TABLE 6

Patient household (n=282) dissaving score associations with health and socioeconomic variables

Variable	Dissaving score	Univariate logistic regression		Multiple logistic regression
Variable	Mean	Unadjusted OR (95% CI)	p-value	Adjusted OR (95% CI)	p-value
Catastrophic costs
Incurred	0.58	2.4 (1.5–3.9)	0.001	1.8 (1.1–3.1)	0.02
Not incurred	−0.43
Poverty
Poorer	0.37	2.3 (1.4–3.7)	0.001	1.8 (1.1–3.0)	0.03
Less poor	−0.35
Food insecurity
High	0.3	2.6 (1.5–4.5)	0.001	2.2 (1.2–3.8)	0.008
Low	−0.26
Secondary education
Incomplete	0.36	1.7 (1.0–2.7)	0.03
Complete	−0.165
Employment
Unpaid/no work	0.16	1.1 (0.68–1.8)	0.6
Paid work	−0.23
Symptom duration
Longer	0.09	1.4 (0.84–2.2)	0.2
Shorter	−0.068
Type of TB
Non-MDR	0.008	1.1 (0.49–2.6)	0.8
MDR	−0.09
Sex
Female	0.07	1.1 (0.66–1.7)	0.8
Male	−0.04

The patient cohort had a median average dissaving score of 0. Higher (more positive) scores indicate greater dissaving and hence greater financial shock. Health and socioeconomic variables were analysed for association with having a greater than average dissaving score by univariate logistic regression. Multiple logistic regression was then performed with stepwise exclusion of non-contributory (p>0.1) variables. The variables that have blank cells in the multiple logistic regression columns were those non-contributory variables excluded from the final model. The variable “secondary education” was entered but was significantly associated in the multiple regression model. Secondary education, employment, symptom duration, type of TB and sex all refer to the patient. A complementary linear regression analysis of the association of a higher dissaving score with health and socioeconomic variables showed a similar pattern of significance with a higher dissaving score being independently associated with incurring catastrophic costs (coefficient 0.30 (95% CI 0.047–0.55), p=0.02) and having greater food insecurity (coefficient 0.38 (95% CI 0.12–0.64), p=0.004). TB: tuberculosis; MDR: multidrug resistant.