Table 3.
PptChange at Baseline GDR of | |||||
---|---|---|---|---|---|
Mode of Purchase | Intervention | Regression Coefficient | 30% | 45% | p-Value |
Retail | Mailing | −0.248 | −4.94 | −6.03 | < .0001 |
Advertising | −0.006 | −0.13 | −0.15 | .875 | |
Generic sampling | −0.001 | −0.02 | −0.02 | .951 | |
Physician incentive | −0.016 | −0.33 | −0.40 | .616 | |
Doubling copayments for brand name drugs | 0.553 | 8.60 | 9.55 | <.0001 | |
Mail order | Mailing | −0.079 | −1.63 | −1.95 | .232 |
Advertising | −0.130 | −2.65 | −3.19 | .297 | |
Generic sampling | −0.083 | −1.71 | −2.04 | .301 | |
Physician incentive | −0.100 | −2.06 | −2.46 | .406 | |
Doubling copayments for brand name drugs | 0.19 | 2.84 | 3.27 | .668 |
Each percentage-point change (PptChange) in the above table was derived using the formula where base is the baseline GDR, coefficient is the estimated regression coefficient of the intervention, δ is the change in the value of the intervention variable (δ=1 for the interventions; δ=log(2) for doubling the copayment ratio), and exp(.) denotes the exponential function. Note that this formula is just a mechanism for translating the parameter estimate to the probability scale; it does not compute the interaction or marginal effect that has been defined as in Ai and Norton (2003) but rather estimates the change in the probability that would have occurred had there not been an intervention.
GDR, generic dispensing rate