Page 62 - Contributed Paper Session (CPS) - Volume 8

CPS2184 M Lutfor Rahman
(dependent variable: infected or non-infected), the results of fitting the simple
and multiple logistic regression models to the TB data are shown in Table 4.
Table 5 presents the results of hierarchical and network logistic regressions.
In the Table 4, the crude estimates show that the covariates age, diabetes
mellitus, exposure duration, sleep together, and eat together are significant at
1% or 5% level of significance. Whereas, the variables biological product,
symptomatic period, age, contact type, and exposure duration are found to be
significant in multiple logistic regression. The household contacts are 6.819
times more likely to be infected than casual contacts. The variables in multiple
logistic regression have been selected by stepwise backward method based
on likelihood ratio.
Table 4: Estimates of parameters in simple (crude) and multiple logistic
regression (adjusted) analysis of TB networks in Portugal.
Characteri Category Binary Logistic Regression
stics
Simple Logistic Multiple Logistic (Model 1)
Coeffici p- Crude Coeffici p- Adjusted
ent val OR ent val OR
ue (95% ue (95%
CI) CI)
Intercept -3.070 0.00
0
Index patient characteristics
Biological Bronchoalv 1 1
product eolar
studied lavage
Sputum 0.026 0.91 1.027 0.673 0.04 1.960
7 (0.623,1.69 5 (1.016-
2)NS 3.782) *
0.004 0.06 1.004 0.007 0.00 1.007
Symptomat 4 (1.000,1.00 6 (1.002-
ic period 9)NS 1.013) **
days
Chest Without 1
radiograp cavitation
hy -0.031 0.89 0.970
Cavitation 9 (0.602,1.56
1) NS
Contacts’ characteristics
Sex Female 1
Male 0.421 0.08 1.523
1 (0.949,2.44
4)NS
Age in 0.020 0.00 1.020 0.024 0.00 1.024
years 1 (1.009,1.0 3 (1.008-
32)** 1.041) **
No 1

51 | I S I W S C 2 0 1 9