Search

Objective: The objective of this study was to evaluate the performance of the Framingham Diabetes Risk Scoring Model (FDRSM) in a Canadian population, using the Canadian Primary Care Sentinel Surveillance Network (CPCSSN) database., Methods: We analyzed the records of 571 631 patients, between the ages of 45 and 64, between 2002 and 2005, by extracting the most recent laboratory and examination results, including age, sex, body mass index, fasting blood glucose, high-density lipoprotein, triglycerides and blood pressure. We calculated the risk scores of these patients based on the FDRSM. We tracked these patients for 8 years to find out whether or not they were diagnosed with diabetes. We used the area under the receiver operating characteristics curve (AROC) to estimate the discrimination capability of the FDRSM on our study sample and compared it with the AROC reported in the original Framingham diabetes study., Results: The AROC for our main research sample of 1970 patients for whom all risk factors and follow-up data were available was 78.6% compared to the AROC of 85% reported in the FDRSM. We found that 70.1% of our main sample had risks lower than 3%; 16.3% had risks between 3% and 10%; and 13.6% had risks greater than 10% for diabetes over the following 8-year period., Conclusions: The discrimination capability of the FDRSM Canadian electronic medical records is fair. However, building a more accurate model for predicting diabetes based on the characteristics of Canadian patients is highly recommended., (Copyright © 2015 Canadian Diabetes Association. Published by Elsevier Inc. All rights reserved.)

Maintaining continuity of patient care can be difficult during transitions between care settings, when quality of care and patient safety are at risk. We sought to understand the role of inter-provider documentation and information exchange across successive care settings for complex, older patients. Data were gathered as part of a multi-site ethnographic study that followed 23 older post-surgical hip fracture patients across multiple care transitions. Documentation and information exchange help build a "picture" of complex patients, but care continuity is challenged by late, missing or unnecessary information, duplicated assessments and lack of interoperability of electronic systems. This study demonstrates the importance of appropriate documentation in the transition process, but points to the need for comprehensive system-level solutions which standardize information collection and sharing along the continuum of care.

Objective: As clinical text mining continues to mature, its potential as an enabling technology for innovations in patient care and clinical research is becoming a reality. A critical part of that process is rigid benchmark testing of natural language processing methods on realistic clinical narrative. In this paper, the authors describe the design and performance of three state-of-the-art text-mining applications from the National Research Council of Canada on evaluations within the 2010 i2b2 challenge., Design: The three systems perform three key steps in clinical information extraction: (1) extraction of medical problems, tests, and treatments, from discharge summaries and progress notes; (2) classification of assertions made on the medical problems; (3) classification of relations between medical concepts. Machine learning systems performed these tasks using large-dimensional bags of features, as derived from both the text itself and from external sources: UMLS, cTAKES, and Medline., Measurements: Performance was measured per subtask, using micro-averaged F-scores, as calculated by comparing system annotations with ground-truth annotations on a test set., Results: The systems ranked high among all submitted systems in the competition, with the following F-scores: concept extraction 0.8523 (ranked first); assertion detection 0.9362 (ranked first); relationship detection 0.7313 (ranked second)., Conclusion: For all tasks, we found that the introduction of a wide range of features was crucial to success. Importantly, our choice of machine learning algorithms allowed us to be versatile in our feature design, and to introduce a large number of features without overfitting and without encountering computing-resource bottlenecks.