. Author manuscript; available in PMC: 2021 Jul 8.

Published in final edited form as: Lancet Infect Dis. 2017 Aug 18;17(11):1190–1199. doi: 10.1016/S1473-3099(17)30447-4

Table 2. Multivariable Cox regression analysis of factors associated with tuberculosis in the derivation cohort.

	Unadjusted hazard ratio (95%CI)	Adjusted hazard ratio (95%CI)	p value	Regression coefficient	Points assigned in risk score^*
Contacts
BMI	0.87 (0.84–0.91)	0.87 (0.83–0.91)	<0.0001	–0.138	Value of BMI
History of previous tuberculosis	2.0 (1.4–2.9)	1.8 (1.2–2.6)	0.005	0.566	−4
High-risk age group (15–19 or >50 years)	1.5 (1.1–2.0)	1.3 (0.96–1.8)	0.09	0.272	−2
Index cases
Sustained exposure to index case	1.6 (1.2–2.2)	1.8 (1.3–2.4)	0.0003	0.573	−4
Exposure to male index case	1.5 (1.0–2.1)	1.7 (1.2–2.4)	0.001	0.554	−4
Households
Lower community household socioeconomic position^†	1.4 (1.0–2.0)	1.3 (0.95–1.8)	0.1	0.281	−2
Exposed to indoor air pollution	1.7 (1.3–2.4)	1.4 (0.97–1.9)	0.07	0.302	−2
Any household member with a history of tuberculosis	1.7 (1.2–2.3)	1.7 (1.2–2.3)	0.001	0.530	−4
Fewer windows per room	1.6 (1.1–2.2)	1.6 (1.2–2.2)	0.004	0.469	−3

The test of proportional hazards assumption for the entire model was χ²=7·27, p=0·61. BMI=body-mass index.

Calculated by multiplying the Cox regression coefficient by a constant (–7·25) and rounding to the nearest integer; the constant was chosen so that the exact BMI value could be used in the score, using the reciprocal (1/–0·138).

^†

Measured with a household poverty index that combined 12 variables characterising education, access to services and material living conditions into a continuous variable that was dichotomised into two equal categories. The following interactions showed no significant associations and made no significant differences to the model when tested using the likelihood-ratio test, and, therefore, were excluded from the final model: sustained exposure and sex of index case (p_interaction=0·8); sex of index case and index case smear positivity status (p_interaction=0·7); exposure to indoor air pollution and fewer windows per room (p_interaction=0·5); sustained exposure and fewer windows per room (p_interaction=0·6); and sex of index case and socioeconomic position (p_interaction=0·07). Among the household characteristics, the only significant associations were between exposure to indoor air pollution and socioeconomic position (p<0·0001); exposure to indoor air pollution and fewer windows per room (p=0·0003); and indoor air pollution and any household member with a history of tuberculosis (p<0·0001).