Leaked data suggests the Department for Work and Pensions is holding Atos to extremely tight tolerances. Photograph: Janine Wiedel Photolibrary/Alamy
Ask Atos, the company responsible for executing the work capability assessment (WCA), or the Department for Work and Pensions, which defines how the WCA is conducted, and they will tell you that they have no targets for the number of people who pass. Yet a new report from the Centre for Welfare Reform, How Norms Become Targets, uses a leaked set of Atos data to suggest that the DWP is holding Atos to extremely tight tolerances on its results.
Atos and the DWP admit to the existence of “statistical norms” and that these are used to manage the performance of individual healthcare professionals carrying out the assessments. Campaigners have long claimed that these norms function as de facto targets, but were surprised by the detail of the data logged and matched against acceptable ranges. Not only are there figures for overall numbers of people awarded the points needed to qualify for the employment and support allowance (ESA), figures also exist for individual prognoses, for the points awarded, even for the word count of the summary findings. And each Atos region is expected to stray no further than 20% from the national average.
Combining any data-gathering system with pressure to meet expectations will drive staff to converge on the “official” numbers; norms will become de facto targets, and no manager wants to be forced to justify their region’s figures. Unfortunately pressure to meet expectations is exactly the process described to the Guardian by Atos whistleblower Dr Greg Wood, who went public after being repeatedly asked to change assessments, including at least one case that conflicted with his professional medical opinion.
Whether the Atos data represents targets, as the CWR report suggests, or norms, as Atos insists, the numbers themselves are deeply problematic. An assessor is expected to see about 40 people per week, 65% of whom, the data shows, are expected to fail the assessment. The remainder will be split between the ESA support group (14.5%) and the work-related activities group (20.5%). Those who pass, in either group, are then further divided across five sub-groups, which specify whether an applicant should be expected to recover in six-24 months or "longer term" (in practice three years). Only 2.6% of WCAs are expected to result in a "longer-term" prognosis, which effectively means an assessor can allocate the prognosis to just one person a week; allocating a single extra person across a month hovers on breaching the allowed 20% variation. All other applicants, no matter their disability, are labelled as expected to have recovered within two years or less.
This suggests an explanation for some of the stranger Atos rulings, where people with lifelong or degenerative disabilities have been toldtheir conditions are expected to improve in six months. Any assessor struggling to keep down their average for longer-term prognoses has to be tempted to assign a shorter prognosis instead. Someone with a long-term disability should theoretically face an assessment once every three years, but if they are consistently assigned to the six-months prognosis group, they will potentially face not one but six assessments in that period.
Similar problems exist for points awarded during the assessment. The "descriptors", if matched, are worth six, nine or 15 points, and many disabled people will match multiple descriptors. It takes 15 points to qualify for the ESA, yet the report suggests national averages of 2.1 points for physical issues and 3.6 points for mental issues. The only way to maintain such low averages is by scoring several people at zero points for each one who passes. Are cases of people who score zero points when they clearly should pass simply evidence that the assessor saw too many seriously ill and disabled people that week? There is already an outcry over the state of the work capability assessment, with roughly one in six of all assessments successfully appealed against, at a cost to the Tribunals Service, and the taxpayer, of more than £75m per year. If that failure rate is not caused by poor quality work at the healthcare professional level but is a consequence of the Atos management system, which in turn is driven by the contractual requirements placed on it by DWP, then shouldn't that outcry be louder still?
Guardian
