Establishing the value for money of investing in an intervention that releases resources and reduces future expected health gains among recipients

To set the context we review NICE documentation on appraisal and social value judgements, detailing what it implies for how committees should interpret and apply the threshold, and whether specific mention is given to interventions that decrease health. We consider the extent to which committees have been presented with incremental cost-effectiveness ratios (ICERs) located in the South West quadrant of a cost-effectiveness plane (less cost and less health than the comparator).  

Next, we review concepts relating to willingness to pay and willingness to accept, and link these to the NICE decision making context as a social decision maker, making decisions on behalf of others, about which interventions to include in universal health coverage. We summarise previous papers that have considered the application of a kinked threshold, and provide an illustration of the practical implications of applying different thresholds by quadrant of the cost-effectiveness plane.

NICE Committees rarely see ICERs in the South West of the cost-effectiveness plane. The application of a different threshold in these circumstances would not address concerns about disinvestment or individual loss aversion. NICE recommendations can lead to investment and disinvestment in services regardless of where the ICER is in the cost-effectiveness plane. The ICER, composed of the average difference in cost divided by the average difference in effect, does not enable consideration of the distribution of individual health losses and gains. Application of a kinked threshold is incompatible with maximising health gains from available resources. While NICE does depart from health maximisation, by applying different thresholds in end of life and highly specialised technologies, it would be difficult to develop a case of support for a different threshold to be applied when an ICER is presented in the South West quadrant.