Pay for Whose Idea of Performance?

In the US, the latest fashionable approach to controlling health care costs and improving quality is "pay for performance" (of physicians) (P4P). We have posted several times about why P4P may not yet be ready for prime time (most recently here and here.) Our main concerns were:

• Use of outcome measures not adequately adjusted for disease severity and other relevant patient characteristics may lead to perverse incentives.
• Process-based measures may assess processes not under physicians' control, and hence be invalid.
• Use of often inaccurate administrative data may threaten measures' validity.
• Measures intended to control costs will not control quality.
• Measures focused on only a few processes will distract from improving quality for other patients, physicians, and problems.

David Healy's recent discussion of how sponsors may manipulate research studies to increase the likelihood of getting results that support their vested interests suggests yet another problem with pay for performance (see our post here).
Healy concentrated on ways in data from pharmaceutical company sponsored studies of selective serotonin reuptake inhibitors (SSRIs) for depression were manipulated to minimize the apparent harms of these drugs.
There is already anecdotal evidence that pharmaceutical companies have manipulated how data about SSRIs is disseminated, even suppressing unfavorable data. A meta-analysis that included previously unpublished data suggested that the risks of treatment with some SSRIs for children and adolescents outweighed their benefits, although data from published trials showed benefits commensurate or outweighing risks. (Whittington CJ, Kendall T, Fonagy P et al. Selective serontonin reuptake inhibitors in childhood depression: systematic review of published versus unpublished data. Lancet 2004; 363: 1341-1345.) The authors concluded, "drug sponsors who withhold trial data (or do not make full trial reports available) undermine the guideline programme, which can ultimately lead to recommendations for treatments that are ineffective, cause harm, or both."
Two years ago, the Canadian Medical Association Journal reported that "an internal document advised staff at the international drug giant GlaxoSmithKline (GSK) to withhold clinical trial findings in 1998 that indicated the antidepressant paroxetine (Paxil in North America and Seroxat in the UK) had no beneficial effect in treating adolescents." (Kondro W, Sibbald B. Drug company experts advised staff to withhold data about SSRI use in children. Can Med Assoc J 2004; 170: 783.)
There is also evidence that pharmaceutical companies manipulated the discussion of SSRIs by supporting the publication of ghost-written articles. For example, David Healy also determined that Pfizer Inc. paid Current Medical Directions, a medical information company, to prepare 85 articles on sertraline (Zoloft). Of these 55 were published, often in widely read and prestigious journals, outnumbering other articles on the drug published from 1998-2000. (Healy D, Cattall D. Interface between authorship, industry and science in the domain of therapeutics. Brit J Psychiatr 2003; 183: 22-27. Link is here.)
Thus, Healy's latest article combined with previously available information raises strong doubts about the integrity of the data available to support the wide-spread use of SSRIs in depression. Furthermore, it is not clear whether physicians' generally favorable attitudes to treatment of depression with SSRIs were unduly influenced by articles cloaked in academic raiments, but actually ghost-written at the behest of pharmaceutical companies. This suggests we physicians ought to be re-thinking how we treat depression, and when and if to treat depression with SSRIs.
However, some P4P schemes incorporate measures of treatment of depression with SSRIs.
For example, the proportions of patients treated acutely and chronically with antidepressant drugs are two of 26 clinical performance measures in the Ambulatory Care Quality Alliance "Recommended Starter Set" of measures. This group of measures is one of the more widely measured prototypes to support P4P. (Note that although the measure does not specifically limit antidepressants assessed to SSRIs, these are certainly now the drugs most widely used as "antidepressants.") This measure was actually obtained from the National Committee on Quality Assurance, and appears to be part of the HEDIS measures already used in a voluntary system to rate health plans.
Thus it appears that P4P risks becoming a way to promote products whose manufacturers have suppressed clinical research, manipulated its design, performance, and analysis, or manipulated the dissemination of its results.
Credible P4P programs must only use measures clearly supported by clinical research data of unquestioned integrity. They should not be use measures designed to promote the vested interests of particular organizations, such as, but not limited to pharmaceutical companies, ahead of the interests of patients.
Physicians must clearly ask whose idea of performance are we supposed to be paid for?