Setting pay for performance targets: Do poor performers give up?

We examine the effect of a health plan's pay for performance incentives on the percentage of outpatient drug prescriptions that are filled with generic rather than brand-name drugs in physicians' practices in an established physician network - the generic prescription rate (GPR). The financial reward was based on the performance of the entire network, but the network implemented rewards at the practice level. Practice-level rewards were awarded on an all-or-nothing basis if the GPR met or exceeded specialty-specific targets that increased each year. Although that design gave the practices a strong incentive to meet the target, practices performing far below the target might 'give up', costing the network its reward. Using a partial adjustment model, we estimate that in the absence of pay for performance, the average equilibrium value of GPR was 58.3%. We estimate that GPR would be maximized if the target were set at 77%. The GPR-maximizing target would induce an improvement in average GPR from 58.3% to 65.8% or 7.5 percentage points. When the target is set above 80%, practices with equilibrium GPR below 58.3% will 'give up' in the sense that they will not improve relative to their equilibrium value.