Table 3.
Multivariable analysis of factors associated with the school-level hookworm PRR among 199 schools surveyed during year 5 survey in Kenya
| Covariate | Hookworm (Coefficient (95% CI), P-value)* | |
|---|---|---|
| Prevalence | PRR | |
| Baseline prevalence, % | ||
| < 1 | −3.10 (−4.46;−1.75), P < 0.001 | |
| 1–20 | −2.09 (−3.07;−1.11), P < 0.001 | |
| 20–50 | Insufficient obs | |
| > 50 | Reference | |
| Treatment covariates | ||
| Treatment rounds with ABZ | ||
| Three | Reference | |
| Four | 20.16 (−0.86;47.08), P = 0.059 | |
| Six | 26.76 (−6.88;68.10), P = 0.110 | |
| Seven | 27.23 (1.13;65.41), P = 0.042 | |
| School WASH covariates | ||
| Overall pupil per latrine ratio | ||
| < 30 | 1.47 (0.59;2.36), P = 0.001 | −10.53 (−30.29;8.89), P = 0.285 |
| 30–60 | 1.28 (0.50;2.07), P = 0.001 | −2.07 (−21.64;14.65), P = 0.706 |
| 60–90 | 2.11 (0.97;3.25), P < 0.001 | −50.78 (−76.76;−22.85), P < 0.001 |
| > 90 | Reference | |
| School feeding program | −1.09 (−1.85;−0.33), P = 0.005 | |
PRR = prevalence relative reduction.
Regression coefficients of association together with their 95% CIs were determined using multivariable mixed-effects linear regression models with a random intercept at county and subcounty levels. Statistical significance of the coefficients was determined by the absence of zero overlapping in the 95% CIs (values in bold). In all the models, negative values of the coefficients indicated a decrease in the rate of school-level infection or low values of PRR, whereas positive values of the coefficients indicated an increase in the rate of school-level infection or high values of PRR.