Background. About 5--15% of the North American and European populations pass a urinary stone in their lifetime. The most accurate way of detecting degree of urinary obstruction is with a radiograph called an intravenous pyelogram or IVP. Objectives. (1) To assess the feasibility of developing a clinical decision rule that would identify the clinical factors that are sensitive in predicting which patients with suspected ureteral calculi have severe obstruction and require urgent pyelography by: (i) defining severe obstruction; (ii) determining the incidence of severe obstruction; (iii) identifying potential predictor variables; and (iv) calculating a sample size for the definitive decision rule study. (2) To prospectively assess clinical predictors of complications from urinary calculi after discharge from the ED by: (i) determining the incidence of complications; (ii) identifying preliminary predictor variables; (iii) calculating a sample size for the definitive study. (Abstract shortened by UMI.)

Context. Subarachnoid haemorrhage is a type of haemorrhagic stroke with an annual incidence of 1 per 10,000 people and accounts for 1% of emergency department visits for acute headache. Objectives. (1) Phase 0: To determine patient characteristics, association of predictor variables with subarachnoid haemorrhage, mean length of stay determined by investigation(s) performed, and an estimate of the incidence of subarachnoid haemorrhage in alert patients with headache and normal neurological examination at the Ottawa Hospital. (2) Phase 1: To develop and pre-test standardized clinical assessments in alert patients with possible subarachnoid haemorrhage and apply these standardized clinical assessments to alert patients with possible subarachnoid haemorrhage. With these clinical assessments, determine the statistical association between the clinical findings and the diagnosis of subarachnoid haemorrhage, and develop a preliminary decision rule to determine which patients require investigation to rule out subarachnoid haemorrhage. (Abstract shortened by UMI.)

Introduction. Optimal management of paroxysmal atrial fibrillation (PAF), a common presenting complaint in emergency departments (EDs), remains undetermined. Methods. Six month prospective observational study at three EDs. Patients had clinically stable PAF for less than 48 hours. Conservative (rate control) and aggressive (pharmacologic and/or electrical cardioversion) treatment were analyzed. Results. 169 patients were analyzed, 32 treated conservatively and 137 aggressively. 83.9% of aggressively treated patients converted in the ED, 8.0% were admitted, and 52.3% stayed in sinus rhythm for four weeks. The corresponding proportions for conservative treatment were 34.4%, 37.5%, and 30.0%. There were 15 ED complications (2 rate control, 4 pharmacologic, and 9 electrical), two required admission (one pharmacologic and one electrical). No thromboembolism occurred by four-week follow-up. Conclusions. The results of this study---the first prospective study of ED treatment of PAF---will be used to plan a randomized controlled trial which will compare the two treatments.

Introduction. The treatment of cellulitis with intravenous antibiotics administered in Emergency Departments is a new phenomenon with significant inter-physician variation. A clinical trial will address many of the questions surrounding this practice. Previous trials have been flawed because of the absence of a validated objective outcome measure. Methods. Eligible patients with cellulitis were prospectively recruited for an observational cohort study and underwent daily measurements of their infection. These measurements were evaluated for their feasibility, inter-rater reliability and criterion validity (compared with the primary outcome of treatment failure versus clinical response; a classification based on physician treatment decisions) Results. Only the infection size and change in size over time performed well, obtaining statistical significance in all domains. Conclusions. The change in size of infection over time is a valid and reliable reflection of clinical decisions for patients with cellulitis, and should be used as the primary outcome for clinical trials of cellulitis therapy.

This thesis demonstrates the application of a feedforward backpropagation-trained artificial neural network using the weight-elimination cost function to the estimation of in-hospital mortality for coronary artery bypass grafting patients from the San Francisco Heart Institute in Daly City, California, USA. The highly-skewed a priori statistics due to the low mortality rate present difficulties for modelling this data. Artificial training and test datasets with higher mortality rates were developed to improve the classification performance of the artificial neural networks. Sensitivity was considered the most important measure of performance for this work. Given that current mortality risk models are unable to accurately identify high-risk patients (those who do not survive the surgery), focussing on increasing the sensitivity rate will indicate when more of the patients who are difficult to classify are correctly identified. The final result was an increase in sensitivity when training with a dataset with a higher mortality rate than the test set. This dataset modification approach resulted in only small changes for other performance measures (specificity, predictive positive value, predictive negative value, and correct classification rate), and thus helped to achieve the main goal of the study.

There are no available guidelines for the diagnosis of meniscal tears in patients with osteoarthritis of the knee. The goal of the present study was to prospectively study a cohort of such patients referred for arthroscopy and develop a clinical prediction rule to select patients who would benefit from arthroscopy. Patients with osteoarthritis of the knee refractory to conservative medical treatment were selected for arthroscopic evaluation. A standardized assessment protocol was administered to each patient and repeated by a second observer when feasible to permit calculation of the kappa coefficient ($\kappa$) for interobserver agreement. Arthroscopic determination of unstable meniscal tears was recorded by one observer who reviewed a video recording and was blinded to preoperative data based on objective criteria. Those variables which had the highest interobserver agreement and the strongest association with meniscal tear by univariable methods were then entered into a logistic regression model. This study underscored the difficulty in using clinical variables to predict unstable medial meniscal tears which were addressed at arthroscopy. (Abstract shortened by UMI.)

Introduction. Surrogate marker responses are imperfect indicators of response to antiretroviral therapy in HIV. It is proposed that the area under the curve of surrogate marker response will be superior to peak response, or to that measured after a period of therapy. Methods. The database from a study of ritonavir in advanced HIV was used. Using logistic regression, the specificity of the surrogate marker level at baseline, change between baseline and at time points to week 16, peak response, and area under the curve of the response to week 16 and week 40 were determined. The predictive values, likelihood ratios and receiver operating characteristic curves were determined for those of highest specificity. Results. Specificities increased from baseline to week 16. Peak responses were inferior to time period surrogate marker changes, whereas the areas under the curve were comparable or better than the time period surrogate marker changes. The highest specificity at any time point was for the CD4 change at week 8 (55.90%), whereas the highest overall specificity was for the 16 week AUC for the CD4% (69.63%). However, the PPV, NPV, likelihood ratios and ROC curves demonstrated poor performance overall for these surrogate markers. Too few subjects had viral load testing for this marker to be assessed. Discussion. Within the limits of this study, it was demonstrated that the CD4 and CD4% were the surrogate markers most associated with clinical outcome, with the CD4% AUC to 16 weeks having the highest overall specificity and the week 8 CD4 having the greatest specificity for clinical use. However, all surrogate markers had specificities below 70%.

Background. The discharge summary is an important method of communicating information necessary for continuing patient care. Deficiencies in discharge summary content, accuracy, and timeliness are a common problem. To improve summaries, studies have assessed physician-in-training (housestaff) education, alteration of the summary format; and collating information within computer databases into a summary. The most promising technology for improving discharge summaries is the clinical database. Primary objective. To compare medical discharge summaries generated from a clinical database with those created by narrative letter dictation. Design. Randomized controlled trial assigning patients to database (DB) or dictated (dictated-random or D-R) summaries. This was preceded by a prospective cohort study where summaries for all patients were created by dictation (dictated-cohort or D-C group). Inclusion criteria and study setting. All patients admitted to and discharged alive from the general internal medicine service in a tertiary care teaching hospital. Intervention. For patients randomized to the DB group, housestaff completed forms capturing hospital and discharge information important for the discharge summary. This information was entered into a database along with portions of the admission history and physical-examination form. Information in the database was collated into a report to create a database discharge summary. For patients in the D-R and the D-C group, housestaff used narrative dictation to create summaries. (Abstract shortened by UMI.)

For patients with an acute myocardial infarction (AMI), the time from symptom onset to initiation of medical treatment is a critical determinant of outcome both for survival and for preservation of myocardial tissue. This is true because the effectiveness of thrombolytic therapy is, inversely related to time between onset of symptoms and initiation of therapy. Studies have shown that the time that patients take to decide to seek help accounts for most of the delay, a factor that has not improved despite efforts in education campaigns. Thus far, only two studies on delay time have been conducted in Canada. The study objective was to describe the delay time and investigate the factors associated with delay in response to symptoms of AMI in a Canadian setting. (Abstract shortened by UMI.)

The objective of this study was to systematically review the literature and to statistically summarize the evidence evaluating the efficacy of acute normovolemic hemodilution (ANH). Review of 1575 citations identified 24 prospective, randomized, controlled trails of ANH containing a total of 1216 patients. When all trials were pooled ANH reduced the likelihood of exposure to allogeneic blood and the total units of allogeneic blood transfused. Marked heterogeneity of the results was present and adverse events were incompletely reported. Subgroup analyses could not account for all heterogeneity noted in this overview. Under reporting of small trials with negative results was suspected. In trials using a protocol to guide perioperative transfusion, ANH failed to reduce either the likelihood of transfusion or the units administered. It is possible that experimental design bias in favor of reducing transfusion in the ANH group is in part responsible for the reported efficacy of this technique.