Measuring the Relationship between Hospital Costs and Quality of Care: An Example of Acute Myocardial Infarction in Edmonton, Alberta, Canada

This study explores the relationship between hospital costs and quality of care for Acute Myocardial Infarction (AMI) in the Edmonton area hospitals. The importance of this relationship is realized when policy makers face decisions about cost minimization and quality maximization during times of health care budget constraints. This study uses regression modelling with increasing specifications as well as various robustness checks to ensure the accuracy of the results. The Model specifications include demographics, AMI risk adjustments, Hospital fixed effects, and year fixed effects. Semi-parametric regression removes the assumption of linearity to determine the true relationship between hospital cost and AMI quality. Higher AMI quality is associated with a 39% increase in hospital costs after adjustments and controls. The semi-parametric regression shows a fairly linear relationship between cost and AMI quality. This study suggests that Canadian policy and decision makers should take caution during budget cuts and implementing cost containment programs. The results suggest that reducing AMI budgets may have a negative effect on the quality of AMI care patients receive in Edmonton, Alberta. The linear relationship suggests that the return on the quality of AMI is consistent for each dollar invested with no economies of scale.


Introduction
The relationship between quality of care and cost is essential to all policy makers. This study explores the relationship between hospital costs and quality of care for Acute Myocardial Infarction (AMI) in www.scholink.org/ojs/index.php/rhs Research in Health Science Vol. 2, No. 4, 2017 315 Published by SCHOLINK INC.
the Edmonton area hospitals. This relationship is increasingly important for policy makers who have strict budget constraints and make decisions about cost minimization and quality maximization. To our knowledge this is the first Canadian study on the cost-outcome tradeoff for AMI.
This study focuses on the quality and cost of AMI care for Edmonton, Alberta, Canada. There are several advantages in this well-defined population. The first reason is that AMI requires immediate medical attention, which removes problems with patient selections between hospitals. The second is that hospitals that provide better care can substantially improve the quality relating to AMI (Stargardt, Schreyögg, & Kondofersky, 2014). The third is that the quality of care can be measured by mortality in well-defined patient groups (Hä kkinen, Rosenqvist, Iversen, Rehnberg, & Seppä lä , 2015 Supplement).
Lastly, the existence of any possible relationship between quality and cost may differ between different heterogeneous sample groups such as the difference between cities, provinces, and countries, which may explain the inconsistent findings in existing literatures.

Semi-Parametric Analysis
If evidence of an association exists, then it is important for policy makers to know the true functional form of the relationship between quality and hospital costs. Semi-parametric regressions allowed us to relax the assumption of linearity from multi-linear regression analysis. Our model will resume the use of the linearity assumption on all parameters except the quality measurement as shown in equation 6, where F is an unknown function and the , coefficient remains linear. This function will be depicted in a graphical form to allow the interpretation of its true functional form. , = ( , ) + , + , + , + , + + ,  Note. *, **, *** indicates 1%, 5%, 10%, significance levels respectively.

Results
1 Based on patient mortality.

Robustness Check
Following the CIHI methodology for CMG+ cost estimation each patient's RIW was multiplied with the provincial CPWC from years 2006/2007 to 2008/2009. However due to changes in CIHI procedures these CPWC are no longer publicly available. This study will include these CPWC for future references (Note 2). To ensure further robustness of our results, this study replaces all micro-costs used in the previous analysis with the CMG+ cost estimates. The association under all specification were consistent with the previous findings when using the CMG+ cost estimates. All robustness check results are available upon request from the corresponding author.

Conclusion and Discussion
This study cautions Canadian policy and decision makers on budget cuts and cost containment programs relating to AMI. Our model finds evidence of a robust positive association between the level of AMI care and hospital cost. In other words, reducing hospital expenditure is associated with a decrease in AMI quality for Edmonton, Alberta. These results suggest that policy makers should take extreme caution when implementing any cost containment program as it may have a negative effect on patient health.
These results have undergone various robustness checks including increasing model specifications and replacing the micro-costing data with CMG+ cost estimates. These variations ensured the robustness of a positive association between the quality of AMI care and hospital costs. Similar positive association between AMI quality and hospital costs can be found in studies from California (Romley, Jena, &  Goldman, 2011), Germany (Stargardt, Schreyögg, & Kondofersky, 2014), Sweden (Hä kkinen et al., 2014), and United States Veterans hospitals (Schreyögg & Stargardt, 2010). Interestingly, this study's semi-parametric approach confirms a fairly linear relationship between quality and cost, which suggest that economics of scale and diminishing marginal returns may not be applicable. This means that the return on the quality of AMI is constant for each dollar invested.
Two major strengths of this study lie in the data set used. The first is that the data set is population based and not a sample. This data set contains all patients between fiscal years 2006 to 2009 who were admitted for AMI in Edmonton Alberta. The second, is that all costs came from the Alberta Health Services MIS which contains actual patient costs that remove the need for further estimation of costs.

Limitations
A major limitation in all AMI outcome-cost studies is the definition of quality being used. An ideal study would need to incorporate some true measure of AMI quality instead of the quality proxy. To our knowledge, there are no measures of the true quality of AMI and to derive such measurement would require the help of experts and physicians in AMI care. Other limitation includes that AMI treatment like PCI and CABG may be done after the initial hospitalization and in another hospital and increasingly also as outpatient operation. This may have impact to our cost and procedure outcomes of the study.
This study is restricted to the Edmonton area hospitals, which may reduce the variations between cost and quality indicators. A higher-level provincial study would be needed to provide more insight on the nature between hospital cost and AMI quality. This study also suffers from the inability to control for unobservable variable bias and reverse causality. This study has explored all previously proposed IV's as well as health area-based instrumentations (Chu & Ohinmaa, 2016), which were not possible to use since the majority of patients were from the same health region.