Results of mixed logit models and probability analyses
| Coefficients | SE* | P value | Relative impact (%)§ | ||
| Breast cancer | |||||
| Attributes† | |||||
| 1) Volume | 100 patients (ref.) | – | – | – | – |
| 350 patients | 0.208 | 0.120 | 0.08 | 10.3 | |
| 450 patients | 0.232 | 0.083 | 0.01 | 11.5 | |
| 2) Waiting time 1 (diagnosis—tumour resection) | 15 days (ref.) | – | – | – | – |
| 25 days | −0.219 | 0.077 | 0.01 | 10.9 | |
| 3) Waiting time 2 (diagnosis—combination surgery) | 25 days (ref.) | – | – | – | – |
| 45 days | −0.524 | 0.123 | <0.01 | 25.6 | |
| 4) Preserved breast contour | 60% (ref.) | – | – | – | – |
| 80% | 0.279 | 0.131 | 0.03 | 13.8 | |
| 90% | 0.434 | 0.138 | <0.01 | 21.4 | |
| 5) Combination surgery | 10% (ref.) | – | – | – | – |
| 30% | 0.379 | 0.184 | 0.04 | 18.7 | |
| 60% | 0.855 | 0.184 | <0.01 | 40.3 | |
| 6) Tumour residual | 1% (ref.) | – | – | – | – |
| 5% | −1.999 | 0.332 | <0.01 | 69.9 | |
| SD of random parameters | |||||
| 6) Tumour residual | 5% | 0.964 | 0.182 | <0.01 | |
| Number of individuals | 84 | ||||
| Number of observations | 972 | ||||
| Log-likelihood | −194.763 | ||||
| Colon cancer | |||||
| Attributes‡ | |||||
| 1) Volume | 50 patients (ref.) | – | – | – | – |
| 150 patients | 0.129 | 0.084 | 0.12 | 6.5 | |
| 200 patients | 0.217 | 0.047 | <0.01 | 10.8 | |
| 2) Waiting time (diagnosis—tumour resection) | 60% (ref.) | – | – | – | – |
| 80% | 0.307 | 0.088 | <0.01 | 15.2 | |
| 90% | 0.147 | 0.072 | 0.04 | 7.3 | |
| 3) Tumour residual | 85% (ref.) | – | – | – | – |
| 95% | 0.157 | 0.059 | 0.01 | 7.9 | |
| 4) Complications | 10% (ref.) | – | – | – | – |
| 20% | −0.172 | 0.048 | <0.01 | 8.6 | |
| 5) Failure to rescue | 5% (ref.) | – | – | – | – |
| 15% | −0.439 | 0.071 | <0.01 | 20.9 | |
| SD of random parameters | |||||
| 5) Failure to rescue | 15% | 0.480 | 0.094 | <0.01 | |
| Number of individuals | 145 | ||||
| Number of observations | 1386 | ||||
| Log-likelihood | −409.463 | ||||
Ref. reference category in the dummy coding for which the derived utility is standardised to zero.
*Reflects the bootstrapped SE
†Based on theory, signs of coefficients were expected to be positive for attributes 2 and 6, and negative for attributes 1 and 3–5 (shown in table 1)
‡Based on theory, signs of coefficients were expected to be positive for attributes 1–3, and negative for attributes 4 and 5 (shown in table 1)
§Expressed as the percentage change in choice probability relative to the base alternative (reference levels across all attributes).