The adoption of big data analytics in healthcare applications is overwhelming not only because of the huge volume of data being analyzed, but also because of the heterogeneity and sensitivity of the data. Eective and ecient analysis and visualization of secure patient health records are needed to e.g., nd new trends in disease management, determining risk factors for diseases, and personalized medicine. In this paper, we propose a novel community cloud architecture to help clinicians and researchers to have easy/increased accessibility to data sets from multiple sources, while also ensuring security compliance of data providers is not compromised. Our cloud-based system design conguration with cloudlet principles ensures application performance has high-speed processing, and data analytics is suciently scalable while adhering to security standards (e.g., HIPAA, NIST). Through a case study, we show how our community cloud architecture can be implemented along with best practices in an ophthalmology case study which includes health big data (i.e., Health Facts database, I2B2, Millennium) hosted in a campus cloud infrastructure featuring virtual desktop thin-clients and relevant Data Classication Levels in storage.

No abstract

This chapter provides ad hoc tools to assist in estimating the non-compliance rate of biosecurity risk material (BRM) in a given pathway when certain data are unavailable. We use the inspection of international mail as a case study. Estimating the noncompliance rate of a pathway is essential in order to assess the risk of the environment, and to make defensible decisions about the allocation of inspection effort. Counts of articles inspected and articles found to have BRM are necessary for estimating the pathway non-compliance rate, and inspection counts by cohort (sub-pathway) are needed in order to perform profiling within a pathway, for example, identifying and prioritizing high-risk countries of origin for mail articles. Detailed information is usually kept on non-compliant mail articles that have been intercepted, but sometimes not on the articles that were inspected and passed, which is needed to report the inspection effort undertaken. Hence, the inspection effort by cohort may be unknown. In many pathways, endpoint surveys are performed after all inspection activity has been completed in order to estimate the efficacy of the inspection. We can estimate the number of mail articles of each cohort that undergo each type of inspection just using the endpoint survey. However, extra information is usually available, such as the totals by cohort or by inspection method, which could make the estimate of the number of mail articles more accurate. We present and demonstrate raking, which is a simple statistical tool that improves the estimates of cohort counts that arise from a survey using the known marginal totals of the process. The endpoint survey is also used for estimating the amount of undetected non-compliance, called leakage, by cohort, which is needed for profiling. However, leakage estimates by cohorts are usually variable because sample sizes are low. We demonstrate an empirical Bayes approach for reducing the variability of leakage estimates, and other ad hoc solutions. Finally, we suggest the use of Receiver Operating Characteristic (ROC) and leakage curves to assess the effect of these and other approaches upon the inspectorate's performance.