First a summary and then the disease management blog will weigh in:
The demo tested “Pay for Performance” among 10 large (at least 200 docs) physician groups. The “performance” used by these groups had everything to do with the approaches used in classic disease management. It included patient education, post-discharge programs, medication reviews, case management, interactive voice response and care coordination. Most of the groups also had an electronic health record.
The “pay” was used to either 1) develop or 2) expand preexisting in-house DM programs. The number of programs per physician group ranged 2 to 9. The pay was also a “gamble”: the groups had to front much of the money in hope of qualifying for the payment bonus at a later date
In order to get the pay, the 10 organizations had to perform by 1) reducing overall spending among its assigned Medicare beneficiaries by more than 2% (making this a gainshare) and 2) achieve improvements in 10 NCQA-pedigreed diabetes clinical process measures. The comparator was a group of Medicare beneficiaries that were retrospectively assigned.
Most of the groups achieved most of the diabetes improvements in the first year, but the only ones that got a gainshare were the Marshfield Clinic and the University of Michigan Faculty Group Practice (calculated at $4.6 M and $2.8M, respectively). One other unnamed group achieved 2% savings but didn’t exceed it (sucks to be you). Marshfield and Michigan also seemed to benefit by being measured against comparator beneficiaries that experienced very high Medicare spending.
Marshfield had only two programs: 1. telephony for heart failure and 2. anticoagulation management. The University of Michigan also had only two programs. They were monitoring and self-care education by care managers for 1. frail patients with many co-morbidities and 2. those with a recent inpatient stay.
In contrast, the average number of clinical programs for all of the physician groups was between 4 and 5 (Billings Clinic led the way with 9 of them) with total program costs among the participants that ranged from just over $400,000 to as high as $2.9 million. Interestingly Michigan had the lowest total program costs and Marshfield had the highest costs.
In their report, GAO noted the approach was “reasonable” but disappointing. It urged caution in interpreting the data because they are from the 1st of a 3 year process. In addition, only one participant had all of its programs up and running for the first 12 months of the demo.
To make matters worse, there were delays in getting claims data from CMS back to the physician groups. It was also hard for the physician groups to interpret the data dumps once they arrived. As a result, the participants were “flying blind” without the kind of feedback necessary to “tweak” their programs. GAO has recommended this be fixed, which may make a difference in the future.
The Disease Management Care Blog weighs in:
Even the GAO noted that large physician groups account for less than 1% of all physician care in the U.S., making the lessons learned here practically inapplicable to the other 99% of health care delivery in the U. S.. Smaller physician groups will never be able to implement any of the approaches used in the PGP demo because they don’t have access to the necessary capital, have yet to implement an EHR and have little experience with P4P programs. Last but not least, if large physician organizations are as efficient as they say they are, the outcomes to date gives them every incentive to avoid such schemes. Based on the results from this demo, they can expect to spend more than they gainshare in the first fiscal year.
And now for something completely different:
1) Given the random distribution of claims expense, the disease management blog wonders if the data spread in the above graph has nothing to do with the interventions hatched by the participants. I can’t see any correlation between the intensity of the interventions, the money invested and the savings that were achieved. Toss a bunch of numbers in the air and some will come back lower or higher. Based on the report, Michigan and Marshfield may have been lucky, not smart.
2) One reason for the lack of performance could be because large physician groups are fundamentally postured to seize every dollar of revenue in a FFS environment. One fat DRG beats 5 well-intentioned case managers.
3) This wasn’t disease management. It appears to me that the best interpretation is that these physician groups used programs that met the definition of the Chronic Care Model. Just like the COACH heart failure study, the programs for each of these groups were under the leadership of and had proximity to physicians, used patient self management support interventions, used computerized clinical data bases/registries/EHR, relied on teaming, used clinical guidelines and had supportive reimbursement. I am very confident that if we had asked, the leaders at each of the participating physician group practices would have denied that their interventions were "disease management," and that what they were offering was distinctly different.
Let’s hope future outcomes in this demo show that CCM is not a bust. Or is inconsistent with other negative disappointing literature on the lack of cost savings in CCM.