using electronic data sources.

We sought to determine the accuracy of two electronic methods of identifying pancreatic cancer in a cohort of pancreatic cyst patients, and to examine the reasons for identification failure. We used the International Classification of Diseases, 9(th) Edition (ICD-9) codes and natural language processing (NLP) technology to identify pancreatic cancer in these patients. We compared both methods to a human-validated gold-standard surgical database. Both ICD-9 codes and NLP technology achieved high sensitivity for identifying pancreatic cancer, but the ICD-9 code method achieved markedly lower specificity and PPV compared to the NLP method. The NLP method required only slightly greater expenditures of time and effort compared to the ICD-9 code method. We identified several variables influencing the accuracy of ICD-9 codes to identify cancer patients including: the identification algorithm, kind of cancer to be identified, presence of other conditions similar to cancer, and presence of conditions that are precancerous.

Adequacy of hospital discharge summaries in documenting tests with pending results and outpatient follow-up providers.

BACKGROUND: Poor communication of tests whose results are pending at hospital discharge can lead to medical errors. OBJECTIVE: To determine the adequacy with which hospital discharge summaries document tests with pending results and the appropriate follow-up providers. DESIGN: Retrospective study of a randomly selected sample PATIENTS: Six hundred ninety-six patients discharged from two large academic medical centers, who had test results identified as pending at discharge through queries of electronic medical records. INTERVENTION AND MEASUREMENTS: Each patient's discharge summary was reviewed to identify whether information about pending tests and follow-up providers was mentioned. Factors associated with documentation were explored using clustered multivariable regression models. MAIN RESULTS: Discharge summaries were available for 99.2% of 668 patients whose data were analyzed. These summaries mentioned only 16% of tests with pending results (482 of 2,927). Even though all study patients had tests with pending results, only 25% of discharge summaries mentioned any pending tests, with 13% documenting all pending tests. The documentation rate for pending tests was not associated with level of experience of the provider preparing the summary, patient's age or race, length of hospitalization, or duration it took for results to return. Follow-up providers' information was documented in 67% of summaries. CONCLUSION: Discharge summaries are grossly inadequate at documenting both tests with pending results and the appropriate follow-up providers.

Which observations from the complete blood cell count predict mortality for hospitalized patients?

BACKGROUND: Information on the prognostic utility of the admission complete blood count (CBC) and differential count is lacking. OBJECTIVE: To identify independent predictors of mortality from the varied number and morphology of cells in the complete blood count defined as a hemogram, automated five cell differential count and manual differential count. DESIGN: Retrospective cohort study and chart review. SETTING: Wishard Memorial Hospital, a large urban primary care hospital. PATIENTS: A total of 46,522 adult inpatients admitted over 10 years to Wishard Memorial Hospital-from January 1993 through December 2002. INTERVENTION: None. MEASUREMENTS: Thirty-day mortality measured from day of admission as determined by electronic medical records and Indiana State death records. RESULTS: Controlling for age and sex, the multivariable regression model identified 3 strong independent predictors of 30-day mortality-nucleated red blood cells (NRBCs), burr cells, and absolute lymphocytosis-each of which was associated with a 3-fold increase in the risk of death within 30 days. The presence of nucleated RBCs was associated with a 30-day mortality rate of 25.5% across a range of diagnoses, excluding patients with sickle-cell disease and obstetric patients, for whom NRBCs were not associated with increased mortality. Having burr cells was associated with a mortality rate of 27.3% and was found most commonly in patients with renal or liver failure. Absolute lymphocytosis predicted poor outcome in patients with trauma and CNS injury. CONCLUSIONS: Among patients admitted to Wishard Memorial Hospital, the presence of nucleated RBCs, burr cells, or absolute lymphocytosis at admission was each independently associated with a 3-fold increase in risk of death within 30 days of admission.

Patients with elevated liver enzymes are not at higher risk for statin hepatotoxicity.

BACKGROUND & AIMS: Studies that evaluate the risk of hepatotoxicity from statins in hyperlipidemic subjects with elevated baseline serum transaminases are lacking. We conducted a study to test the hypothesis that patients with elevated baseline liver enzymes have higher risk of statin hepatotoxicity. METHODS: Our study consisted of the following 3 cohorts of patients seen between January 1, 1998 and June 31, 2002: Cohort 1: 342 hyperlipidemic patients with elevated baseline enzymes (AST >40 IU/L or ALT >35 IU/L) who were prescribed a statin; cohort 2: 1437 hyperlipidemic patients with normal transaminases who were prescribed a statin; and cohort 3: 2245 patients with elevated liver enzymes but who were not prescribed a statin. The effect of statins on liver biochemistries was assessed over a 6-month period after statins were prescribed. Elevations in liver biochemistries during follow-up were categorized into mild-moderate or severe based on predefined criteria. RESULTS: The incidence of mild-moderate elevations and severe elevations in liver biochemistries in cohort 1 were 4.7% and 0.6%, respectively. Compared with cohort 1, individuals in cohort 2 had lower incidence of mild-moderate elevations (1.9%, P = 0.002) but not severe elevations (0.2%, P = 0.2). However, between cohorts 1 and 3, there were no differences in the incidence of mild-moderate elevations (4.7% vs. 6.4%, respectively, P = 0.2) or severe elevations (0.6% vs. 0.4%, respectively, P = 0.6). Statin discontinuation during the follow-up was similar between cohorts 1 and 2 (11.1% vs. 10.7%, respectively, P = 0.8). CONCLUSIONS: These data suggest that individuals with elevated baseline liver enzymes do not have higher risk for hepatotoxicity from statins.