Reducing the Floor Effect in the SF-6D: A Feasibility Study

The SF-6D suffers from a floor effect where for patient groups in severe health a significant number of patients report the
lowest level of health possible for some dimensions, meaning the SF-6D cannot capture a deterioration in health for these
patients. This paper presents a feasibility study aimed at reducing this floor effect. A modified SF-6D classification system
was proposed that incorporated an additional level in each of the physical functioning and role limitations dimensions. The
modified classification system was valued by a Portuguese general population sample (n = 140) using the Portuguese SF-6D protocol. A sample of 82 health states were valued and several regression models were estimated
to produce preference weights to predict health state values for all states defined by the modified classification system.
Estimations at the individual level were performed using 950 health state valuations. Models were analyzed in terms of logical
consistency of coefficients, overall fit and predictive ability and were compared to Portuguese SF-6D models. The additional
severity levels included in the modified classification system have significant decrements in health state values. These additional
severity levels do not significantly impact on the modelled preference weights (the regression coefficients) of other levels
across all dimensions. This feasibility study modified the SF-6D to reduce the floor effect. This study presents one option
and further research in this area is encouraged.