Table 5.
Primary and sensitivity analyses for CHICO trial dispensing outcome
| Analysis | No (I:C) | Rate* (95% CI) | Adjusted rate ratio (95% CI)† | P value† | |
|---|---|---|---|---|---|
| Intervention (I) | Control (C) | ||||
| Primary analysis | 144:149 | 0.155 (0.138 to 0.174) | 0.157 (0.140 to 0.176) | 1.011 (0.992 to 1.029) | 0.25 |
| Pre-specified sensitivity analyses: | |||||
| Per protocol: excluding non-compliers in intervention arm‡ | 78:149 | 0.162 (0.143 to 0.184) | 0.161 (0.142 to 0.183) | 1.052 (1.029 to 1.076) | <0.001 |
| 0-4 year olds only | 144:149 | 0.225 (0.200 to 0.253) | 0.225 (0.200 to 0.253) | 1.037 (1.014 to 1.060) | 0.001 |
| 5-9 year olds only | 144:149 | 0.093 (0.083 to 0.103) | 0.096 (0.087 to 0.107) | 0.965 (0.935 to 0.997) | 0.03 |
| Worst case scenario: including “age unknown” as aged 0-9 | 144:149 | 0.183 (0.162 to 0.206) | 0.184 (0.163 to 0.208) | 1.002 (0.986 to 1.018) | 0.84 |
| Excluding internal pilot practices (n=48) | 117:128 | 0.150 (0.133 to 0.169) | 0.153 (0.136 to 0.183) | 1.026 (1.006 to 1.048) | 0.012 |
| Adjusting for No of months affected by covid-19 (0-12)§ | 144:149 | 0.155 (0.138 to 0.174) | 0.157 (0.140 to 0.176) | 1.002 (0.984 to 1.021) | 0.79 |
| Post hoc sensitivity analyses: | |||||
| Complier average causal effect analysis¶ | 144:145 | 1.023 (0.897 to 1.166) | 0.74 | ||
| Excluding individual practice months of data from March 2020** | 105:121 | 0.231 (0.208 to 0.257) | 0.249 (0.224 to 0.277) | 0.967 (0.946 to 0.989) | 0.003 |
| Amoxicillin items only (macrolides excluded) | 144:149 | 0.120 (0.107 to 0.136) | 0.124 (0.110 to 0.139) | 0.994 (0.974 to 1.014) | 0.56 |
| Adjusting for random effects at PCN level | 144:149 | 0.150 (0.141 to 0.161) | 0.151 (0.141 to 0.162) | 0.942 (0.916 to 0.969) | <0.001 |
| Adjusting for delayed start (number of months)†† | 144:149 | 0.155 (0.138 to 0.174) | 0.157 (0.140 to 0.176) | 1.043 (1.023 to 1.063) | <0.001 |
CCG=clinical commissioning group; CI=confidence interval; PCN=primary care network.
Rates taken from random effects Poisson regression, incorporating CCG as random effect.
Random effects Poisson regression, adjusting for baseline dispensing rate and incorporating CCG as random effect.
Excluding practices in intervention arm that did not comply with intervention ([number of uses/list size of children]<0.05) or for which compliance was unknown; excludes practice that merged with another practice that had taken part in control arm.
Including numerical variable (0-12) to indicate how many months were affected by covid-19.
Adjusting for baseline rate and number of months affected by covid-19.
Excluding follow-up months affected by covid-19.
Some practices asked to delay their start date and covariate was included to indicate number of months they delayed.