William O. Cleverley
James O. Cleverley
Value-based purchasing (VBP) is on a path to becoming the new reality for the future of U.S. health care. But will hospitals be able to manage the potential costs associated with the improved quality that VBP will require?
At a Glance
- An analysis of the relationship between quality and cost using a sample of 3,081 Medicare-reimbursed acute care hospitals found that cost appears to increase somewhat with quality improvement.
- However, an analysis that focused on the top 10 Medicare severity-adjusted DRGs (MS-DRGs) in terms of volume, based on national statistics, found that improving quality can help to reduce costs.
- An examination of mortality associated with a single MS-DRG found that cost increases steadily as mortality rates increase, supporting a conclusion that poor outcomes lead to greater expenditures.
With the signing of the Affordable Care Act, the Centers for Medicare & Medicaid Services (CMS) has been set well on the path to implementing value-based purchasing (VBP) into Medicare's payment structure. CMS has long viewed VBP as an important long-range solution for Medicare's fiscal challenges. A statement from the CMS Report to Congress: Plan to Implement a Medicare Hospital Value-Based Purchasing Program (Nov. 21, 2007) says, "Value-based purchasing (VBP), which links payment to performance, is a key policy mechanism that CMS is proposing to transform Medicare from a passive payer of claims to an active purchaser of care."
As the VBP demonstrations outlined in the reform legislation unfold, value is to be determined through performance in quality dimensions, and those dimensions will be assessed from available quality metrics that currently exist in the Reporting Hospital Quality Data for Annual Payment Update (RHQDAPU) program commonly referred to as Hospital Compare.
Under the VBP demonstrations, hospitals are to be rewarded or penalized by Medicare based upon their relative performance to quality metrics. Those hospitals with overall performance higher than some percentile ranking of all hospitals might receive a bonus payment, while those with scores below a percentile ranking might be penalized. There is also discussion for providing some additional weight to hospitals that have made significant improvements in the quality metric scores. Although the exact details of the VBP system are as yet uncertain, full implementation of VBP seems inevitable. It also seems highly likely that VBP systems will be introduced into many commercial payer programs-most likely on a basis that will mirror the Medicare program.
No one will argue the importance of quality, especially in health care, but the critical financial question is, what is the cost of added quality? If a hospital can take actions that will improve its quality score and realize an additional $1 million in payment, should it do so if it costs $2 million to realize those improvements?
Some argue that quality improvements will actually reduce costs. However, quality enhancements often require additional resources once production efficiency has been reached. Simple economics will then dictate that improvements in quality are not possible if they are not compensated at levels commensurate with required cost increases. This economic reality runs counter to the public's perspective, which demands quality improvements regardless of cost. Ultimately, however, hospitals need to ensure that high-quality services are provided and that long-term financial viability is not sacrificed.
Analytical evidence about the relationship of cost to the quality metrics used in the Hospital Compare database can provide insight regarding these concerns. It is important to recognize that using this evidence requires the assumption that these quality metrics do in fact measure quality, which may not be the case, despite CMS's decision to use them in its VBP program. It is therefore important to assess the relationship of cost to the current Hospital Compare metrics.
The Hospital Compare Database and Hospital Quality Index™
The Hospital Compare metrics are the result of voluntary data submission by participating hospitals and patients. Over the past several years, the number of hospitals reporting the metrics has increased as the quality of the data submitted has improved. At the time of this analysis, the Hospital Compare database contained information in the following areas:
- Process-of-care measures-indicators of how well a hospital provides care consistent with known best practice in five areas (heart attack, heart failure, pneumonia, surgical care, and child asthma)
- Outcome-of-care measures-indicators of 30-day risk-adjusted mortality and readmission rates
- Patient survey data-information on patient experience from patient surveys
- Medicare payment and volume-information on payment and the number of discharges by diagnosis-related group (DRG) for each hospital
To examine the relationship of quality to cost, a measure of composite quality performance is required. The Hospital Quality Index™ (HQI) was developed for this purpose using Hospital Compare data. The HQI evaluates hospital performance from only the process-of-care and outcome-of-care areas.
For this analysis, we examined 25 process-of-care metrics over the period of April 2008 through March 2009 in the areas of heart attack, heart failure, pneumonia, and surgical infection prevention. These process-of-care areas refer to medical standards for treatment protocols, such as giving patients experiencing heart attack aspirin upon arrival in the emergency department. Hospitals report the percentage of time they met these standards in each of the 25 areas. From these data, it was possible to determine each hospital's percentage above or below the U.S. average, as well as how often the hospital performed at or above the highest performing hospitals in the country. Hospitals received high process-of-care composite scores in the HQI when a higher number of areas were reported and when performance in those areas exceeded both the U.S. average and high-performance levels.
We measured outcome quality for the period of July 2005 through June 2008 using risk-adjusted mortality and readmission rates established for each facility by Medicare. These rates are provided for hospitals in three areas: heart attack, heart failure, and pneumonia patients. The mortality rates estimate the risk-adjusted frequency of death within 30 days of patient discharge. From the data in these areas, we created a composite score to evaluate the percentage by which each hospital was above or below U.S. average levels. Hospitals that had lower levels of mortality and readmission had better composite scores.
The final step to create the HQI was to combine the composite scores for the process-of-care and outcome-of-care areas, as described above, so that the HQI could serve as an overall quality score for each hospital.
Data Analysis-Cost and the HQI
We tested the relationship between quality and cost using a sample of 3,081 acute care hospitals that were reimbursed by Medicare under the prospective payment system (PPS). All non-PPS paid hospitals were excluded from the analysis, except that Maryland hospitals were included. Data used for the study were for 2008-the last complete year of available data at the time of
The exhibit below presents a summary of the quality-cost relationships. It shows the impact of increasing quality on a number of other independent variables. Hospitals are divided into four quartiles based on their HQI score. For example, U.S. hospitals in the lowest quartile in terms of HQI values had a median HQI of 89.2. Conversely, U.S. hospitals in the highest quartile had a median HQI of 102.5.
For each quartile, we computed the median values for a number of additional metrics. The first variable tested was Medicare cost per discharge (case mix and cost-of-living adjusted). The cost-per-discharge metric makes it possible to assess the relationship between quality and cost. We used an inpatient measure of cost because the HQI is currently composed only of inpatient metrics. The data do suggest a positive relationship between cost and quality: Lower quality is related to lower cost. Moreover, as quality increases, cost also increases, but not by the same rate of change. For example, the rate of cost increase between the "lowest HQI" quartile and the "low HQI" quartile was 1.9 percent, whereas the rate of cost increase from the "high HQI" quartile to the "highest HQI" quartile was 1.1 percent. The bottom-line finding is that cost does appear to increase with quality improvement as measured by the HQI metric, but the relationship is not strong.
We also reviewed the relationship between hospital size, as measured by net patient revenues, and HQI. The relationship between size and quality appears to be much stronger than the relationship between cost and quality. Larger hospitals appear to have significantly higher HQI scores than smaller hospitals. This relationship is consistent with findings of prior studies on quality that have detected an association between higher volume and better outcomes. The median net patient revenue for hospitals in the highest quality quartile was $174.9 million, whereas the median net patient revenue for hospitals in the lowest HQI quartile was only $53.8 million.
The third relationship tested was quality and profitability. Some in the hospital industry have argued that only profitable hospitals can afford quality enhancements. Literature in other industries supports the notion that providers of higher-quality services or products are likely to be more profitable. The data described above suggest that operating margins do increase as quality scores rise. It was not possible to determine whether the higher quality scores led to greater profitability, or whether higher profits permitted quality enhancements. The median operating margin for hospitals in the highest quality quartile was 2.2 percent compared with 0.1 percent for those in the lowest quality quartile.
The above data analysis suggests that three factors- size, profit, and cost-are related to the HQI metric. To verify these relationships, we used a multivariate regression analysis with the HQI metric as the dependent variable and net patient revenue, operating margin, and cost per discharge as the independent variables. This analysis corroborates the finding that size, as measured by net patient revenue, and to a lesser extent margin are significant factors, but it also suggests that cost per discharge is not. In short, volume appears to be the biggest key determinant for the HQI metric. Cost has a much weaker relationship.
Data Analysis-Cost and Mortality at the MS-DRG Level
Because volume was so positively correlated with the HQI, we decided to test the relationship of volume and quality at the level of specific Medicare severity-adjusted DRGs (MS-DRGs). We used discharge mortality rates as the measure of quality and the number of cases discharged as the measure of volume. It is important to note that discharge mortality rates are outcome measures and are not process measures. The earlier results based on the HQI measure included both outcome- and process-based measures. We used mortality rates in this portion of the analysis because the data are readily available and because they are widely regarded as the ultimate test of quality.
We also used the estimated cost for the specific MS-DRG for that hospital adjusted for cost of living. Data were taken from the 2008 Medicare Provider Analysis and Review (MedPAR) file, which contains data for all Medicare inpatient claims. Because the MS-DRG categorization system adjusts for severity, the mortality rates across identical MS-DRGs should be comparable. We chose the top 10 MS-DRGs in terms of volume, based on national statistics. Of these MS-DRGs, only MS-DRG 470 (Major joint replacement or reattachment of lower extremity without major complications or comorbidities [MCC]) pertains to surgical cases. The other nine MS-DRGs pertain to medical cases. Only hospitals with volumes greater than 20 for that specific MS-DRG were included, which removed hospitals with very low volume where extreme mortality rates-either very high or very low-were recorded due to low volumes. The 10 MS-DRGs are listed in the exhibit below.
The exhibit below summarizes the results of the regression analysis. Each of the 10 regression equations was stated in the following format:
Mortality Percentage = Intercept + (B1 x Number of Discharges) + (B2 x Cost per Discharge Cost of Living Adjustment)
The first relationship tested was increased volume (number of discharges) to decreased mortality. The coefficient was used to determine this relationship. In each of the 10 MS-DRGs, we found that mortality improved with higher volume of cases. This finding is consistent with the previous HQI relationship with increased quality by increased size. However, although the finding is consistent for all 10 MS-DRGs, the relationship is only significant at the 5 percent level in six of the 10 cases.
Next, we examined the relationship of mortality and cost per case. Again, the coefficient was used to test whether higher costs were associated with lower mortality. In nine of the 10 cases, higher mortality rates were found to be associated with higher costs. Statistical significance of this relationship at the 5 percent level was achieved in five of the 10 cases. This finding is, perhaps, more striking in that it suggests there is both a quality and cost incentive to keep mortality low. In short, hospitals with poor outcomes (in this case, higher mortality) are more likely to have higher costs. This finding validates the perception of many policy analysts: Improving quality can, in fact, help to reduce costs to create a win-win scenario. By contrast, our initial analysis, which measured quality on a combination of both process and outcome measures, found that higher quality was associated with higher cost, which contradicts the expectation of most policy analysts.
Data Analysis-MS-DRG 291 Heart Failure and Shock with MCC
We next examined some specific relationships for one of the MS-DRG groups, MS-DRG 291-Heart Failure and Shock with MCC. The exhibit below shows the relationships for the four quality quartiles that are based on mortality percentage. Although the statistical relationship between volume, as measured by the number of MS-DRG 291 cases, and mortality was significant, the pattern in the exhibit is not clear cut. Volume is the lowest for the highest mortality group. However, the actual trend does not appear strong when presented in this tabular format. The quartile with the lowest mortality rate had a median of 61 cases, compared with 77 and 74 for the two quartile groups with the next highest levels of mortality rates.
The relationship between cost and mortality is also displayed in the exhibit. Cost increases steadily as mortality rates increase. To be clear, the expenditure of more money does not produce poor outcomes-namely higher mortality; rather, poor outcomes cause a greater expenditure of money. The relationship suggests that hospitals with the highest mortality rates will spend about $1,100 more per case for this MS-DRG.
The third variable included in the exhibit below reflects the testing of a coding relationship.
MS-DRG 291 is a member of the "Heart Failure and Shock" MS-DRG family, which also includes MS-DRG 292 (with CC) and MS-DRG 293 (without CC). Of the three MS-DRGs, no. 291 has the highest severity, the highest weight, and greatest level of reimbursement under Medicare. The test was to determine whether hospitals in all four quality groups had the same distribution of patients in each of the three MS-DRGs. However, the data in the exhibit disclose that hospitals that have a larger percentage of patients assigned to the MS-DRG reflecting the highest severity-that is, no. 291-also have the lowest mortality rates. The group exhibiting the highest quality, with a mortality percentage of 2.27 percent, had nearly 41 percent of all patients assigned to MS-DRG 291, compared with 35 percent assignment for the group demonstrating the lowest quality.
What is not clear is whether this finding reflects true differences in patient severity or differences in the aggressiveness of the hospitals' coding. Hospitals that "under-code" by placing more patients in the "lowest-weighted" MS-DRG could show higher rates of mortality because their MS-DRG 291 cohort would not include as many cases with truly lower severity. Conversely, hospitals that aggressively code could have some cases with lower severity in MS-DRG 291, which might bring their mortality rate down. Whatever the underlying cause, the relationship between coding and mortality appears clear and is statistically significant: The lowest mortality rates are seen in hospitals with a greater percentage of patients in the highest severity MS-DRG.
A Win-Win Situation?
Our primary purpose was to assess the relationship between quality and cost. Specifically, is there an increase to cost that is associated with an increase in quality? This question is especially critical should Medicare, and perhaps other payers, begin a more comprehensive VBP program. Using the HQI, which is a blend of the measures most likely to be used by Medicare in any VBP program, we found that cost does in fact increase as quality scores improve. This finding implies that hospitals might face a trade-off between increased payments resulting from higher quality scores and increased costs necessary to reach those levels. However, the relationship between cost and HQI scores was not statistically significant, suggesting that the requirements of the VBP may not actually raise costs.
On further testing the relationship between mortality and cost in 10 MS-DRGs, we found that higher mortality rates were often associated with higher costs. The implication here is that poor quality is often associated with higher costs, which would imply that a VBP program related to an outcome measure such as mortality might be a win-win scenario for both payers and hospitals: Improvements in quality would actually lead to lower costs. The strength of this conclusion is diluted somewhat, however, by the fact that statistical significance was found with only five of the 10 MS-DRGs.
Paying for better quality is clearly a reasonable contract feature for both payers and hospitals. The major problem, however, appears to be the measurement of quality. If the metrics used do not lead to actual improvement in health outcomes, then the system fails the patient. Further, if the incentives do not cover the costs to achieve better outcomes, then hospitals will face a dilemma in deciding whether to pursue the quality initiative at the risk of financial distress.
William O. Cleverley, PhD, is president, Cleverley + Associates, Worthington, Ohio, and a member of HFMA's Central Ohio Chapter (firstname.lastname@example.org).
James O. Cleverley is principal, Cleverley + Associates, Worthington, Ohio, and a member of HFMA's Central Ohio Chapter.
Publication Date: Monday, January 03, 2011