Table 2. Determinants of cough severity.
| Ln (TOTAL TIME COUGHING) | EPISODE FREQUENCY | |||
|---|---|---|---|---|
| Bivariable | Multivariable | Bivariable | Multivariable | |
| Beta Coefficient (95% Confidence Interval) | Beta Coefficient (95% Confidence Interval) | Rate Ratio (95% Confidence Interval) | Rate Ratio (95% Confidence Interval) | |
| (p-value) | (p-value) | (p-value) | (p-value) | |
| Treatment day | -0.66 (-0.91, -0.40) | -0.66 (-0.92, -0.41) | 0.66 (0.52, 0.84) | 0.61 (0.46, 0.79) |
| (p<0.001) | (p<0.001) | (p = 0.001) | (p<0.001) | |
| Treatment day ^2 | 0.06 (0.02, 0.10) | 0.06 (0.02, 0.10) | 1.03 (1.00, 1.06) | 1.04 (1.00, 1.07) |
| (p = 0.002) | (p = 0.001) | (p = 0.063) | (p = 0.028) | |
| Age* | 0.11 (-0.26, 0.47) | n/a | 0.98 (0.82, 1.16) | n/a |
| (p = 0.575) | (p = 0.785) | |||
| Sex = Female | -0.05 (-0.59, 0.49) | n/a | 1.05 (0.63, 1.75) | n/a |
| (p = 0.843) | (p = 0.846) | |||
| Smoker | 0.98 (-0.21, 2.17) | 0.59 (-0.06, 1.25) | 2.01 (1.34, 3.01) | 1.82 (0.90, 3.73) |
| (p = 0.108) | (p = 0.077) | (p = 0.001) | (p = 0.094) | |
| MODS positive | 0.71 (0.31, 1.10) | n/a | 1.73 (1.14, 2.63) | n/a |
| (p<0.001) | (p = 0.010) | |||
| Prior TB | 1.04 (0.27, 1.82) | 1.44 (0.66, 2.22) | 2.66 (1.51, 4.66) | 4.03 (2.19, 7.42) |
| (p = 0.009) | (p<0.001) | (p = 0.001) | (p<0.001) | |
| Diabetes | 0.59 (-0.24, 1.43) | 0.86 (0.09, 1.63) | 1.38 (0.81, 2.37) | 1.75 (1.13, 2.71) |
| (p = 0.164) | (p = 0.028) | (p = 0.240) | (p = 0.012) | |
| HIV positive | 0.03 (-0.83, 0.88) | 0.56 (-0.21, 1.34) | 0.73 (0.40, 1.33) | 1.40 (0.71, 2.78) |
| (p = 0.952) | (p = 0.150) | (p = 0.305) | (p = 0.366) | |
| Drug resistant TB | -0.03 (-0.83, 0.77) | -0.35 (-1.09, 0.39) | 1.33 (0.45, 2.77) | 0.73 (0.37, 1.45) |
| (p = 0.942) | (p = 0.351) | (p = 0.784) | (p = 0.366) | |
Shown here are the results of bivariable and multivariable Tobit regression models in which the outcome of interest is the log-transformed TOTAL TIME COUGHING (log seconds per hour). The lowest observed value was taken as the lower limit of the model. This approach allows recordings where no cough episodes were observed to be included in the analysis. To model COUGH EPISODE FREQUENCY, bivariable and multivariable negative binomial regression models were constructed. Both models included a random intercept to account for the correlation between recordings from the same study participant. Microbiology (MODS result) was collinear with treatment day (the relationship between characteristics of cough and treatment day is explained by microbiological response), therefore the final multivariable includes treatment day and not MODS result. Treatment day was adjusted for in the model using a quadratic term (treatment day ^2) to reflect the non-linear, rapid decrease in cough observed early in treatment [5].
*Age per 10 years, centered at 34 years.
(Total participants = 69, total recordings = 359)