Text

theses

Advances in information acquisition techniques and the widespread use of information technologies in healthcare services has resulted in an incredible opportunity for health administrators to utilize data analytics and models to provide better health care,manage risks and improve patient outcomes. In this study two different data analytical models - a Logic Model and a Predictive Model – were formulated using data sets obtained from the Health Cost Utilization Project (HCUP) and the Recovery Audit Contractor (RAC) reports for use in Medicare and Medicaid patient hospitalization outcomes research. Objectives: The overall goal of the study was to (1) to design an appropriate analytical model to explain the operations of the RAC process and identify the hospitalization factors that affect the efficient recovery of claims (2) to formulate a predictive model by using HCUP’s Nationwide Inpatient Sample datasets to help predict those hospitalization factors above affecting the RAC claims recovery process, and (3) to determine other relevant hospital, regional and patient related variables that play a statistically significant role in both the RAC and the Hospitalization Outcomes Models. Methods: To meet the aforementioned objectives data was extracted from both the RACTrac Website and Reports (for developing the RAC Process Model) and the HCUP Nationwide Inpatient Sample (NIS) database. Several analytical models currently in vogue in both health and finance were investigated and it was decided to adopt a Logic Model to describe the RAC claims recovery process and with its help identified the hospitalization factors related to the claims and payment issues. Secondly the Multiple Linear Regression Model was found to be the most suitable predictive model type for the 6 hospitalization factors identified from the RAC Logic Model. Lastly several descriptive and inferential statistics were employed to infer relationships among several patient and hospital variables with the RAC regions and their outcomes. Results: Both Length of Stay (LOS) and Total Charges were found to be intimately related to the RAC claims recovery process and accordingly they both were employed in the development of the Multiple Linear Regression Model with several independent variables such as DRG, RAC region, Payer type (Medicare, Medicaid, Private), Number of Diagnoses and Number of Procedures resulted in a reasonably good fit (54 % to 59 %) of the model in explaining the variance of the outcome of Total Charges and not a very good fit for the LOS which was expected since LOS is not a linear variable and subject to too many constraints and hence not easily predictable. The ANOVA Tests revealed several interesting relationships between the independent variables listed above and the RAC regions with implications of import for the RAC claims recovery process. Conclusion: This study is significant because it demonstrates the validity of the use of analytical models such as Logic Model and the Multiple Linear Regression Model in predicting Hospitalization Outcomes of interest to not only the RAC claims recovery process relevant to this study but also in other health administrative settings involving planning of budget and resource allocation. The complex process of RAC claims recovery mechanism has been duly modeled by the Logic Model technique thus making it available for future configuration modification and studies into their effect on theclaims recovery process.

ETD_8298

Ph.D.

Includes bibliographical references

by Kathleen M. Kirk

doi:10.7282/T3FF3WDW

ETD doctoral

The author owns the copyright to this work.