The Centers for Medicare and Medicaid (CMS) released the second year of quality performance measures in August 2015. CMS provided data on each of the 33 quality measures and two composite scores for 333 Medicare Shared Savings Program (MSSP) Accountable Care Organizations (ACOs). Among the ACOs, the majority (n= 241, 72 percent) did not generate savings.
We explored differences in characteristics and quality measures between all ACOs that did and did not generate savings and estimated the impact of those differences on the magnitude of savings reported. (See Exhibit 1 and Exhibit 3.) The most influential factor, accounting for almost one-third of the variance in savings generated, was each ACO’s per-member benchmark expenditure — the amount below which an ACO must reduce spending to be eligible for savings.
On average, those who received savings had a target expenditure of $915 more per member (Exhibit 1). ACOs that received savings had an average lower expenditure of $732 per member; those that did not receive savings had an average higher expenditure of $165 per member than their established benchmark. In other words, the two groups spent virtually an identical amount per member with savings determined by the baseline benchmark.
The second most important factor in determining variance in savings was patients’ rating of doctor (Exhibit 3). Perhaps, unexpectedly, this was inversely related to the amount of savings — higher physician ratings led to lower shared savings amounts. While we may speculate about the possible reasons—effective patient-doctor relationships likely include an element of mutual decision-making in which difficult conversations take place—the relationship between physician ratings and savings remains to be elucidated. The per-member benchmark and patient’s rating of doctor together account for 40 percent of the variance in savings.
There were no differences in the average number of beneficiaries assigned to ACOs who generated savings and those that did not (Exhibit 1). Three-quarters of ACOs with savings were in the second and third years of participation in MSSP. Similar findings were previously reported when examining the PY2 results for 86 successful ACOs.
Quality Scores
Differences were noted in the total quality scores between the two groups of ACOs, with 24 percent of ACOs receiving savings having a total score of 90 or greater. Exhibit 2 lists 13 of the 33 quality measures that showed a significant (p<0.05) or marginal (p >0.05 – p<0.10) difference between those ACOs according to whether or not savings were received. At least one measure in each of the four quality domains emerged as a statistically significant differentiator.
Among the eight measures under the preventive health domain, three were greater for those ACOs generating savings: adult weight screening and follow-up, tobacco use and intervention, and depression screening. However, it appears there is considerable room for improvement in these measures as the averages were below the 90th percentile for quality scores.
Half of the 12 measures for at-risk populations were higher for ACOs earning savings. These included lower LDL-cholesterol among diabetics, blood pressure control among hypertensives, complete lipid profiles and LDL control among those with ischemic vascular disease (IVD), and three measures for patients with coronary artery disease (CAD): drug therapy for lowering LDL-cholesterol, angiotensin-converting enzyme (ACE) inhibitor or angiotensin receptor blocker (ARB) therapy for patients with CAD and diabetes and/or left ventricular systolic dysfunction (LVSD), and the composite score for patients with CAD. For the last two measures, reporting only was required in participation years 1 and 2.
While most differences in quality rating between ACOs achieving savings and not doing so were in the direction one would anticipate, three were not. Health and functional status scores and the percentage qualifying for electronic health records (EHR) were higher for those without savings whereas admissions for ambulatory sensitive condition (ASC) of chronic obstructive pulmonary disease (COPD) or asthma in older adults were higher in ACOs that generated savings. One could argue most of these differences were small, and therefore likely to reflect spurious findings. Data in subsequent years should help determine if these findings persist.
Also worth noting is that one-fourth of all ACOs receiving savings (23 out of 92) were only required to report data in order to be eligible to receive savings. Furthermore, benchmarks for Health Status/Functional Status scores are not yet defined in any participation year.
Laying Out The Model
To further understand the relationship between ACO characteristics and quality measures for the 92 programs generating savings, we undertook a multi-variable approach using stepwise linear regression. Variables in the model included the 33 quality measures, the two composite measures, the number of assigned beneficiaries, and the per-member benchmark. Exhibit 3 shows the results of the final model, with eight variables selected for inclusion.
The first variable selected was the per-member benchmark, accounting for nearly one-third (31 percent) of the variance in the amount of savings generated. The remaining variables chosen were, in order: the patient’s rating of the doctor (inverse relationship), blood pressure control in diabetes, the total number of assigned beneficiaries (inverse relationship), health status/functional status (inverse relationship), LDL cholesterol control in diabetics, and the CAD composite score (inverse relationship). Together these eight factors accounted for almost 60 percent of the variance in the savings per member.
We also explored the relationship between amount of savings and the various factors in a sub-set of the ACOs with greater than 5,000 beneficiaries and who also successfully reported quality measures in performance years 2 and 3. The results were similar, with the same first four factors in the previous regression model accounting for 52 percent of the variance in savings; the greatest contributor was the per-member benchmark (partial R-square=0.3081).
Summary
The findings indicate the per-member benchmark is the strongest predictor of receiving savings and the amount of savings. But while success in savings to date is largely influenced by the established per-member benchmark, several quality measures are logically related to the magnitude of savings. Opportunities remain for improving patient outcomes. Additional time and experience in selecting quality metrics may be required to strengthen the relation measures of care quality and cost savings.
Exhibit 1: Comparison Of ACOs With And Without Savings
Exhibit 2: Quality Measures Differentiating MSSP Organizations With And Without Savings
Exhibit 3: Results Of Multivariable Linear Regression
Author’s Note
Dr. Egan has received research support from Medtronic and Quintiles, and income as a consultant from AstraZeneca and Medtronic.
from Health Affairs Blog http://ift.tt/1IkiR2n
No comments:
Post a Comment