Table 3.
Hierarchical multivariable logistic regression analysis of Tuberculosis treatment abandonment
| Variable | Categories | Odds ratio | 95% CI |
|---|---|---|---|
| Notification year (ref = 2007) | 2008 | 0.98 | [0.79–1.22] |
| 2009 | 1.05 | [0.84–1.30] | |
| 2010 | 1.15 | [0.92–1.44] | |
| 2011 | 1.39 | [1.11–1.73] | |
| 2012 | 1.17 | [0.92–1.48] | |
| 2013 | 1.42 | [1.12–1.81] | |
| 2014 | 1.71 | [1.34–2.16] | |
| Age (years; ref. = 0–9) | 10–19 | 1.40 | [0.71–2.75] |
| 20–29 | 1.95 | [1.00–3.79] | |
| 30–39 | 1.71 | [0.88–3.32] | |
| 40–49 | 1.30 | [0.67–2.52] | |
| 50–59 | 1.13 | [0.58–2.22] | |
| 60–69 | 0.71 | [0.36–1.42] | |
| >69 | 0.65 | [0.32–1.31] | |
| Male | 1.21 | [1.08–1.35] | |
| Race (ref = white) | Black | 1.20 | [0.95–1.52] |
| Yellow | 2.02 | [1.31–3.11] | |
| Brown | 1.28 | [1.09–1.49] | |
| Indigenous | 1.11 | [0.46–2.73] | |
| Unknown | 1.17 | [0.91–1.49] | |
| Education (ref = none) | Primary incomplete | 0.90 | [0.68–1.19] |
| Primary complete | 0.98 | [0.73–1.33] | |
| Secondary incomplete | 0.86 | [0.65–1.14] | |
| Secondary complete | 0.68 | [0.49–0.93] | |
| High school incomplete | 0.57 | [0.41–0.79] | |
| High school complete | 0.41 | [0.29–0.57] | |
| College incomplete | 0.29 | [0.14–0.56] | |
| College complete | 0.33 | [0.19–0.59] | |
| Unknown | 0.78 | [0.60–1.02] | |
| Not applicable (age < 7) | 0.75 | [0.34–1.67] | |
| Pregnant at diagnosis | 0.47 | [0.19–1.19] | |
| HIV test (ref = negative) | Positive | 2.10 | [1.72–2.55] |
| Not done | 1.93 | [1.71–2.16] | |
| Alcohol use (ref = No) | Yes | 1.76 | [1.53–2.03] |
| Unknown | 1.26 | [0.91–1.74] | |
| Diabetes (ref = No) | Yes | 0.59 | [0.46–0.77] |
| Unknown | 0.84 | [0.62–1.13] | |
| Any other aggravating condition (ref = No) | Yes | 1.39 | [1.20–1.61] |
| Unknown | 0.89 | [0.73–1.09] | |
| TB type (ref = Pulmonary) | Extrapulmonary | 0.74 | [0.62–0.87] |
| Both | 0.74 | [0.51–1.09] | |
| First baseline culture (ref = Negative) | Positive | 1.63 | [1.15–2.31] |
| Unknown | 1.62 | [1.20–2.19] | |
| DOT recommended (ref = No) | Yes | 1.43 | [1.23–1.65] |
| Unknown | 1.10 | [0.75–1.62] | |
| DOT throughout treatment (ref = No) | Yes | 0.69 | [0.60–0.79] |
| Unknown | 0.74 | [0.50–1.12] | |
| Bairro-level literacy rate | 10 percentage points | 0.83 | [0.61–1.13] |
| Bairro-level sewerage coverage | 10 percentage points | 1.05 | [1.02–1.09] |
| Bairro-level: homicide rate | One standard deviation | 0.92 | [0.85–1.00] |
| District (ref = District I) | II | 1.18 | [0.90–1.56] |
| III | 0.69 | [0.52–0.92] | |
| IV | 0.91 | [0.66–1.24] | |
| V | 0.99 | [0.73–1.34] | |
| VI | 1.04 | [0.77–1.40] | |
| Constant | 0.13 | [0.01–2.20] | |
| Bairro-level random effect variance | 0.07 | [0.03–0.12] |
This model contained only covariates significant in bivariate models (results of bivariate models are shown in Additional file 3: Table S2). For bairro-level variables, odds ratios are those associated with a change in the covariate of the indicated amount