Pay for performance: adding to healthcare disparities

a large Medicare study assessed the utility of the "value-based payment modifier", aquality of care incentive, finding that there was no improvement in care, but actually exacerbating health care disparities (see doi:10.7326/M17-1740):

​Details:

--history:

    --jan 2017: Medicare implemented Merit-based Incentive Payment System (MIPS) for fee-for-service Medicare patients, a pay-for-performance program to reward clinicians who improved quality of care and reduced spending, its effects are pending

        --but from 2013-16, there was similar system, the Value-Based Payment Modifier (VM):

            --2013: practices with >100 clinicians were required to meet reporting requirements or incur small reduction in 2015 payment, but performance-based payment adjustments were optional

            ​--2014: VM mandatory for all practices with >10 clinicians (excluding those in ACOs or other alternative payment options).

                ​-- In practices with >100 clinicians: pay-for-performance payments could be more, neutral or less, depending on performance

                ​--10-100 clinicians: pay-for-performance payments could only be more or neutral

                --<10 clinicians: unaffected

            --2015: all practices with >10 clinicians were exposed to the full VM incentives (more money, neutral, or lose money)

            --base payment adjustments ranged from -2% to +2% on the basis of 2014 performance, -4% to +4% on the basis of 2015 performance. BUT there was provision of budget neutrality, so since lots of practices got fined, those that did well actually received bonuses of 16-32% in 2016!!!  And, many of the performance measures were minimally adjusted by patient characteristics (ie, with no adjustment for patients' social or clinical characteristics)

--the goal of this study was to use these different standards in the different years to compare performance differences: between practices with >10 vs >100 clinicians and partial or full VM incentives in 2014; and practices with>100 clinicians in 2015: an ersatz RCT. And also to look at the effect of adjusting for additional patient characteristics and performance outcomes, comparing those practices with a larger proportion of low-income and medically complex patients to practices with those with fewer such patients

--they compared practices of 50-150 clinicians (above and below the cutpoint of 100), assuming that there would be little difference in performance within this range; and similarly 2-30 clinicians for the lower cutpoint

--a random sample of 20% of fee-for-service Medicare beneficiaries were assessed, noting if beneficiaries were also on Medicaid (a marker of income level), whether their Medicare was for a disability, or whether they had end-stage renal disease; they also included the presence of 27 chronic conditions

--outcomes: hospitalization for ambulatory care-sensitive conditions, all-cause 30-day readmissions, Medicare spending, and mortality

--they then compared practices bonuses and penalties in 2014 in practices with >100 vs >10 clinicians (felt to reflect the difference attributable to VM incentives):

--analysis repeated for 2015, when practices with 10-99 clinicians were exposed to penalties comparing practices bonuses and penalties in 2015 in practices with >100 vs >10 clinicians (also felt to reflect the difference attributable to VM incentives):

 

 

Results:

 --2014 data: hospitalizations for ambulatory care-sensitive conditions (ACSC), readmissions, Medicare spending, and mortality were essentially unchanged, very small not statistically significant differences, adjusting for patient characteristics, eg

    --exceeding threshold of >10 clinicians was associated with​an average of 0.0027 more hospitalizations for ACDCs per beneficiary (-0.0003 to 0.0056, p=0.08) than expected

    --exceeding threshold of >100 clinicians was associated with​​an average of 0.002 fewer hospitalizations for ACDCs per beneficiary (0.006 to 0.003, p=0.48) than expected

--2015 data: similar 

--review of the numbers in the article's supplement showed that in data from 2012, 2014 and 2015, ​30-day all-cause readmissions and mortality increased or decreased by a max of 0.006 in those with >100 clinicians. Medicare spending​ (part A and B) was increased $382 in 2012, then -$152 in 2014 and -$1 in 2015.

--practices serving poorer patients (on Medicare and Medicaid, "dually eligible"): higher rates of hospitalizations for ACSCs, Medicare spending and mortality, controlling for sex and age. BUT, adjusting for clinical and social factors: large decreases in differences between practices in the highest vs lowest quartile of the share of dually eligible patients by 55.9% for hospitalizations for ACSCs, 11.9% for Medicare spending, and 34.8% for mortality (all p<0.001)​

--under the new MIPS, this change by including social/clinical characteristics of patients would lead to 2.9-16.7% of practices moving from above to below the exceptional-performance threshold and 1.6-9.9% of practices from below to above the threshold.

--under VM, this change by including social/clinical characteristics of patients , would have moved 4.8-25.7% of practices out of eligibility for bonuses for a given measure (ie up to a quarter of practices getting bonuses would not have by adjusting for these social/medical patient factors) and 3.7-24.9% of practices out of eligibility for penalties (and up to a quarter of practices penalized would not have been if such adjustment happened).

​

Commentary:

--pay-for-performance in the VM initiative (and likely the  new iteration of MIPS) did not have any real effect on important clinical outcomes, such as preventable hospitalizations, mortality, or even cost. and seem to have made the rich richer and the poor poorer (ie, giving money to the well-healed practices in well-healed communities and taking money away from the clinical practices in poorer communities.  sort of like the new proposed US tax plan.....).  there are many potential reasons:

    --there are fundamental differences in the patient populations: poorer patients tend to have more medical as well as social problems. clinicians appropriately need to focus on what are the most important problems for the individual patients, often at the expense of checking off another box on the quality metric (and, by the way, some of the Medicare quality metrics, as well as those from other insurers, are not necessarily associated with significant improvements in health, though they are really easy to measure and report on: the low hanging fruit, which may not be very nutritious or healthy...). eg, making sure to check the A1c in diabetics every 6 months may not improve care when i, as the patient clinician, know that the issue is that the patient has been homeless and the real intervention to improve their outcomes is access to reasonable housing and food. or the patient has not been able to get their medications, so there is no real utility in checking a blood test that i know in advance is going to be out-of-range. and the above study did show that controlling for several medical and  social characteristics of patients did minimize the differences in the measured outcomes. But, the effect of working in these poorer communities is that they will more likely be penalized by pay-for-performance initiatives

    ​--there are fundamental differences in the actual clinical services available. the well-healed clinical practices tend to have more space, more support staff, and more medically-sophisticated/educated/Westernized, non-culturally diverse patients.  past studies have found that they already do a lot more of the "quality metrics" than clinics in poorer communities. so the combo of their already doing many of these "quality" criteria, the added support staff who can (and have) identified patients who need specific tests, and can utilize high educational-level English literature to give to patients to peruse will lead them to benefit from pay-for-performance initiatives​ with minimal extra effort. 

--also, the actual financial incentives were small (penalties of 2% in 2014 and 4% in 2015), which would likely be insufficient for clinicians working in poorer communites to really improve their clinical services enough to effect these quality metrics: would they really be able to hire more people dedicated to performing the necessary tasks to get better numbers?? could they have enough support staff to really call patients at home who missed their appointments, or have been out of care for awhile and should come in for an appointment, or figure out how to improve health metrics when older patients go home to their native countries in the cold months but their insurance will not pay for enough medications to cover them until they return in the warmer weather? and ultimately do the tasks for improving these quality metrics end up in fact falling on the already overworked clinicians/nurses etc?

--as the authors comment in the study regarding the implementation of the VM pay-for-performance system, "inadequate risk adjustment for clinical and socio-economic factors may lead to sustained transfers of payments away from practices serving poorer and sicker patients for reasons not related to quality or efficiency of care"

so,  the results in this study are not so surprising.  earlier studies from 1-2 decades ago found that those well-funded, efficient, well-staffed medical services serving higher income people were already doing most of the "high-quality" care metrics, and the pay-for-performance outlays just gave them more money for what they were mostly already doing (and, if necessary, they could easily change systems to optimize their financial gain). And those serving lower income patients, who have many more social issues, tended to be understaffed and unable/unwilling to deal with these financial incentives: often because doing check-box type medicine really may not be so important for patients who have very large and dominant social/medical needs; and these needs are more important to the patients (and providers) and their health than ordering a less important test; and these health facilities do not have the additional personnel to help implement these "quality metrics" anyway.  This study confirms the results of older studies: this type of financial incentive makes the rich get richer and the poor get poorer.... And such a system thereby can exacerbate the health inequities, potentially decreasing the ability of those working with poorer and sicker patients to have adequate resources to deal with the multitude of social and medical problems of their patients, which is especially true now that we in the US seem to be galloping towards decreasing the already-inadequate social services for our populations.​