We screened 2252 children with asthma indicated on their school health records (), and 713 were eligible for the program. We enrolled 530 children from 67 schools and preschools, for a participation rate of 74% (530 of 713 children). A total of 523 subjects (99%) had data for the primary analysis.
Enrollment flow diagram. ETS indicates environmental tobacco smoke.
There were no differences in demographic characteristics between children in the 2 study groups (). The children’s mean age was 7.1 years, and more than half were male (58%), black (63%), and covered by Medicaid (73%). In addition, 54% of the children lived witha smoker and 69% reported using a preventive medication at baseline. Asthma symptoms were similar at baseline, with children in both groups having an average of 8 symptom-free days per 2 weeks.
Population Demographic Variables and Baseline Morbidity
summarizes the primary outcomes by group. During the peak winter season, children in the treatment group experienced more symptom-free days compared with children in the control group (mean symptom-free days per 2 weeks, 11.6 vs 10.7, respectively). Children in the treatment group also had significantly fewer nights with symptoms, fewer days with activity limitation, less rescue medication use, and fewer days absent from school due to asthma compared with children in the control group. The analyses were repeated controlling for baseline symptoms and covariates. Treatment was associated with an additional 0.92 symptom-free day per 2 weeks, on average, compared with the control group (95% confidence interval [CI], 0.50 to 1.33). Full-year outcomes also showed a significant treatment effect (difference=0.73 symptom-free days per 2 weeks; 95% CI, 0.43 to 1.04). While children’s cotinine level was associated with the outcomes, the intervention effect was independent of change in cotinine levels. The mean (SD) end-of-year cotinine values were 1.05 (1.49) μg/L for children in the treatment group and 1.16 (1.94) μg/L for children in the control group (difference=−0.11 μg/L; 95% CI, −0.19 to 0.42) (to convert to nanomoles per liter, multiply by 5.675).
Children in the treatment group were less likely to have an exacerbation requiring prednisone throughout the study period compared with children in the control group (12% vs 18%, respectively). When controlling for covariates, the intervention was associated with a 36% reduction in asthma exacerbations compared with the control group (relative risk=0.64; 95% CI, 0.41 to 1.00). For emergency department visits, acute office visits, and hospitalizations individually, the visit rates were lower for children in the treatment group but there were no significant differences. At the end of the study, 49% of children in the treatment group were receiving fluticasone and 51% were receiving fluticasone with salmeterol. In the control group, 60% reported using preventive asthma medication (29% fluticasone, 11% fluticasone with salmeterol, and 21% other).
illustrates the number of symptom-free days per 2 weeks by month for the children in the treatment and control groups during the peak winter season. Both groups demonstrated improvement over time (P=.005 for the control group; P<.001 for the treatment group). However, children in the treatment group experienced more symptom-free days compared with children in the control group, and by February they were experiencing a mean of 1.2 additional symptom-free days per 2-week period (95% CI, 0.59 to 1.80).
Mean symptom-free days per 2 weeks by month. In November, January, and February, children in the treatment group had significantly more symptom-free days than children in the control group (P=.02, .005, and <.001, respectively).
We next considered the primary study outcome separately for children without and with smoke exposure in the home (). There was no significant treatment×smoke exposure interaction effect (P=.07). Stratified analysis showed positive treatment effects for symptom-free days for children without smoke exposure (mean symptom-free days per 2 weeks, 11.6 for the treatment group vs 10.5 for the control group; adjusted difference=0.91; 95% CI, 0.31 to 1.51) and for children with smoke exposure (mean symptom-free days per 2 weeks, 11.6 for the treatment group vs 10.9 for the control group; adjusted difference=0.96; 95% CI, 0.39 to 1.52), suggesting an effect of the intervention for both groups of children. Similar results also were present for the other outcome measures.
Outcomes for the Peak Winter Season Stratified by Smoke Exposure
Comparisons between subjects with complete data and those without complete data showed that baseline symptom-free days and race were associated with missing outcomes. Missing data were imputed using a multiple imputation algorithm based on multiple regression,35
and multiple imputation inference was carried out.36
The result showed a treatment effect similar to that shown by the primary analysis; the treatment group had 0.91 additional symptom-free day per 2 weeks compared with the control group (95% CI, 0.49 to 1.33). There were no significant adverse events in either group.