Table 1. Moderating effect of number of prescribed ATC codes on health care costs during first quarter of PIM use.
| NEG | EG | Difference | ||||||
|---|---|---|---|---|---|---|---|---|
| Costs in € | Effect of ATC | (SE) | Effect of ATC | (SE) | Moderating effect of ATC | (SE) | p-value | R2 |
| Medication | 70.39 | (0.36) | 58.87 | (0.46) | -11.52 | (0.58) | 0.000 | 0.164 |
| Outpatient physician services | 33.28 | (0.17) | 28.76 | (0.21) | -4.47 | (0.27) | 0.000 | 0.107 |
| Hospital treatment | 79.54 | (1.54) | 240.04 | (2.02) | 160.50 | (2.54) | 0.000 | 0.007 |
| Rehabilitation | 5.40 | (0.18) | 15.03 | (0.25) | 9.63 | (0.31) | 0.000 | 0.007 |
| Medical supplies | 2.54 | (0.04) | 2.12 | (0.05) | -0.42 | (0.07) | 0.000 | 0.014 |
| Total costs | 195.99 | (1.66) | 333.20 | (2.18) | 137.20 | (2.74) | 0.000 | 0.090 |
The calculation of moderating effects of the number of prescribed ATC codes in the incident quarter of PIM use is based on fully saturated linear mixture regression models with maximum likelihood estimators with a quadratic term for the number of prescribed ATC codes. Calculation of moderating effect of ATC is based on the interaction effect between the study group (EG vs. balanced NEG) and the number of prescribed ATC codes in the 1st quarter of the post-period. The statistical fit of the single models is shown in the last column in form of the Maddala-R2 which is based on the maximum-likelihood.