Results of mixed-effects multilevel linear regressiona
| Variable | Main analysis | Sensitivity analysis excluding practices outside the 1st and 99th centiles (extreme outliers) | Sensitivity analysis GMS/PMS practices |
|---|---|---|---|
| Observations | 25 614 | 25 333 | 15 745/9359b |
| Groups (practices) | 6900 | 6872 | 4826/3004b |
| Log IMD score, beta-coefficient | 0.0060 | 0.0030 | 0.0045/0.0089 |
| Log IMD score, P-value (95% CI) | 0.003 (0.0021 to 0.010) | 0.061 (−0.00013 to 0.0062) | 0.034/0.016 (0.00034 to 0.0086)/(0.0016 to 0.016) |
| Log IMD score effect size: change in slope over 4 years (%) if variable increases by 10% | 0.060% | 0.029% | 0.045%/0.089% |
| ICC | 0.81 | 0.75 | 0.82/0.77 |
↵a Dependent variable = adjusted total NHS payments per patient.
↵b Some practices changed contract type during the study and were included in both the GMS and PMS analyses only during the years during which they had that contract. This accounts for why the sum of each group exceeded the total number of groups, while the sum of each group’s observations did not exceed the total number of observations. GMS = General Medical Services. ICC = intraclass correlation. IMD = Index of Multiple Deprivation (1 = most deprived and 10 = least deprived). PMS = Personal Medical Services.