Table 3. Comparison of antibiotic use among children under the control and intervention treatment conditions during the STAR stepped wedge cluster randomized trial (as treated analysis).
Children seen during control periods were evaluated and managed using the iCCM SCJA; children seen during the intervention condition were evaluated and managed using the STAR SCJA.
| Observed proportions1 | Adjusted intervention effect2 (95% CI) | ||||||
|---|---|---|---|---|---|---|---|
| Control | Intervention | Prevalence difference | Odds ratio | Estimate (SD-BC2)3 | |||
| Within-period | Between-period | ||||||
| Antibiotics at initial visit | All children | 539/587 (91.8%) | 448/633 (70.8%) | −24.6% (−36.1%, −13.1%) |
0.18 (0.06, 0.49) |
0.045 (0.024) | 0.040 (0.018) |
| mRDT+ | 285/310 (91.9%) | 228/318 (71.7%) | −21.1% (−37.7%, −4.6%) |
0.18 (0.06, 0.59) |
0.063 (0.045) | 0.041 (0.032) | |
| mRDT − | 254/277 (91.7%) | 218/313 (69.6%) | −26.6% (−39.8%, −13.4%) |
0.26 (0.11,0.61) |
0.056 (0.018) | 0.054 (0.027) | |
| Antibiotics at any time | 540/584 (92.5%) | 458/620 (73.9%) | −22.4% (−33.5%, −11.3%) |
0.17 (0.06, 0.51) |
0.049 (0.029) | 0.048 (0.023) | |
1Three children seen during an intervention period were evaluated using the iCCM SCJA; they were included in the control condition.
2The prevalence difference and odds ratio estimates are based on GEE with an identity link and logit link, respectively, adjusting for categorical periods and stratum effects. The GEE analysis assumes marginal Bernoulli distributions for any antibiotic use and employs bias-corrected standard errors to adjust for the moderately small number of clusters (villages).
3The correlation parameters for a nested exchangeable working correlation structure are estimated with MAEE and reported for the logistic model only (see S1 Appendix and S4 Table).
GEE, generalized estimating equation; iCCM, Integrated Community Care Management; MAEE, matrix-adjusted estimating equation; mRDT, malaria rapid diagnostic test; SCJA, Sick Child Job Aid.