Table 2.
Bivariable and multivariable-adjusted effect of household use of wood as a cooking fuel on treatment failure at 48 hours among N=284 children 1 to 23 months of age presenting with pneumonia to a tertiary hospital in Gaborone, Botswana, April 2012 to April 2014.
| Bivariable Models | Full Multivariable Model | Reduced Multivariable Modela | ||||
|---|---|---|---|---|---|---|
| RR (95% CI) | P b | RR (95% CI) | P b | RR (95% CI) | P b | |
| Exposure of interest | ||||||
| Use of wood as a cooking fuel | 1.62 (1.19, 2.21) | 0.002 | 1.48 (1.08, 2.03) | 0.01 | 1.44 (1.09, 1.92) | 0.01 |
| Demographics | ||||||
| Age | ||||||
| 1 to <6 months | 3.36 (2.22, 5.07) | <0.0001 | 3.48 (2.27, 5.35) | <0.0001 | 3.23 (2.14, 4.88) | <0.0001 |
| 6 to <24 months | 1 (ref) | 1 (ref) | 1 (ref) | |||
| Male gender | 1.10 (0.80, 1.52) | 0.57 | 1.02 (0.77, 1.34) | 0.91 | ||
| Birth weight <2,500 grams | 1.71 (1.25, 2.34) | 0.001 | 1.27 (0.95, 1.69) | 0.11 | ||
| HIV exposure statusc | ||||||
| HIV-infected | 2.34 (1.55, 3.53) | <0.0001 | 1.63 (1.08, 2.46) | 0.02 | ||
| HIV-exposed, uninfected | 1.78 (1.27, 2.51) | 0.001 | 1.32 (0.85, 2.07) | 0.22 | ||
| HIV-unexposed | 1 (ref) | 1 (ref) | ||||
| Nutrition and infant feeding practices | ||||||
| Malnutritiond | ||||||
| None | 1 (ref) | 1 (ref) | ||||
| Moderatee | 1.71 (1.14, 2.57) | 0.01 | 1.55 (0.99, 2.43) | 0.06 | ||
| Severef | 1.35 (0.81, 2.22) | 0.25 | 1.66 (0.93, 2.95) | 0.08 | ||
| Currently breastfeeding | 0.90 (0.65, 1.25) | 0.54 | 0.90 (0.61, 1.33) | 0.61 | ||
| Socioeconomic factors | ||||||
| Maternal education level | ||||||
| None or primary | 2.30 (1.32, 4.00) | 0.003 | 1.44 (0.83, 2.48) | 0.19 | ||
| Secondary | 1.47 (0.90, 2.42) | 0.13 | 1.31 (0.82, 2.07) | 0.26 | ||
| Tertiary | 1 (ref) | 1 (ref) | ||||
| Electricity in home | 0.69 (0.50, 0.94) | 0.02 | 1.15 (0.63, 2.12) | 0.65 | ||
| Municipal water or private water source | 0.54 (0.39, 0.75) | 0.0002 | 0.72 (0.49, 1.06) | 0.09 | ||
| Refrigerator in home | 0.67 (0.49, 0.92) | 0.01 | 0.95 (0.54, 1.68) | 0.86 | ||
| Household food insecurity in the prior 4 weeks | ||||||
| Smaller meal than needed because of a lack of resources | 1.12 (0.73, 1.73) | 0.60 | 1.06 (0.53, 2.13) | 0.87 | ||
| Fewer meals in a day because of a lack of resources | 1.01 (0.64, 1.59) | 0.98 | 0.66 (0.31, 1.39) | 0.28 | ||
| Travel of more than 1 hour to clinic or hospitalg | 1.04 (0.60, 1.81) | 0.88 | 0.72 (0.48, 1.09) | 0.12 | ||
RR, risk ratio; CI, confidence interval
Reduced multivariable model was constructed using a change-in-estimate approach. At each stage, the covariate for which removal caused the smallest change in the RR of the exposure was dropped. Only those variables whose removal from the model changed the RR for exposure by more than 10% were retained in the reduced multivariable model.
Wald χ2 P-values
HIV exposure status could not be established for N=2 children
Data on nutritional status were missing from N=15 children
Moderate malnutrition defined as weight-for-length between -3 and -2 standard deviations on World Health Organization (WHO) growth curves or, for children ≥6 months of age, mid-upper arm circumference (MUAC) between 115mm and 125mm
Severe malnutrition defined as weight-for-length <-3 standard deviations on WHO growth curves, MUAC <115mm (for children ≥6 months), or bilateral edema of nutritional origin
Data on duration of travel to clinic or hospital were missing from N=1 child