Table 2 Sensitivity analysis for aVE of 3 doses of rotavirus vaccine in preventing hospital admission for AGE among children aged 6–59 months
Type of analysis | aVE | 95% CI lower limit | 95% CI upper limit | Interpretation |
|---|---|---|---|---|
Unmatched regression model | 54% | 45% | 62% | Main analysis model used in the study |
Matched regression model | 52% | 35% | 64% | Values similar to main model; not adapted because of loss of sample size due to matching |
Propensity score matched regression model | 64% | 36% | 80% | Values similar to main model; not adapted because of loss of sample size due to matching |
PSW regression model | ||||
Inverse probability of treatment weights | 56% | 47% | 63% | Among eligible children who were fully vaccinated or unvaccinated (N = 7,226), for estimating PSW analysis, the PS was estimated as mean PS of 0.24 and s.d. of 0.06; children in the study have 24% probability of receiving the treatment. The PS values are normally distributed in this population group. The standardized mean differences before and after weighting were assessed for covariates to see their balance between treated and control groups. Since all covariates have s.m.d. <0.1 after weighting, the model is considered well balanced |
Stabilized treatment weights | 56% | 47% | 63% | |
Overlap treatment weights | 48% | 38% | 57% | |
E value to measure residual confounding | Measured value for OR 0.46 is 3.77 | An unmeasured confounder would need to increase the odds of both being vaccinated and getting infected by at least 3.77× each to fully explain away the observed effect | ||
