Improving sensitivity of machine learning methods for automated case identification from free-text electronic medical records.

BACKGROUND: Distinguishing cases from non-cases in free-text electronic medical records is an important initial step in observational epidemiological studies, but manual record validation is time-consuming and cumbersome. We compared different approaches to develop an automatic case identification system with high sensitivity to assist manual annotators. METHODS: We used four different machine-learning algorithms to build case identification systems for two data sets, one comprising hepatobiliary disease patients, the other acute renal failure patients. To improve the sensitivity of the systems, we varied the imbalance ratio between positive cases and negative cases using under- and over-sampling techniques, and applied cost-sensitive learning with various misclassification costs. RESULTS: For the hepatobiliary data set, we obtained a high sensitivity of 0.95 (on a par with manual annotators, as compared to 0.91 for a baseline classifier) with specificity 0.56. For the acute renal failure data set, sensitivity increased from 0.69 to 0.89, with specificity 0.59. Performance differences between the various machine-learning algorithms were not large. Classifiers performed best when trained on data sets with imbalance ratio below 10. CONCLUSIONS: We were able to achieve high sensitivity with moderate specificity for automatic case identification on two data sets of electronic medical records. Such a high-sensitive case identification system can be used as a pre-filter to significantly reduce the burden of manual record validation.

Risk of ischemic stroke associated with antidepressant drug use in elderly persons.

Competing hypotheses have been formulated about a possible association between selective serotonin reuptake inhibitors (SSRIs) and ischemic stroke. However, the relationship between antidepressant drug use and ischemic stroke is still unclear. The aim of the study was to assess the association between the use of different types of antidepressants and the risk of ischemic stroke in elderly outpatients. A population-based, nested, case-control study was conducted in persons 65 years and older in the Integrated Primary Care Information database (1996-2005). Cases were all patients with a validated first ischemic stroke. Controls were matched on year of birth, sex, and index date. Exposure to antidepressants was divided in current, past, and nonuse and further categorized by type (SSRI, tricyclic, and other antidepressants), dose, and duration. Conditional logistic regression was used to compare the risk of ischemic stroke between users of antidepressants and nonusers. Overall, 996 incident ischemic strokes were identified. Current use of SSRIs was associated with a significantly increased risk as compared with nonuse (odds ratio, 1.55; 95% confidence interval, 1.07-2.25) in elderly patients, particularly when used for less than six months. No associations were observed for current use of tricyclic and other antidepressant drugs. To summarize, compared with nonuse, only SSRI use seems to be associated with an increased risk of ischemic stroke in elderly patients, particularly as a short-term effect.

Association of community-acquired pneumonia with antipsychotic drug use in elderly patients: a nested case-control study.

BACKGROUND: According to safety alerts from the U.S. Food and Drug Administration, pneumonia is one of the most frequently reported causes of death in elderly patients with dementia who are treated with antipsychotic drugs. However, epidemiologic evidence of the association between antipsychotic drug use and pneumonia is limited. OBJECTIVE: To evaluate whether typical or atypical antipsychotic use is associated with fatal or nonfatal pneumonia in elderly persons. DESIGN: Population-based, nested case-control study. SETTING: Dutch Integrated Primary Care Information database. PATIENTS: Cohort of persons who used an antipsychotic drug, were 65 years or older, and were registered in the IPCI database from 1996 to 2006. Case patients were all persons with incident community-acquired pneumonia. Up to 20 control participants were matched to each case patient on the basis of age, sex, and date of onset. MEASUREMENTS: Risk for fatal or nonfatal community-acquired pneumonia with atypical and typical antipsychotic use. Antipsychotic exposure was categorized by type, timing, and daily dose, and the association with pneumonia was assessed by using conditional logistic regression. RESULTS: 258 case patients with incident pneumonia were matched to 1686 control participants. Sixty-five (25%) of the case patients died in 30 days, and their disease was considered fatal. Current use of either atypical (odds ratio [OR], 2.61 [95% CI, 1.48 to 4.61]) or typical (OR, 1.76 [CI, 1.22 to 2.53]) antipsychotic drugs was associated with a dose-dependent increase in the risk for pneumonia compared with past use of antipsychotic drugs. Only atypical antipsychotic drugs were associated with an increase in the risk for fatal pneumonia (OR, 5.97 [CI, 1.49 to 23.98]). LIMITATIONS: Antipsychotic exposure was based on prescription files. Residual confounding due to unmeasured covariates or severity of disease was possible. CONCLUSION: The use of either atypical or typical antipsychotic drugs in elderly patients is associated in a dose-dependent manner with risk for community-acquired pneumonia.