257323 Applying a pharmacy-based metric in mortality prediction models for patients with acute myocardial infarction

Sunday, October 28, 2012

Raymond Kuo, PhD , Institute of Health Policy and Management, National Taiwan University, Taipei City, Taiwan
Kuo-Piao Chung, PhD , Institute of Health Policy and Management, College of Public Health, National Taiwan University, Taiwan, Taipei City, Taiwan
Mei-Shu Lai, MD, PhD , Graduate Institute of Epidemiology and Preventive Medicine, National Taiwan University, Taipei City, Taiwan
Background: Most mortality predictions models are based on the diagnoses derived from administrative data or electronic medical records. However, automated pharmacy claim data also has been used for measuring a patient's comorbidity or risk adjustment in healthcare utilization. This study aimed to verify that the Pharmacy-based Metric, a morbidity measurement based on prescription data, can be used for predicting in-hospital mortality and 30-day mortality for patients with Acute Myocardial Infarction (AMI).

Methods: The 2007 2009 National Health Insurance Database of Taiwan was used in this study. The information regarding the prescriptions includes prescriptions dispensed at outpatient clinics, in-hospital pharmacies, and at community pharmacies. The prescription codes from the claim data were first mapped to the WHO ATC codes, and then to the thirty-two classes of the Pharmacy-based Metric. Three diagnosis-based morbidity measures: The AHRQ Clinical Classifications Software (CCS), Deyo's Charlson Comorbidity Index (CCI) and Elixhauser's Index, were adopted as competitors to the Pharmacy-based Metric used in this study. Patient's age, gender and AMI location were also included in the prediction models. Multilevel logistic regression was applied to account for the clustering characteristics of the data. All models were fitted by the maximum likelihood method with Laplace approximation for a more robust parameter estimation.

Result: A total of 39 300 patients admitted with Acute Myocardial Infarction between 2007 and 2009 were selected for this study. The C-statistics for the Pharmacy-based Metric model predicting in-hospital mortality in 2007 to 2009 were 0.768, 0.762, and 0.751, respectively. The in-hospital mortality prediction model adjusted by the Pharmacy-based Metric performed better than the other three alternative models adjusted by CCS, CCI, and Elixhauser's Index (ranging from 0.745 to 0.763). The C-statistics for the Pharmacy-based Metric model predicting 30-day mortality in 2007 to 2009 were 0.762, 0.762, and 0.750, respectively. This model also provided a superior performance compared to the other three alternative models adjusted by CCS, CCI, and Elixhauser's Index.

Conclusion: The Pharmacy-based Metric was found to be valid for predicting the mortality for patients with Acute Myocardial Infarction. Automated pharmacy claim data is useful in developing risk adjustment models for mortality prediction. Based on the superior performance of the pharmacy-based models presented in this study, models that predict risk adjusted outcomes and are used for comparing hospital performance might consider incorporating morbidity measures based on prescription data.

Learning Areas:
Administration, management, leadership
Biostatistics, economics

Learning Objectives:
Develop prescription-based mortality prediction models for patients with Acute Myocardial Infarction (AMI).

Keywords: Drugs, Mortality

Presenting author's disclosure statement:

Qualified on the content I am responsible for because: I am the postdoc research fellow who did the analysis for this study and also the author who prepared the draft. My research interests include risk adjustment methodology for quality of care as well as healthcare utilization, comparative-effectiveness research, and health insurance payment design. The submission of this abstract and acting as the presenter has been authorized by the principal inspector and is agreed by all other co-authors.
Any relevant financial relationships? No

I agree to comply with the American Public Health Association Conflict of Interest and Commercial Support Guidelines, and to disclose to the participants any off-label or experimental uses of a commercial product or service discussed in my presentation.