Dissertation Title： "Hospital Quality and Patient Trust in Care for Colorectal Cancer"Quality of care has been of growing importance to U.S. health care researchers and policy makers. There is a widespread desire to improve medical outcomes and enhance the patient-physician relationship. This dissertation focuses on care for colorectal cancer, primarily using California Cancer Registry data, and presents the analysis of three prominent quality issues: trust in physicians, choice of hospitals for surgery, and hospital profiling.

In Paper One, I explored the impact of patients’ experiences during medical care on their trust in physicians using a survey of 1,602 colorectal cancer patients. I used logistic regression on trust scores to assess the association of specific health care domains with the likelihood of having low trust, controlling for patient sociodemographic and clinical covariates. During medical care experiences, problems in access, patient education and particularly coordination and continuity of care seriously affected patients’ trust in their health care providers, suggesting that efforts to improve these aspects of medical services may enhance patients’ trust.

In Paper Two, I studied the effect of patient characteristics on hospital choice. Using data on 38,237 patients, associations of patient attributes with surgical care at high-volume hospitals, teaching hospitals, and hospitals with different level of adjusted mortality were assessed with multivariate regression analysis. Patients with more severe clinical status, of minority ethnicity, and of less privileged socioeconomic status were significantly less likely to be admitted to the high-quality institutions for cancer surgery, suggesting the potential benefit of improving access for patient outcomes.

In Paper Three, I evaluated the utility of cancer registry data for profiling of hospitals by survival of colorectal cancer patients. With 5 years of data from California supplemented with hospital discharge abstracts, I fitted univariate and bivariate hierarchical models with hospital random effects and used a Bayesian algorithm to provide valid estimates of hospital profiles for survival. For larger hospitals, adjusted 30-day survival rate can be estimated with adequate precision to support useful and valid comparisons of hospital quality, especially with additional information from the bivariate-outcome model, suggesting that the already-collected cancer registry data can find a new use for profiling cancer care providers.