Skip to main content

Table 5 Bivariable and multivariable Gompertz regression analysis for predictors of time to lost to follow-up among DR-TB patients, University of Gondar Comprehensive Specialized Hospital, September 2010–December 2017, (N = 332)

From: Incidence and predictors of lost to follow-up among drug-resistant tuberculosis patients at University of Gondar Comprehensive Specialized Hospital, Northwest Ethiopia: a retrospective follow-up study

Variable Eventa Censoredb CHR (95%CI) AHR (95%CI)
Median Age (IQR) 29 (25, 45) 30 (22, 40) 1.02 (1.00, 1.05) 1.03 (0.99, 1.05)
Sex
 Male 20 186 1  
 Female 16 110 1.34 (0.69, 2.58)  
Body mass index (kg/m2)
 < 18.5 25 196 1.22 (0.55 2.71)  
 > =18.50 8 78 1  
Housing Conditionc
 Homeless 10 36 2.20 (1.06, 4.57) 2.51 (1.15, 5.45)
 Having Home 26 222 1 1
Alcohol drink
 No 30 249 1  
 Yes 5 45 0.91 (0.35, 2.35)  
Co-morbidity
 No 30 247 1  
 Yes 6 49 1.12 (0.46, 2.68)  
Year of treatment initiationc
 < =2012 6 92 1 1
 2013–2014 22 106 3.08 (1.25, 7.59) 3.25 (1.30, 8.13)
 > =2015 8 98 1.42 (0.49, 4.11) 2.07 (0.71, 5.99)
Base line sputum smear result
 Negative 6 42 1  
 Positive 27 231 0.77 (0.32, 1.87)  
Residency
 Urban 19 150 1  
 Rural 17 146 0.97 (0.51, 1.86)  
Cigarette smoking
 No 32 245 1  
 Yes 3 51 0.47 (0.14, 1.53) 0.40 (0.12, 1.35)
HIV co-infection
 No 27 219 1  
 Yes 9 75 1.05 (0.49, 2.24)  
Number of previous TB treatment
 < 2 11 101 1  
 > =2 25 193 1.10 (0.54, 2.24)  
Occupation
 Employed 26 183 1  
 Unemployed 9 108 0.62 (0.29, 1.32) 0.74 (0.32, 1.68)
  1. LR chi2 (6) = 17.14 probability > chi2 = 0.0088
  2. CHR Crude Hazard ratio, AHR Adjusted Hazard Ratio, CI Confidence Interval
  3. Eventa in this study was patients lost during follow-up; Censoredb was either cured, completed, death, transfer out, still on treatment
  4. The shape parameter gamma was found to be −0.0811245 (95% CI: −0.1407007, −0.0215482) which is negative. This indicates that the hazard of lost to follow-up decreases exponentially with time
  5. cSignificantly associated variables in both bivariable and multivariable model