Table 4.

Top 15 most important variables for predicting tuberculosis screening yield (Approach 2b, CART with categorical outcome; n = 73 neighborhoods).

Importance ranking	Variable	Relative variable importance score
1	Percent of tuberculosis patients with a prior tuberculosis episode	100.0
2	Percent of historic tuberculosis patients that were aged 15–44 years	62.3
3	Percent of households that own a vehicle	49.2
4	Proportion of the population that is female	32.5
5	Percent of population that have completed only a primary school education	31.4
6	Percent of population that have any post-secondary school education	28.2
7	Historic tuberculosis case notification rate	27.6
8	Percent of population that worked for pay in the past week	27.1
9	Percent of households that own a refrigerator	26.8
10	Percent of residences that are in informal or non-permanent structures	26.0
11	Historic tuberculosis case notification rate for individuals 15–44 years old	25.7
12	Percent of historic tuberculosis patients that were aged < 15 years	25.6
13	Historic tuberculosis case notification rate for females	24.8
14	Historic tuberculosis case notification rate for individuals > 44 years old	24.1
15	Population density (population per km2)	24.0