Table 5. Estimated regressions of average total and average on-site cost per patient-year of antiretroviral therapy in 45 facilities in Zambia.
| Individually estimated linear regressions | Criterion for defining dummy variable | Jointly estimated regressions of logged dependent variable on dummy variables∧a | Point estimates and 95% confidence intervals for predicted average cost per patient∧b | |||||
| Column numbers | (1) | (2) | (3) | (4) | (5) | (6) | (7) | (8) |
| Dependent variables | Averagetotal cost | Averageon-site cost | Averageoff-site cost | NA | Natural log of average total cost | Natural log of average on-site cost | Average total cost | Average on-site cost |
| Explanatory variables | ||||||||
| Constant | 951.66** (0.050) | 395.99 (0.255) | 555.67 (0.200) | Constant | 6.824**(0.000) | 6.208**(0.000) | 919.8 (787.6–1,074.1) | 496.6 (412.5–597.8) |
| Proportion of patients adherent | 248.02 (0.545) | 256.83 (0.433) | −8.82 (0.973) | >.91 | 0.104**(0.067) | 0.156**(0.021) | 1,020.3 (894.3–1,164.1) | 580.3 (495.6–679.3) |
| Average baseline CD4 | −1.77+ (0.132) | −0.60 (0.336) | −1.16 (0.229) | >144 | −0.128**(0.008) | −0.108*(0.061) | 859.5 (741.8–995.9) | 470.1 (394.2–560.7) |
| Proportion of patients initiatingat WHO Stage 4 | 711.34* (0.090) | 256.83 (0.397) | 454.76 (0.251) | <.10 | −0.044 (0.361) | −0.103*(0.071) | 686.3 (574.8–819.5) | 352.4 (285.1–435.6) |
| Dummy = 1 if Hospital | −241.05** (0.000) | −152.35** (0.000) | −88.70 (0.004) | = 1 if Hospital | −0.225**(0.000) | −0.288**(0.000) | 575.1 (461.1–717.3) | 297.2 (228.2–387.0) |
| Months of ART service provision | −7.32** (0.004) | −2.46 (0.200) | −4.86** (0.023) | >24 | −0.177**(0.005) | −0.170**(0.022) | 468.8 (403.4–544.7) | 300.3 (250.9–359.3) |
| Patient-years of ART per yr of operation | −0.05* (0.075) | −0.03+ (0.198) | −0.03 (0.276) | >300 | −0.204**(0.008) | 0.010 (0.910) | 423.0 (370.5–483.0) | 276.0 (235.5–323.4) |
| >800 | −0.103**(0.07) | −0.084 (0.21) | ||||||
| Dummy = 1 if Rural | −42.45 (0.657) | −111.73* (0.050) | 69.28 (0.320) | = 1 if Rural | −0.035 (0.626) | −0.185** (0.033) | 408.3 (332.9–500.9) | 229.4 (179.7–292.9) |
| Dummy = 1 if Private | 26.33 (0.594) | −65.98** (0.040) | 92.31** (0.005) | = 1 if Private | 0.007 (0.919) | −0.133+ (0.131) | 411.4 (312.3–542.0) | 200.8 (144.4–279.2) |
| Number of observations | 45 | 45 | 45 | 45 | 45 | |||
| R-squared | 0.636 | 0.539 | 0.556 | 0.782 | 0.659 | |||
| p of Early Start∧c | 0.093 | 0.498 | 0.057 | 0.043 | ||||
| p of Economies of Scale | 0.084 | 0.206 | 0.283 | 0.000 | ||||
Note:**p<0.05,
p<0.1,+p<0.2. p-values are given in parentheses below the coefficients in columns (1) through (6).
Coefficients in columns (5) and (6) are estimated by Zellner’s seemingly unrelated regression method using Stata’s sureg command.
Predicted values in columns (7) and (8) are calculated from the coefficients in columns (5) and (6) respectively, as explained in the text.
Values of p for “Early Start” are the significance level at which the joint hypothesis can be rejected that neither baseline CD4 nor proportion of patients initiating at WHO stage IV influences the dependent variable. For the two equations estimated jointly (columns (5) and (6)) the p-value is the probability that these four estimated coefficients could occur if all four parameters were actually zero.