A Pragmatic Policy Trial Testing the Behavioral Economic Principles Loss Framing, Loss Aversion, and Social Pressure in Physician Pay-for-Performance
Discussant: Fredric Blavin
Methods: The design was a quasi-experimental comparison involving patients treated by physicians in the Trinity Organization versus Non-Trinity physicians at Advocate. Non-Trinity physicians were matched based on pre-intervention (2015) performance level and 2014-2015 historic trend. The patients were assigned by attribution to physicians who were cluster-randomized to 3 arms: 1) An active control using loss framing with performance incentivized by “money left on the table” (9 physicians caring for 666 patients); 2) Loss framing with loss aversion by pre-funding incentives in a health system bank account in the physician’s name (13 physicians caring for 1,506 patients); and 3) Loss framing with social pressure by making the performance bonus determined by individual (50%) and group (50%) performance (13 physicians caring for 1,358 patients). Using a difference-in-differences design, we compared changes at Trinity as a whole (loss framing) to propensity-matched non-Trinity physicians with physician fixed-effects and patient risk-adjustment. The primary outcome variable was the proportion of evidence-based quality measures met for patients with chronic conditions.
Results: The Trinity RCT included 3,530 patients randomized via 35 physicians and the non-Trinity matched cohort included 3,801 patients across 35 physicians. There were no differences across physician characteristics. Trinity patients were more likely to be black (68% vs. 17%; P < 0.001) and younger (64 vs. 66 years; P < 0.001). There were small differences in patient characteristics by Arm - Arm 2 had the highest average number of chronic conditions (Arm 1, 1.54 conditions; Arm 2, 1.63 conditions; Arm 3, 1.60 conditions; P < 0.001) and greatest proportion of black patients (Arm 1, 64%; Arm 2, 59%; Arm 3, 80%; P < 0.001).
Unadjusted results of the quasi-experimental analysis showed that Trinity patients experienced a significantly larger increase in evidence-based care rates compared to Non-Trinity patients (83.55% in 2015 to 88.69% in 2016 vs. 85.39% in 2015 to 87.42%; difference-in-difference 3.1% points; P<0.001). In adjusted analysis, Trinity patients received a larger increase in evidence-based care relative to non-Trinity patients from 2015 to 2016 (Trinity 86.49% to 90.70% vs. Non-Trinity 87.50% to 89.27%; difference-in-difference 2.4%; P<0.001). The RCT within Trinity did not reveal significant additional improvement between the Arm 2 (loss framing + loss aversion) and Arm 3 (loss framing + enhanced social pressure) versus loss framing (Arm 1) (P = 0.52).
Conclusions: In the first trial to test behavioral economic principles applied to P4P, loss framing resulted in significantly improved quality for chronic care patients relative to a gain-framed comparison group. Adding loss aversion and enhanced social pressure did not lead to further quality improvement. Use of behavioral economic principles in P4P design may enhance quality improvement and improve the value of care, though specific design elements need to be evaluated systematically.