Since 2005, Medicare began the Reporting Hospital Quality Data for Annual Payment Update (with the incredibly unintelligible acronym of RHQDAPU). RHQDAPU at first just required hospitals to report quality measures. The Health Reform VBP initiatives, however, will begin to pay hospitals based on their performance on these metrics. The 2007 CMS report claims that any VBP plan should contain the following 7 components.
A Performance Assessment Model that is used to score a hospital’s performance on a specified set of measures, generating a Total Performance Score for each hospital.
Translation of the VBP Total Performance Score into an incentive payment.
A measure development process, including selection criteria for choosing performance measures for the financial incentive, and candidate measures for VBP Program start.
A phased approach to transition from RHQDAPU to VBP.
Redesigned data submission and validation infrastructure to support the VBP Program requirements.
Enhancements to the Hospital Compare website to support expanded public reporting of performance results.
An approach to monitoring VBP impacts, including potential impacts on health disparities.
Below I discuss aspects of hospital VBP in more detail.
There are a large variety of measures that Medicare could use to evaluate hospitals. These include outcome measures (e.g., mortality), process measures (e.g., administering beta blockers after a heart attack), structural measures (e.g., nurse/bed ratio) and patient satisfaction measures ( HCAHPS ). Ideally, Medicare could use outcomes measures exclusively. However, risk adjustment methods are never perfect so physicians who treat patients with unobservable (from a data point of view) adverse health factors will be unfairly hurt by VBP. In addition, measuring “health” as a outcome is difficult. Mortality is easy to measure, but most patients want more from their treatments than just avoiding death; quality of life is also important.
How does Medicare know that the reported quality measures are accurate? One means to accomplish this is through data validation. Medicare can audit charts to make sure the care the hospital claims to have provided was actually received by the patients. One could randomly select hospitals to receive these audits every quarter. A more efficient means would stratify hospitals into high, medium and low risk. Medicare should sample a larger share of hospitals in the high-risk strata than the low-risk strata.
Hospitals may also enter incorrect data on their quality submissions or claims data. Medicare may need to allow time for corrections or appeals so that the information Medicare receives is correct.
Attainment vs. Improvement
Which hospitals should receive the biggest bonus: those who attain the highest quality marks or those who improve the most. By ignoring improvement, hospitals in the bottom quality score percentiles may have little incentive to improve patient care since it is unlikely they will ever make it to the top scoring rungs. On the other hand, relying only on improvement to rate hospitals is unfair to hospitals who exhibit long-term clinical excellence.
To account for these trade-offs, the authors calculate both an attainment score. Both scores are rated on a scale from 0-10. Hospitals much reach a minimum quality threshold to receive an attainment score above 0. Additionally, there is a quality benchmark above which all hospitals receive an attainment score of 10. In between the attainment threshold and benchmark, hospitals can score anywhere between 0 and 10. A nonlinear function is used to calculate and exact score and then this figure is rounded. Hospitals whose score improves from the prior year can receive a positive improvement score. More improvement leads to a better score.
As both measures are ranked from 0-10, hospitals end up receiving the higher of the attainment or improvement score.
Although Medicare can calculate the attainment and improvement scores for each quality metric, the agency must use a single metric to determine how the hospital’s overall performance will affect their reimbursement rates. The question is how to weight each quality measure. The simplest approach is to just average the scores across all the metrics. Some quality measures, however, may be more important than others. In this case, Medicare may wish to consult with clinical experts to construct a weighting scheme to compute a single metric upon which to base payment.
Additionally, not all hospitals will be scored in all measures. For example, “some hospitals may not perform percutaneous coronary intervention; therefore, this measure would not apply to them.” Thus, the weighting will apply to only quality measures with which the hospital is eligible. Further, Medicare may only want to evaluate hospitals who reach a certain minimum number of observations for each quality measure to ensure an adequate sample size.
VBP is supposed to save money. If Medicare is going to pay hospitals bonuses, this money will have to come from somewhere. The CMS report offers the idea that 5% of the base DRG payments could be used as VBP. In essence, this means that the worst hospitals will see a 5% decrease in revenue. The best hospitals, however, will not maintain the same reimbursement levels. The report suggests using 80% of these funds for VBP, but keeping 20% as cost savings to Medicare. Thus, even the best hospitals will receive 1% lower payments under VBP.
The authors note that the base DRG payments “would include geographic and DRG relative weight adjustments, but payments for capital costs, IME (indirect medical education), and DSH (disproportionate share hospital) would not be part of the basis of the incentive payment.”
The goal of VBP is to improve quality, but implementing a VBP program can have unintended consequences. These include:
Hospitals can “teaching to the test” by focusing their efforts on improving quality in measured area by sacrificing quality in unmeasured dimensions.
Hospitals may drop or avoid caring for patients who are sicker or more difficult to manage, as observed through reduced access to care and increases in transfers.
It is possible that VBP may increase disparities in care by region, race/ethnicity, or other factors. For instance, physicians may decide not to care for immigrants for whom English is a second language if their drug adherence rates are lower, thus causing lower VBP scores for the physician.
Hospitals can game the data by reporting care that was not provided in order to secure incentives.
Hospitals may undertake actions to shift un-reimbursed costs to other payers.
A study by Chen et al. (2010) gives just such an example. Taiwan’s National Health Insurance ( NHI ) instituted a voluntary diabetes mellitus pay for performance (DM-P4P) program in 2001. In its first incarnation before the end of 2006, the DM-P4P program provided phsyicians with rewards for process (e.g., hemoglobin A1C testing). In the second period starting at the end of 2006, the NHI began to pay extra bonuses based on intermediate outcomes measures (e.g., percent of patients with A1C≥9.5%, percentage of patients with LDL≥130 mg/dl). Using a logistical model, the authors find that “older patients and patients with more comorbidities or more severe conditions are prone to be excluded from P4P programs.” Thus, the patients who would benefit most from the P4P program (i.e., the sickest patients) are precisely the ones excluded from the P4P program. Based on their analysis, the authors conclude that patient participation in the P4P program should be mandatory.
Even if participation in P4P is mandatory, however, unethical providers could still select patients to participate in P4P. For example, only diabetic patients are eligible for the P4P A1C measure. Thus, if the physician does not classify the individual as diabetic, they will not receive the A1C test. Ensuring the correct individuals participate in the P4P program is far from simple.