Healthcare constitutes one of the most important components in our lives. Disease or illness may mean a significant down turn in a personal life. Therefore the critical position of healthcare cannot be overemphasized. Healthcare is normally defined as the management or treatment of any health problem through the services that might be offered by, e.g., medical, nursing, and so on. The effect of the management or treatment is closely related to the detection of the disease. More specifically, determining when and/or where the disease presents as early as possible is usually conductive to an effective control and also to a better result of the management or treatment. Statistical techniques are such valid methods that can be used for the early detection in the healthcare industry....[ Read more ]

Healthcare constitutes one of the most important components in our lives. Disease or illness may mean a significant down turn in a personal life. Therefore the critical position of healthcare cannot be overemphasized. Healthcare is normally defined as the management or treatment of any health problem through the services that might be offered by, e.g., medical, nursing, and so on. The effect of the management or treatment is closely related to the detection of the disease. More specifically, determining when and/or where the disease presents as early as possible is usually conductive to an effective control and also to a better result of the management or treatment. Statistical techniques are such valid methods that can be used for the early detection in the healthcare industry.

According the objectives, disease detection usually can be classified into three categories: spatial, temporal and spatiotemporal detections. Spatial detection focuses on detecting specific locations within a large area which are subject to abnormal patterns, e.g., higher than average risks of the disease. For a given geographical area, temporal detection is of interest to signal, as early as possible, the presence of the disease. While spatiotemporal detection, as a more complex objective, investigates both when and where the abnormal patterns of the disease may appear. In this thesis, we are particularly interested in the spatial and the temporal disease detections, both of which have attracted sufficient attention from academics in the past few decades.

In studies of the spatial detection, scan statistics are widely applied to investigate healthcare related questions. However, to date, there is a scarcity of previous research on the scan statistics. They assume the independence of each population group which is divided based on, e.g., gender, age and race. Such an assumption ignores the natural interaction among multiple population groups and thus those derived statistical models may be less capable of disease detection. Compared with the spatial detection, the studies of the temporal detection are more mature. Among proposed techniques, we focus on the application of statistical process control (SPC) in purpose of deriving more powerful and applicable control charts for temporal disease detection.

This thesis seeks to make three major contributions to the area of disease detection. First, a new scan statistic is constructed for the spatial detection based on a multivariate lognormal model, which is able to describe the interaction among population groups. Second, a chart with probability control limits is proposed for the online monitoring of Poisson count data, which is the most common data type in healthcare related problems. In this study, we release the assumption of the population distributions and are able to guarantee the theoretical distribution for the in-control average run length in the proposed control chart. Third, a robust control chart is suggested for monitoring the covariance matrix of some medical indicators so as to detect any pathological deterioration expressed by the changes in the variance and/or covariance elements.

The three approaches strengthen the utility of statistical techniques in the healthcare industry. We have also performed numerical simulations to compare them with corresponding counterparts. In addition, real-data examples have demonstrated the effectiveness of these techniques. They can be implemented into practice with reliability.