Provider Monitoring and Pay-for-Performance when Multiple Providers Affect Outcomes: An Application to Renal Dialysis

Abstract

Purchasers routinely measure and reward quality and efficiency. Monitoring includes "report cards" for hospitals, surgeons, dialysis facilities and health plans and can indirectly influence care by altering patient flows or informing quality improvement initiatives. "Pay-for-performance" (P4P) systems incorporate direct incentives for measured performance.

To implement performance measures, an appropriate unit of analysis must be selected. Although performance is generally measured at one level (e.g., surgeon or hospital), outcomes depend on multiple providers' actions. In the dialysis context, practice variations across dialysis facilities and nephrologists may independently influence outcomes. The dialysis facility has been the unit of measurement for quality reports and P4P proposals. Although the facility is a convenient unit for measurement, potential drawbacks exist. Selecting the facility implicitly attributes responsibility for practices of non-employee physicians. Ignoring nephrologists' incentives level may miss improvement opportunities. Facility-level measures fail to provide patients with guidance regarding choice of physician. Due to these limitations, the prevailing focus on the facility should be examined empirically.

Using claims for U.S. Medicare hemodialysis patients in 2004, we determined resource utilization [Medicare Allowable Charges (MAC) per dialysis session for services billed by the dialysis facility other than the dialysis treatment] and clinical outcomes [achieving targets for anemia management (hematocrit (Hct)≥33%) and dialysis adequacy (urea reduction ratio (URR)≥65%)]. For each patient-month, we identified the primary dialysis facility and nephrologist and calculated MAC/session and % of patients achieving targets for all patients treated by each facility/physician pair. Using 9,994 facility-physician pairs, we estimated a mixed model with fixed effects for patient conditions and random effects for facility and physician. Sufficient cross-over existed between facilities and physicians to estimate their separate contributions (in 65% of facilities, multiple physicians treated ≥5 patients, and 55% of physicians treated ≥5 patients in more than one facility).

Mean MAC/session was $81.80, 81% of patients had Hct≥33%, and 92% had URR≥65%. For each measure, outcomes varied substantially at both levels, but variation was more pronounced at the facility level. The standard deviations (SD) across facilities and physicians, respectively, were $19.45 and $6.76 for MAC/session, 6% and 3% for Hct≥33%, and 7% and 3% for URR≥65%.

By using data from facilities with multiple physicians and from physicians treating patients at multiple facilities, it is possible to distinguish the variation in performance attributable to facilities from that attributable to physicians. Similar methods could be employed for other types of providers. If dialysis quality measurement and P4P incentives are targeted to only one provider, the facility is the appropriate focus of such measures and incentives. Nonetheless, the existence of variation across physicians raises issues regarding the extent to which quality reports and P4P places facilities at risk for outcomes they only partially control. Cooperative efforts and alignment of incentives between facility managers and nephrologists to optimize outcomes and efficiency will become increasingly important under P4P programs and proposed reforms to pay for more services prospectively. The observed resource use variation is large, with the facility-level SD of $19.45 per session translating into $155,600 for a typical-sized facility.

Keywords: Quality, Pay-for-Performance, Incentives

JEL Classification: I11, I18