Policy makers have only a limited number of strategic levers they can pull to effect improvements in health system performance, including reforms to market structure, provider payment mechanisms and information systems. Performance management seeks to combine these elements of a health system into a coherent whole that actively promotes desired objectives. Yet Oliver’s paper clearly demonstrates that – whilst the design of such reforms has predominantly been a concern of strategic policy makers – their effectiveness often rests ultimately on the response of the individual clinicians and clinical teams responsible for delivering health services (Oliver, Reference Oliver2015). Too little attention has been paid to the preferences and motivation of health care professionals, and the constraints under which they operate. A typical economic approach seeks to model the trade-off between the clinician’s ‘effort’ and financial reward, with an allowance for the altruism of the professional, in deriving utility from good patient outcomes (Chalkley and Malcomson, Reference Chalkley and Malcomson1998). Yet, there is precious little evidence to support such theory (McGuire, Reference McGuire2011), and Oliver sets out a number of possible alternative models.

In considering the need for performance management, it is important to recognize that there are good reasons for some diversity of treatments and clinical outcomes, given the inevitable heterogeneity of patients. However, countless studies have identified residual variations in practice and outcomes that cannot be explained by patient characteristics or preferences. It is difficult to escape the conclusion that much of the unexplained variation is caused by unwarranted variations in clinical practice (Wennberg, Reference Wennberg2010). There is therefore great scope for some convergence towards better practice, as set out vividly by the Institute of Medicine (2001). With this in mind, it is instructive to consider from the perspective of the health care professional why performance management initiatives have generally failed to secure step changes in performance by reducing unwarranted variation. Therefore, I briefly consider below three possible reasons why clinical professionals may not have responded as intended.

First, they may be unaware of performance guidelines or published performance information, or of the associated incentives. Although targets might be given a high degree of prominence at a strategic level, it is often the case that the intended priorities and incentives are significantly blunted at the operational level. For example, hospital managers may fail to transmit strategic priorities to their clinical staff. In an early study in Scotland, Mannion and Goddard (Reference Mannion and Goddard2001) uncovered a high degree of ignorance amongst clinical staff of important performance metrics that were intended to inform clinical improvements. The traditional autonomy of the medical profession from day-to-day management serves to increase the risk of performance management initiatives failing to secure traction amongst front-line clinicians. However, the increased possibilities offered by information technology and the increased emphasis on maintaining professional competence make it difficult to see why ignorance should be a continuing obstacle to performance management.

The second obstacle is more challenging. Professionals may be aware of guidelines and performance data, but often consider them to be faulty, irrelevant or incomplete. In many systems, a lack of data accuracy and timeliness have been persistent and legitimate barriers to acceptance amongst professionals. A particularly important concern is the accuracy of any ‘risk adjustment’ undertaken to correct for variations in clinical case mix (Iezzoni, Reference Iezzoni2010). Although superficially a technical issue, the credibility of many performance management regimes has been fatally undermined by poor risk adjustment and inappropriate treatment of uncertainty. Furthermore, data limitations have often meant that the only outcome measure available has been mortality, a patently inadequate metric for many interventions. Performance measurement will achieve clinical credibility only when the use of more sensitive metrics, such as patient-reported outcomes, becomes established (Smith and Street, Reference Smith and Street2013).

Third, professionals may be aware of the performance information and recognize its validity, but do not have adequate incentives or capacity to make improvements. Many performance improvement initiatives require considerable investment in new ways of working, new collaborations or new information systems. These reconfigurations may in principle be cost-effective. However, they often require actions by a variety of partners, and the personal commitment required from clinicians in persuading managers, fellow clinicians and patients to accept the reforms may be substantial. Even if relevant incentives are put in place, the institutional constraints may appear insuperable. Moreover, even when necessary change could in principle be secured, the chosen incentives may appear underpowered, or just a temporary fad, and fail to convince those responsible for delivering services to change their practices.

Perhaps the most challenging aspect of performance management as applied to health care is that it can come into direct conflict with the notion of clinical professionalism. A performance management system seeks to elicit prescribed behaviour, or a preferred outcome, from a clinical professional. Yet individual patients vary enormously in both their preferences and clinical conditions. Professionals are trained to respect such variations, and to vary their recommendations and treatments accordingly. It is important therefore to recognize an important paradox at the heart of recent initiatives such as ‘pay-for-performance’, which seek to guide the professional’s behaviour explicitly through financial incentives. No one (least of all George Bernard Shaw) would argue that professional actions can ever be completely independent of financial considerations (Shaw, Reference Shaw1971). Yet, there is nevertheless a strong professional ethic, particularly embodied in medical training, of doing the best for a patient, regardless of any countervailing incentives, whether financial or otherwise. Therefore, unless very carefully designed, the generalizations embodied in clinical guidelines, performance reporting, pay-for-performance or other performance management techniques may at times seriously compromise the medical ethos of personalizing care and ignoring immediate financial return.

There are many reasons for clinical variation. Some of them are virtuous, in the spirit of personalization, others less so, caused by ignorance, laziness or the pursuit of financial reward. Until performance management instruments are seen to be respecting the need for legitimate departures from prescribed practice, they will be treated with suspicion by clinical professionals. Even worse, the emphasis on managerial control and explicit incentives may ‘crowd out’ the natural wish of most professionals to improve the health of their patients (Smith, Reference Smith2002). The evidence to date, as documented in Oliver’s article, suggests that neither the best nor the worst predictions of the impact of performance management have yet materialized. This is not to say that the techniques of performance management do not ‘work’. Rather, it suggests that policy makers should build on early experience to design performance management reforms that reflect a sound understanding of what motivates professionals and what constrains their improvement. A major research effort is needed to that end.