Primary care staff in our study described several ways in which PMs may lead to inappropriate care, may take the focus off of patient concerns and patient service, and may make it more difficult for patients to make informed value-consistent decisions (especially when the patient’s values conflict with interventions deemed appropriate by PMs). These problems have undoubtedly existed to some extent in all healthcare systems even before the advent of modern performance measurement systems. Indeed, PM systems have likely been responsible for improvements in some areas within these quality domains while simultaneously exacerbating problems in other areas (e.g. reducing inappropriate underuse of interventions while increasing inappropriate overuse). Although our research methodology does not allow us to identify which, if any, unintended consequences are inherent in all PM systems, it is noteworthy that few of the unintended negative consequences identified in our research appeared to be directly related to national PM policies. Instead, the influence of PM policy on patient care appeared to be mediated through facility-level efforts to implement these national policies. This finding suggests that greater collaboration may be needed between developers of PMs and those responsible for facility implementation so that PM developers understand the variation in how measures are adapted into practice and local implementers have a thorough understanding of the rationale and evidence underlying the measure.
As hypothesized elsewhere,9–11,22
we found some evidence that PMs can drive inappropriate care, such as polypharmacy. Prescribing drugs involves greater risk but is a more efficient way to meet the PMs than behavioral interventions and providers may feel incentivized by the PMs to provide medications rather than behavioral health counseling. The administration of more drugs than is medically indicated may occur in over half of elderly patients25
and is a risk factor for morbidity and mortality.26–29
The examples of overtreatment provided by participants were sometimes reported second hand. It is unknown whether they represent perception or reality. Even if more perceived than real, perceptions may shape norms and have unintended effects on provider behavior and satisfaction; or to quote the sociologist W.I. Thomas: “If men define situations as real, they are real in their consequences.”30
This work contributes to a growing body of evidence suggesting that PM systems can have negative effects on provider-patient communication. As others have observed,14,31
we found that PMs increase providers’ workload during the clinical encounter which can crowd out education and the discussion of issues that are of higher priority to patients (including, on occasion, the chief complaint). It is therefore perhaps not surprising that the establishment of a national pay for performance system in the UK failed to improve patient ratings of patient-provider communication even though this measure was incentivized.2
PMs that count patient refusals against PM scores may add to the communication problem by making it more difficult for providers to accept a patient’s choice to decline care.
Avenues for Improvement
This work does not provide sufficient evidence to conclude that the frequency and seriousness of unintended negative consequences of PM are sufficient to warrant system-wide change. Our data does indicate that PM systems do carry some risk of negatively impacting the quality of patient care. Organizations wishing to take the lead in improving their PM systems may therefore be interested in exploring ways to minimize the risk of these problems. The following approaches may address several of the negative unintended consequences identified in this research:
- Integrate the development of PMs and local implementation strategies through collaboration between PM system administrators and facility staff.
- Develop PMs monitoring overtreatment to balance the current focus on undertreatment.
- Modify PMs to credit appropriate provider treatment behavior as well as achievement of patient health goals. (See Kerr and colleagues work on “tightly linked” PMs.)32,33
- Track the number of patients who refuse care and exclude them from measure denominators. Although some gaming may occur14,34,35 research suggests that widespread abuse is unlikely.35
VHA is currently making several changes to the national PM system to incorporate insights from this and other research. A newly convened national clinical reminder standardization workgroup will develop reminders that facilitate patient-centered care decisions. New PMs are being considered that will reward clinically appropriate action, even if the patient has not achieved specific targets. Finally, more flexible incentive plans were recently introduced that hold facilities accountable for fewer measures, chosen to address areas with the greatest opportunity for local improvement.
Our methodology allowed us to explore a range of unintended consequences of PM and identify clinic processes and provider actions that appear to lead to their occurrence. However, this methodology is not well suited for assessing the frequency with which the identified effects occur or evaluating the balance of positive and negative effects of PM. Additional quantitative studies are needed to address these issues. Additionally, our findings may not generalize to other healthcare systems. However, most of the VHA’s primary care PMs are similar or identical to HEDIS (Healthcare Effectiveness Data and Information Set) measures which are used by over 90% of U.S. health plans.36
Our data consisted of facility staff self-reports (and in some cases second-hand reports on the behavior of others). Participants may have incorrectly inferred negative effects on patients, may have been reluctant to share perceived deficiencies in personal or facility practices, or may have been more likely than non-participants to express the views summarized above.