Investigating the Autocorrelation in Surgical Data with Ordinal Outcomes

Performance monitoring of care providers is naturally of great importance. As in production environment, by use of well-defined quality characteristics, we can detect any deviation from the desired performance, which can be helpful in avoiding unintended consequences as well as in providing valuable information for quality improvement of the care delivery system. However, unlike the manufacturing processes in which materials are relatively homogeneous, there is a considerable heterogeneity in the characteristics of patients. To have a fair judgment about the performance, the heterogeneity of patients has to be considered. Risk-adjusted control charts account for this heterogeneity. These widely used control charts have been developed on the assumption of independent observations .However ,some empirical evidence suggests that the data related to some of the surgeons have some type of dependence. In this paper, we investigated this for surgical data when they have multinomial ordinal outcomes .The results showed that the autocorrelation is significant.

کلیدواژه ها:

Healthcare ، Risk-Adjusted Control Charts ، Parsonnet Score ، Ordinal Logistic Regression ، Proportional odds ، Autocorrelation.