Table 5.
Results of the negative binomial regression model
| Independent variable | Random effect model | ||
|---|---|---|---|
| IRR (95% CI) | P | ||
| Income | 0.897 (0.853, 0.943) | < 0.001 | |
| Costgap | 1.000 (1.000, 1.000) | 0.671 | |
| Size | street THs# | ||
| ordinary THs | 0.527 (0.211, 1.316) | 0.170 | |
| central THs | 0.664 (0.256, 1.727) | 0.401 | |
| IDNs | No# | ||
| Tight | 0.484 (0.290, 0.810) | 0.006 | |
| Loose | 1.011 (0.672, 1.522) | 0.956 | |
| Merged | 0.881 (0.319, 2.433) | 0.807 | |
| PRP | No# | ||
| Yes | 2.305 (1.570, 3.384) | < 0.001 | |
| Payment | SDAQ# | ||
| SDLQ | 0.868 (0.731, 1.031) | 0.106 | |
| FFS | 0.939 (0.638, 1.382) | 0.749 | |
| Constant | 23.364 (7.920, 68.921) | ||
| Log likelihood | − 875.77356 | ||
| Wald chi2(9) | 82.57 | ||
| Prob>chi2 | < 0.001 | ||
Note: IRR incidence rate ratio, SDAQ single-disease with abundant quota, SDLQ single-disease with limited quota, #, reference group. According to the Hausman test (chi2(8) =5.00, Prob>chi2 = 0.757), we use the random effect model