Table 4.
Per-protocol subgroup estimates for the primary and secondary health care utilisation outcomes, excluding observations in the intervention arm that did not receive at least two CHW home visits in the month preceding the survey, during the three-year trial period overall and at each follow-up time point*
| Outcome | ARC | ARI | C vs I, AORCS (95% CI) | P-value |
|---|---|---|---|---|
|
Prompt treatment within the health sector (n = 13 500)
|
|
|||
| Overall† |
0.52 |
0.57 |
1.22 (1.06-1.40) |
0.005 |
| Time point‡ |
|
|
|
|
| 12 mo |
0.58 |
0.66 |
1.43 (1.21-1.69) |
<0.001 |
| 24 mo |
0.46 |
0.48 |
1.07 (0.89-1.28) |
0.453 |
| 36 mo |
0.52 |
0.55 |
1.13 (0.94-1.35) |
0.184 |
| LR test |
|
|
|
0.0036 |
|
Any prompt treatment (n = 13 493)
|
|
|||
| Overall† |
0.59 |
0.64 |
1.26 (1.09-1.44) |
0.001 |
| Time point‡ |
|
|
|
|
| 12 mo |
0.64 |
0.72 |
1.43 (1.21-1.69) |
<0.001 |
| 24 mo |
0.55 |
0.58 |
1.10 (0.92-1.32) |
0.315 |
| 36 mo |
0.56 |
0.61 |
1.23 (1.03-1.47) |
0.023 |
| LR test |
|
|
|
0.0205 |
|
Health sector evaluation (n = 14 518)
|
|
|||
| Overall† |
0.62 |
0.68 |
1.29 (1.12-1.48) |
<0.001 |
| Time point‡ |
|
|
|
|
| 12 mo |
0.63 |
0.70 |
1.38 (1.17-1.65) |
<0.001 |
| 24 mo |
0.65 |
0.71 |
1.30 (1.09-1.55) |
0.003 |
| 36 mo |
0.58 |
0.61 |
1.17 (0.97-1.39) |
0.094 |
| LR test |
|
|
|
0.1736 |
|
Any care (n = 14 527)
|
|
|||
| Overall† |
0.69 |
0.75 |
1.35 (1.17-1.55) |
<0.001 |
| Time point‡ |
|
|
|
|
| 12 mo |
0.70 |
0.76 |
1.37 (1.16-1.63) |
<0.001 |
| 24 mo |
0.73 |
0.79 |
1.35 (1.13-1.62) |
0.001 |
| 36 mo |
0.63 |
0.69 |
1.31 (1.10-1.57) |
0.003 |
| LR test |
|
|
|
0.8945 |
|
Recommended case management (n = 10 569)
|
|
|||
| Overall† |
0.42 |
0.45 |
1.20 (1.06-1.37) |
0.005 |
| Time point‡ |
|
|
|
|
| 12 mo |
0.46 |
0.49 |
1.19 (0.99-1.42) |
0.061 |
| 24 mo |
0.39 |
0.45 |
1.35 (1.12-1.63) |
0.001 |
| 36 mo |
0.38 |
0.39 |
1.06 (0.86-1.30) |
0.592 |
| LR test |
|
|
|
0.1435 |
|
Prompt, recommended case management (n = 10 569)
|
|
|||
| Overall† |
0.36 |
0.39 |
1.15 (1.01-1.32) |
0.040 |
| Time point‡ |
|
|
|
|
| 12 mo |
0.43 |
0.46 |
1.19 (0.99-1.43) |
0.062 |
| 24 mo |
0.31 |
0.36 |
1.26 (1.04-1.53) |
0.017 |
| 36 mo |
0.33 |
0.33 |
0.99 (0.80-1.22) |
0.890 |
| LR test | 0.1280 | |||
AORCS – cluster-specific adjusted odds ratio, ARC – absolute risk of events in the control arm, ARI – absolute risk of events in the intervention arm, C – control clusters, CI – confidence interval, I – intervention clusters, ICC – intracluster correlation coefficient, LR – likelihood ratio, mo – months
*Two regression models are presented here: regression model 1 controlled for the time effect t = 1, 2, 3, to estimate the intervention effect during the three-year follow-up period overall. Regression model 2 included the interaction term ηit that estimated the intervention effect at each time point. The likelihood ratio test corresponds to the interaction term in model 2. Adjusted models controlled for child’s age (0-11, 12-23, 24-35, 36-59 mo) and sex; baseline cluster-level summary of the outcome; baseline cluster-level summary of household wealth (quintiles), mother’s decision-making power (any, none), and mother’s mobility (none, dependent mobility, independent mobility), which were deemed imbalanced at baseline and likely risk factors; PHC catchment area and cluster distance to PHC (coded as a continuous variable in the models for prompt treatment within the health sector, any prompt treatment, prompt, recommended case management, and pneumonia where the relationship with distance was linear, and otherwise coded as a dichotomous variable using a five-kilometre cut-off), which were the variables on which randomisation was stratified; and symptom (fever, diarrhoea with no blood, cough with fast breathing, combination), only for recommended case management outcomes.
†Regression model 1.
‡Regression model 2.