Generally, high level of readmission is associated with poor patient care, hence, its relation to the quality of care is plausible. Readmission data are easily obtained from hospital databases and it appears more frequently than other adverse outcomes, such as mortality. Frequent patient readmissions have personal, financial and organisational consequences. This has motivated healthcare commissioners in England to use emergency readmission as an indicator in the performance rating framework. For the majority of studies, the definition of readmission varies considerably. The National Health Service (NHS) performance ratings framework defines readmission for adults as an emergency or unplanned admission to hospital within 28 days following discharge. Yet, in the literature, the definition varied according to the purpose of the study, generally from 30 to 90 days. Therefore, the definition of readmission and an appropriate framework for profiling hospitals play an important part in the study of readmissions. This thesis addresses these concerns by developing quantitative models focusing on readmitted patients at the national level. A continuous-time Markov model is used to capture the readmission process, where the optimal time window is computed using Bayes theorem and minimum classification error approach. Using national data (England), we demonstrate the usefulness of the approach in the case of chronic obstructive pulmonary disease, stroke, congestive heart failure and hip & thigh patients, which are known to be the leading causes of early readmission. Our findings suggest that there are marked differences in the optimal width of the time window for the selected clinical conditions. In the context of performance ratings framework, computing time windows for each clinical condition may not be practical, as there are hundreds of clinical conditions. NHS hospitals are rated using aggregated levels of readmission. To address this issue, patients length of stay in the community are clustered into several sub-groups. Using the estimated time window for each sub-group, we classify readmitted patients into 'high' and 'low' risk groups (binary response). The categorical response is further used in predictive modelling and in the profiling process. Identification of a method that provides an adequate predictive outcome is essential. Thus, the transition model is adopted, which allows the incorporation of the patients past history of readmissions. The effectiveness of this model enables the extension into a multilevel transition model, where individual hospitals propensity for first readmission, second readmission, third (and so on) are considered to be measures of performance. Using these measures we define a hospital performance index. From the 167 NHS acute and foundation Trusts in England, we found that the worst performing hospitals for cluster 2 patients were in London. Given the importance of patient readmissions in the NHS, the models and findings in this thesis should be of great interest to the department of health in other countries, healthcare commissioners in England, health and social services planners, and purchasers and providers of healthcare.