Background: Based on survey data from both Canada and abroad, most people would prefer to be cared for and to die in their own homes. Although 70% of elderly patients state a preference for comfort care over high technology life prolonging treatment in an inpatient setting, 54% are still admitted to intensive care units (ICUs). Understanding their wishes regarding end-of-life care, and being able to engage in evidence informed end-of-life discussions has never been so important, in order to empower patients, and to optimize scarce resource management. For the purpose of this thesis, “very old” patients will be defined as those eighty years of age and older. All three manuscripts will be based on data from the Realistic 80 study, a prospective cohort trial of 1671 critically ill very old patients admitted to 22 Canadian ICUs. Objectives: Manuscript 1: To describe the hospital outcomes of the entire cohort of Realistic 80 patients, including their ICU mortality and length of stay, their hospital mortality and length of stay, and their ultimate dispositions. Manuscript 2: To derive a clinical prediction rule for hospital mortality in the medical patient cohort. Manuscript 3: To derive a clinical prediction rule for hospital mortality in the emergency surgical patient cohort. Data Source: A prospective, multicenter cohort study of very elderly medical and surgical patients admitted to 22 Canadian academic and non-academic ICUs. Methods: Clinical decision rule methodology was used to analyze the data set and to create two separate clinical prediction tools, one for critically ill elderly medical patients, and one for critically ill surgical emergency patients. A third manuscript describing general clinical outcomes was also produced. Results of Manuscript 1: A total of 1671 patients were included in this section of the “Realities, Expectations and Attitudes to Life Support Technologies in Intensive Care for Octogenarians: The Realistic 80 Study (a prospective cohort of nearly 2000 critically ill Canadian patients over eighty years old enrolled from 22 ICUs across Canada) that will provide the data for this thesis. The Realistic 80 cohort had a mean age of 84.5, a baseline Apache II score of 22.4, a baseline SOFA score of 5.3, an overall ICU mortality of 21.8%, and an overall hospital mortality of 35%. The cohort had a median ICU length of stay of 3.7 days, and an overall median hospital length of stay of 16.6 days. Only 46.4% of the survivors were able to return home to live. Results of Manuscript 2: Age, renal function, level of consciousness, and serum pH were the important predictors of hospital mortality in critically ill elderly medical patients. Our clinical prediction tool is very good, particularly at the all-important extremes of prognosis, and ready for external validation. Results of Manuscript 3: Renal function and serum pH were the important predictors of hospital mortality in critically ill elderly surgical patients. Our model’s performance is very good, and will serve to inform clinical practice once validated. Conclusions: Very old medical patients have longer ICU stays and higher mortality than their surgical counterparts. Premorbid health status and severity of illness are associated with mortality. Our medical patient clinical prediction tool is very good and ready for external validation. Our surgical emergency clinical prediction tool shows promise, but will require the incorporation of more patients and a repeat derivation phase prior to external validation or clinical implementation.

International tourist hotel industry has becoming a professional management domain nowadays. Due to the increasingly fierce competition, hotels must develop ways to attract customers by meeting their market requirements and preferences, rather than waiting passively for the customers to come. With data mining technology, the hotels can facilitate the discovered characteristics of potential customers to make the right marketing strategies and decisions by targeting at specific groups of customers. Behind the consumers¡¦ behaviors, there are usually indicators for special consuming requirements. However, by browsing the business transaction data, one can usually learn only the consuming requirement volume and is unable to determine the implied and hidden information. This research makes use of data mining technology to explore the customers¡¦ historical data. Specifically, it applies the discovered decision rule to investigate and validate the characteristics of potentially customers ¡X customers who are more likely to book rooms of higher rate. We apply the data mining techniques to the transactional data of a hotel, collected over three years. Our research reveals that there exist characteristics rules for the potential customers and these rules do not change abruptly over the years. The application of these rules to target advertising in hotel domain is verified using the hotel transaction data collected in the subsequent year. The result shows that by targeting at customers of the discovered characteristics rules, higher response rate can be achieved.

Recently, distributed detection has been intensively studied. The prevailing model for distributed detection (DD) is a system involving both distributed local sensors and a fusion center. In a DD system, multiple sensors work collaboratively to distinguish between two or more hypotheses, e.g., the presence or absence of a target. In this thesis, the classical DD problem is reexamined in the context of wireless sensor network applications. For minimize the error probability at the fusion center, we consider the conventional method that designs the optimal binary local sensor decision rule in a channel-aware system, i.e., it integrates the transmission channel characteristics for find the optimal binary local sensor decision threshold to minimize the error probability at the fusion center. And there have different optimal local sensor decision thresholds for different channel state information. Because of optimal multi-bit (soft) local sensor decision is more practical than optimal binary local sensor decision. Allowing for multi-bit local sensor output, we also consider another conventional method that designs the optimal multi-bit (soft) local sensor decision rule in a channel-aware system. However, to design the optimal local sensor decision rule, both of two conventional methods are easily trapped into local optimal thresholds, which are depended on the pre-selected initialization values. To overcome this difficulty, we consider several modified Simulated Annealing (SA) algorithms. Based on these modified SA algorithms and two conventional methods, we propose two novel SA algorithms for implementing the optimal local sensor decision rule. Computer simulation results show that the employments of two novel SA algorithms can avoid trapping into local optimal thresholds in both optimal binary local sensor decision problem and optimal multi-bit local sensor decision problem. And two novel SA algorithms offer superior performance with lower search points compared to conventional SA algorithm.

The goal of this study was to derive a clinical prediction rule for transient ischemic attack (TIA) and non-disabling stroke to predict a treatment change based on echocardiogram. Methods: We conducted a cohort sub-study for TIA and non-disabling stroke patients collected over five years from 8 Emergency Departments. We compiled a list of 27 potential predictors to look for treatment change based on echocardiogram findings. We used a univariate, logistic regression and recursive partitioning analysis to develop the final prediction model. Results: The frequency of treatment change was seen in 87 (3.1%) of 2804 cases. The final model contains six predictors: age less than 50 years old, coronary artery disease history, history of heart failure, any language deficit, posterior circulation infarct and middle cerebral artery infarct on neuroimaging. Conclusions: We have developed a highly sensitive clinic prediction rule to guide in the use of echocardiogram in TIA and non-disabling stroke.

Objective: This study aims to derive a clinically-applicable decision rule to predict the risk of symptomatic vasospasm, a neurological deficit primarily due to abnormal narrowing of cerebral arteries supplying an attributable territory, in aneurysmal subarachnoid hemorrhage (SAH). Methods: SAH patients presenting from 2002 to 2011 were analyzed using logistic regression and recursive partitioning to identify clinical, radiological, and laboratory features that predict the occurrence of symptomatic vasospasm. Results: The incidence of symptomatic vasospasm was 21.0%. On multivariate logistic regression analysis, significant predictors of symptomatic vasospasm included age 40-59 years, high Modified Fisher Grade (Grades 3 and 4), and anterior circulation aneurysms. Conclusion: Development of symptomatic vasospasm can be reliably predicted using a clinical decision rule created by logistic regression. It exhibits increased accuracy over the Modified Fisher Grade alone and may serve as a useful clinical tool to individualize vasospasm risk once prospectively validated in other neurosurgical centres.

A transformative decision rule alters the representation of a decisionproblem, either by changing the sets of acts and states taken intoconsideration, or by modifying the probability or value assignments.Examples of decision rules belonging to this class are the principleof insufficient reason, Isaac Levi’s condition of E-admissibility, Luceand Raiffa’s merger of states-rule, and the de minimis principle. Inthis doctoral thesis transformative decision rules are analyzed froma foundational point of view, and applied to two decision theoreticalproblems: (i) How should a rational decision maker model a decisionproblem in a formal representation (‘problem specification’, ‘formaldescription’)? (ii) What role can transformative decision rules play inthe justification of the principle of maximizing expected utility?The thesis consists of a summary and seven papers. In Papers Iand II certain foundational issues concerning transformative decisionrules are investigated, and a number of formal properties of this classof rules are proved: convergence, iterativity, and permutability. InPaper III it is argued that there is in general no unique representationof a decision problem that is strictly better than all alternative representations.In Paper IV it is shown that the principle of maximizingexpected utility can be decomposed into a sequence of transformativedecision rules. A set of axioms is proposed that together justify theprinciple of maximizing expected utility. It is shown that the suggestedaxiomatization provides a resolution of Allais’ paradox that cannot beobtained by Savage-style, nor by von Neumann and Morgenstern-styleaxiomatizations. In Paper V the axiomatization from Paper IV is furtherelaborated, and compared to the axiomatizations proposed byvon Neumann and Morgenstern, and Savage. The main results in PaperVI are two impossibility theorems for catastrophe averse decisionrules, demonstrating that given a few reasonable desiderata for suchrules, there is no rule that can fulfill the proposed desiderata. In PaperVII transformative decision rules are applied to extreme risks, i.e.to a potential outcome of an act for which the probability is low, butwhose (negative) value is high. / <p>QC 20100622</p>

Soybean is an important source of protein in animal feed, and growing demand for meat consumption worldwide has led to increased soybean production. Over 120 million metric tons of soybean were harvested in the United States in 2018, approximately one-third of the world production. In the Mid-Atlantic region, soybean is one of the most valuable field crops. Major foliar diseases that reduce soybean yield in the Mid-Atlantic region are frogeye leaf spot (FLS) and Cercospora leaf blight. In addition to crop rotation and host resistance, foliar fungicides, often with quinone outside inhibitor (QoI) active ingredients, are used to manage these soybean foliar diseases. Yield benefits of foliar fungicides have been inconsistent and this may be the result of low disease pressure, unfavorable environmental conditions for disease development, or the presence of fungal pathogen populations that have developed resistance to fungicides. The objectives of this research were 1) to develop a pyrosequencing-based assay to rapidly quantify QoI resistance frequencies in Cercospora sojina, the causal agent of FLS, 2) to examine the effects of fungicide application timings, disease pressure, and environmental factors on soybean yield, and 3) to develop a weather-based soybean foliar fungicide application decision aid for the Mid-Atlantic U.S. using a threshold decision rule. A pyrosequencing assay targeting the G143A mutation was designed, and a Virginia survey of C. sojina populations indicated that the G143A mutation conferring QoI resistance is widespread. In small plot fungicide application timing experiments, five weekly fungicide applications starting at beginning pod (R3) resulted in the greatest yield, but for single fungicide applications, R3 or 1 week after R3 resulted in the greatest yields. There was positive relationship between the cumulative number of disease favorable days (mean daily temperature 20-30°C and ≥ 10 hours of relative humidity >90%) from planting to R3 and disease severity at the full pod stage (r = 0.97, P = <0.01). Higher disease severity was associated with greater yield loss (r2 =0.53, P = 0.10) suggesting foliar fungicide applications are more likely to have yield benefits as the number of disease favorable days prior to R3 increase. A disease favorable-days threshold (FDT) using the environmental parameters indicated above was evaluated in on-farm experiments throughout Virginia, Maryland, and Delaware. Based on decision rules, FDT = 8 three weeks prior to R3 was the best predictor of a yield benefit with an R3 fungicide application. The decision aid was also able to correctly predict when a fungicide application would not be profitable ≥90% of the time. This weather-based decision aid along with monitoring of fungicide resistance development within the region will provide soybean growers in the Mid-Atlantic U.S. with tools to maximize yields and profitability. / Doctor of Philosophy / Soybean is the third most valuable field crop in the world, ranked only behind rice and wheat in value. Over 98% of the soybean crop is used for animal feed due to its high protein content. The United States is the largest soybean producer in the world, responsible for one-third of global production. Soybean is the top cash crop in the Mid-Atlantic region. Foliar fungal diseases can reduce the soybean yield by causing lesions on the leaves that reduce photosynthesis and cause premature defoliation. Frogeye leaf spot (FLS) caused by Cercospora sojina is a major yield reducing soybean foliar diseases in the Mid-Atlantic region. Foliar fungicides, often with quinone outside inhibitor (QoI) active ingredients, are used to manage the disease. However, fungicide efficacy has been inconsistent. Inconsistencies may be due to low disease pressure, improper application timing, or fungicide resistance. The purpose of this research was to investigate the fungicide efficacy inconsistencies and to develop management tools to improve yield and maximize profitability. Our objectives were to 1) develop a molecular assay to quantify frequencies of the mutation conferring fungicide resistance in Virginia populations of C. sojina, 2) examine the effects of fungicide application timings, disease severity, and weather on soybean yield, and 3) develop a weather-based soybean foliar fungicide application decision aid for the Mid-Atlantic U.S. The C. sojina fungicide resistance mutation was widespread in Virginia, but overall frequencies were relatively low compared to findings from Midwest and Southern states. In fungicide timing experiments, beginning pod (R3) applications resulted in the most consistent yield benefits, and disease severity and yield loss increased as the number of weather-based disease favorable days prior to R3 increased. We used data from on-farm experiments in Virginia, Maryland, and Delaware to develop a weather-based disease favorable-days threshold that increased the probability that a fungicide application at R3 would have a yield benefit in soybean. The results of our research have led improved fungal disease management recommendations for soybean in the Mid-Atlantic that will maximize yields and profitability.

Detecting and tracking moving objects are important topics in computer vision research. Classical methods perform well in applications of steady cameras. However, these techniques are not suitable for the applications of moving cameras because the unconstrained nature of realistic environments and sudden camera movement makes cues to object positions rather fickle. A major difficulty is that every pixel moves and new background keeps showing up when a handheld or car-mounted camera moves. In this dissertation, a novel estimation method of camera motion parameters will be discussed first. Based on the estimated camera motion parameters, two detection algorithms are developed using Bayes' rule and belief propagation. Next, an MCMC-based feature-guided particle filtering method is presented to track detected moving objects. In addition, two detection algorithms without using camera motion parameters will be further discussed. These two approaches require no pre-defined class or model to be trained in advance. The experiment results will demonstrate robust detecting and tracking performance in object sizes and positions.

Many elderly patients visiting the emergency department for minor injuries are not assessed for functional status and experience functional decline 6 months post injury. Identifying such high-risk patients can allow for interventions to prevent or minimize adverse health outcomes including loss of independence. For the purpose of a planned clinical decision rule to identify elderly patients at high risk of functional decline a survey of physicians was conducted. A random sample of 534 Canadian geriatricians, emergency and family physicians was selected with half randomly selected to receive an incentive. A response rate of 57.0% was obtained with 90% of physicians considering a drop in function of at least 2 points on the 28-point OARS ADL scale as clinically significant. A sensitivity of 90% would meet or exceed 90% of physicians' requirements for a clinical decision rule to identify injured seniors at high risk of functional decline 6 months post injury.

Many elderly patients visiting the emergency department for minor injuries are not assessed for functional status and experience functional decline 6 months post injury. Identifying such high-risk patients can allow for interventions to prevent or minimize adverse health outcomes including loss of independence. For the purpose of a planned clinical decision rule to identify elderly patients at high risk of functional decline a survey of physicians was conducted. A random sample of 534 Canadian geriatricians, emergency and family physicians was selected with half randomly selected to receive an incentive. A response rate of 57.0% was obtained with 90% of physicians considering a drop in function of at least 2 points on the 28-point OARS ADL scale as clinically significant. A sensitivity of 90% would meet or exceed 90% of physicians' requirements for a clinical decision rule to identify injured seniors at high risk of functional decline 6 months post injury.