Model | Enrolled (voluntarily) | Enrolled (PNBSF-subsidised) | Predictions | N | |||||
OR or IRR | ATE | OR or IRR | ATE | Not enrolled in a CBHI | Enrolled (voluntarily) | Enrolled (PNBSF-subsidised) | |||
Model 1: Consulted in a health facility? | Logistic | 2.357** | 0.144** | 1.326 | 0.045 | 0.315 | 0.459 | 0.360 | 418 |
(0.85) | (0.06) | (0.68) | (0.08) | (0.03) | (0.06) | (0.08) | |||
Model 2: No of prenatal care visits | Poisson | 1.173*** | 0.565*** | 0.901 | −0.325 | 3.268 | 3.833 | 2.943 | 197 |
(0.06) | (0.18) | (0.11) | (0.36) | (0.14) | (0.14) | (0.34) | |||
Model 3: Gave birth in a health facility? | Logistic | 7.883*** | 0.349*** | 3.767* | 0.238* | 0.443 | 0.792 | 0.681 | 197 |
(4.51) | (0.08) | (3.02) | (0.12) | (0.04) | (0.07) | (0.12) | |||
Model 4: HH had to forgo medical consultation? | Logistic | 1.129 | 0.026 | 1.240 | 0.047 | 0.357 | 0.383 | 0.404 | 1001 |
(0.23) | (0.04) | (0.27) | (0.05) | (0.02) | (0.04) | (0.04) | |||
Model 5: HH had to forgo medical treatment? | Logistic | 0.646* | −0.072* | 0.680 | −0.065 | 0.263 | 0.190 | 0.198 | 1001 |
(0.16) | (0.04) | (0.19) | (0.04) | (0.02) | (0.04) | (0.04) | |||
Model 6: HH had catastrophic health expenditures (40% threshold)? | Logistic | 1.070 | 0.004 | 0.956 | −0.002 | 0.059 | 0.063 | 0.057 | 1001 |
(0.41) | (0.02) | (0.40) | (0.02) | (0.01) | (0.02) | (0.02) | |||
Model 7: HH had catastrophic health expenditures (30% threshold)? | Logistic | 1.428 | 0.038 | 0.587 | −0.042 | 0.110 | 0.149 | 0.069 | 1001 |
(0.40) | (0.03) | (0.22) | (0.03) | (0.01) | (0.03) | (0.02) | |||
Model 8: HH had catastrophic health expenditures (20% threshold)? | Logistic | 1.456 | 0.057 | 1.174 | 0.023 | 0.170 | 0.228 | 0.193 | 1001 |
(0.36) | (0.04) | (0.34) | (0.04) | (0.02) | (0.04) | (0.04) |
In all regressions, the reference group was individuals or HH not enrolled in a CBHI. OR are provided for logistic regressions, and IRR for Poisson regressions. For logistic models, predictions are predicted probabilities of the dependent variable. For Poisson models, predictions are the predicted number of events. Regressions were weighted by both (1) the inverse probability of treatment obtained in the first-step multinomial logistic model of health insurance enrolment at the individual level (models 1–3) or HH level (models 4–8) and (2) the sampling weights to account for choice-based stratified samples. Standard errors in parentheses (with clustering at the HH level in all individual-level regressions to account for intra-HH correlation). Full results for each model are provided in online supplemental material.
*p<0.05, **p<0.01, ***p<0.001.
ATE, average treatment effect; CBHI, community-based health insurance; HH, household; IRR, incidence rate ratio; PNBSF, Programme National de Bourses de Sécurité Familiale.